[HN Gopher] Stephen Fry - AI: A Means to an End or a Means to Ou...
       ___________________________________________________________________
        
       Stephen Fry - AI: A Means to an End or a Means to Our End?
        
       Author : lordleft
       Score  : 172 points
       Date   : 2024-09-16 14:04 UTC (8 hours ago)
        
 (HTM) web link (stephenfry.substack.com)
 (TXT) w3m dump (stephenfry.substack.com)
        
       | egnehots wrote:
       | well it's human nature to struggle with collective action when
       | the risks are unclear, vague and not shared. stakeholders are
       | juggling immediate, tangible concerns, like climate change,
       | economic stability, and political issues, making it tough to
       | justify moving AI up the priority list.
        
         | ben_w wrote:
         | Extra fun: AI can make a big (+ and -) difference to climate
         | change[0], mess with the economy[1], and get used as a tool to
         | sow political chaos[2].
         | 
         | But sure, humans are necessarily very myopic, it's necessary
         | that we ignore 98% of the issues in the world or we wouldn't be
         | able to even function.
         | 
         | [0] High power use, can help roll-out renewables and storage
         | 
         | [1] What happens when those humanoid robots we see demos of,
         | get good enough to replace all the staff in the factories where
         | they get made? And the rest of their supply chain?
         | 
         | [2] Imagine if the pizzagate conspiracy theorists had had
         | access to an un-censored sound-and-video GenAI tool
        
           | janalsncm wrote:
           | I would like to learn more about what AI can do specifically
           | to solve the climate crisis.
           | 
           | My guess is that a lot of the actual lift would come from
           | industrial automation to create cheaper green products. I
           | guess that is "AI" in some sense.
           | 
           | But if we are building solar panels, the R&D budget should be
           | put towards streamlining the build process. Figure out how to
           | commoditize solar panels so that oil is too expensive.
           | 
           | Building huge "foundation" models like I see huge AI labs
           | doing is a bit like building better visualizations of an
           | impending asteroid impact. It's not really what we need right
           | now.
        
             | ben_w wrote:
             | > I would like to learn more about what AI can do
             | specifically to solve the climate crisis.
             | 
             | Robots that _install_ PV. I think there 's even an
             | YCombinator startup doing exactly that? Needs a higher
             | degree of AI to function outdoors than in a nice fixed
             | factory setting.
             | 
             | (Especially if they also drive the trucks containing the
             | PV, but who knows how long we'll have to wait for _that_
             | AI...)
             | 
             | > Figure out how to commoditize solar panels so that oil is
             | too expensive.
             | 
             | Good news: it already is :D
             | 
             | > Building huge "foundation" models like I see huge AI labs
             | doing is a bit like building better visualizations of an
             | impending asteroid impact. It's not really what we need
             | right now.
             | 
             | Mm. These things give increasing levels of
             | generalisability: the biggest weakness of previous models,
             | was that you could train them to learn one thing very well,
             | and they'd suck at anything else.
             | 
             | In one sense that's still true of the new models, it's just
             | that they're being trained to build a world model from most
             | written text and most pictures and most videos, so they
             | have a very broad range of things they're OK at -- much
             | less likely to be confused by a rattlesnake they mistake
             | for a powerline, for example.
        
       | throwanem wrote:
       | Anyone find a recording? Most things I prefer in prose, but Fry's
       | words suffer badly without his particular delivery.
        
         | megak1d wrote:
         | I share this. I quickly grow irritated by Fry's elaborate
         | language when in written form.
        
           | throwanem wrote:
           | Like any skillful rhetorician, he constructs with an eye
           | toward the intended medium. When that's prose, it reads a lot
           | more smoothly than here.
        
             | bloak wrote:
             | That's right: his books are very readable.
             | 
             | (With a couple of his novels I felt that if he had made
             | them a bit less entertaining and a bit more pretentious and
             | if he weren't already famous as an actor and comedian then
             | probably they would be taken seriously as "literary
             | fiction". I'm thinking of "The hippopotamus" in
             | particular.)
        
         | tagami wrote:
         | I too would like a recording. Until then, soak in each line,
         | seek the cadence of delivery and imagine, as he asks us to do
         | in his lecture,
         | 
         | "...before I go any further I'd like to paint some pictures.
         | Think of them as tableaux vivants played onto a screen at the
         | back of your mind."
        
           | throwanem wrote:
           | I know how to model a speaker's delivery; I find I lack the
           | skill to model his.
        
         | oever wrote:
         | There's a text2speech with a Stephen Fry option. I do not have
         | it, but it must exist.
        
           | throwanem wrote:
           | As well suggest that, hungry for a meal, I visit the nearest
           | toilet.
        
         | dgeiser13 wrote:
         | The speech was given 4 days ago. Video may still be released.
        
       | padjo wrote:
       | The greatest technology humans have invented and the one that
       | currently needs most investment is bureaucratic collective
       | action.
        
         | abeppu wrote:
         | Can you expand on the "greatest technology" half of your
         | comment?
         | 
         | I can see plenty of clear reasons why bureaucratic collective
         | action is in need of investment, but I'm predisposed to suspect
         | that's because currently it's pretty bad.
        
           | throwanem wrote:
           | I believe the claim is that we have no more powerful
           | technology for the coordination of human action at very large
           | scales. If so, I wouldn't disagree.
        
             | padjo wrote:
             | Yeah pretty much. If an extinction level asteroid pops up
             | tomorrow it's not going to be rugged individualism and
             | entrepreneurial spirit that saves the day, it's going to be
             | the ability to efficiently coordinate millions of people to
             | achieve a goal.
        
               | chasd00 wrote:
               | in that particular scenario i'm sure SpaceX will be a big
               | part of the solution so... probably a mix of both. A
               | weapon of mass destruction developed by a nation and a
               | space vehicle developed privately by an individual tired
               | of the status quo.
        
               | throwanem wrote:
               | Developed "privately" only in the rather facile sense
               | this can be said of anything done with the necessary
               | support of massive federal funding, yes. It's bureaucracy
               | that allocates and disburses that money.
        
           | padjo wrote:
           | I suppose by greatest I mean: increasingly important as we
           | scale.
        
         | hansonkd wrote:
         | It's astonishing to me that people do not see the correlation
         | between AI and as you call it "bureaucratic collective action".
         | 
         | I have considered corporations, nation states, etc as forms of
         | artificial intelligence that long predates computers. Just
         | because the intelligence is collective doesn't make it less
         | artificial. and just because the "code" that makes up these
         | collective intelligences are written as laws and governance
         | documents doesn't change the fact that it is an agent that can
         | make "decisions".
         | 
         | I suppose it is hard for people to feel a connection with a
         | group or org as intelligence opposed to an personified
         | "individual" like ChatGPT. Kind of like how it is hard to see
         | an anthill as an organism when we are so much more related to
         | the ant. However just because its hard for us to relate to an
         | anthill, it doesn't make it any less of an organism that
         | behaves like an individual.
         | 
         | In order to advance AI, it is imperative to study and
         | understand how our current "super intelligences" of nation-
         | states and mega-corporations behave and how they interact with
         | others and govern themselves.
        
           | cmrdporcupine wrote:
           | We're swimming in a dominant ideology which barely recognizes
           | the existence of "society" (the Margaret Thatcher "there's no
           | such thing as society" vibe) and frames most of reality in
           | terms of individualism, individuals, and the self.
           | 
           | It should come as no surprise how little understanding there
           | is of the social framing of technology and tools.
           | 
           | From chipping flint to writing code, it's a huge mass of
           | inherited and circulated technique-knowledge.
        
           | kristiandupont wrote:
           | I have been fascinated by this thought experiment for many
           | years as well.
           | 
           | >I suppose it is hard for people to feel a connection with a
           | group or org as intelligence
           | 
           | In theory (one theory, at least), we should have as much
           | chance to understand what such an intelligence thinks as a
           | single brain cell would have of understanding what goes on in
           | a human mind.
        
         | xyzzy123 wrote:
         | What worries me is that in many industries (health & education
         | particularly) there doesn't seem to be any actual limit to
         | admin demand.
         | 
         | Like; admin demands increase until the system _barely_
         | functions and compliance is just barely possible.
         | 
         | BUT administrators would always like more data, controls and
         | process.
         | 
         | So one possible outcome of AI is that computers magically pick
         | up our admin burdens, freeing us to do our jobs in peace.
         | 
         | The _other_ possible outcome is that AI increases our capacity
         | for admin and therefore generates _way more admin_.
         | 
         | A particular risk is where AI generates asymmetric leverage for
         | administrators.
         | 
         | An example of this would be admin requiring lengthy forms to do
         | such-and-such. If a human administrator no longer needs to
         | spend their time validating or checking the forms, there's less
         | pushback on making the forms longer or producing even more of
         | them.
        
           | hansonkd wrote:
           | > AI increases the capacity for admin and therefore generates
           | way more admin
           | 
           | This is the most likely scenario.
           | 
           | The original sales pitch for Computers were originally to
           | free doctors and nurses and admins by making things easier.
           | But instead the administrative work exponentially grew.
        
             | _wire_ wrote:
             | Work goes to where it can get done.
        
         | throwaway13337 wrote:
         | I agree that effective human organization is, by far, our
         | largest problem today. All of our contemporary problems are, at
         | their core, organizational problems.
         | 
         | However, we actually lack an understanding of how large systems
         | are organized.
         | 
         | In the human body, each cell performs its task without a
         | centralized authority but as an individual agent. The cells are
         | selfish and just try to survive. But emergent in this behavior
         | is a human that thinks itself the master of all the cells.
         | 
         | Michael Levin has done surprising things that fall from
         | studying cells of animal bodies and how they can act
         | independently.
         | 
         | We really need a science of agent based systems. I don't know
         | of something that captures it entirely.
         | 
         | What I do know is that top-down hierarchical organizations are
         | fragile and will corrupt with time. Agent based systems are the
         | only way nature will allow to progress much.
        
           | WillAdams wrote:
           | >We really need a science of agent based systems. I don't
           | know of something that captures it entirely.
           | 
           | Wasn't that Stephen Wolfram's _A New Kind of Science_?
        
         | cmrdporcupine wrote:
         | Exactly this.
         | 
         | We need to heavily (re)invest in processes, tools, ethics,
         | technology that boost human social power and our ability to
         | work together.
         | 
         | There are 8 billion people on the planet. So much potential.
        
           | SoftTalker wrote:
           | Every attempt to organize people bureaucratically ends in
           | failure or war. We don't seem to learn this lesson as we keep
           | trying.
           | 
           | No bureaucracy will ever care for individual goals, needs,
           | desires as much as you will for yourself and your family.
           | 
           | We need some rules of the road and limits. History also
           | proves that. But it's a fools errand to think that we can
           | organize 8 billion people to work together, other than within
           | some fairly loose rules that prohibit undesirable extremes of
           | behavior.
        
             | si1entstill wrote:
             | I think both can be true. Most liberal western economies
             | are highly individualistic in terms of form and law, but
             | large firms naturally come into being and thrive. I guess
             | one could argue that this isn't due to the need for
             | "bureaucracy" but instead "labor" and that the bureaucracy
             | is an unavoidable side-effect, but there is undeniably
             | large-scale organization.
             | 
             | Context seems like the most interesting thing to consider.
             | I suspect the attitude and outlook of the individuals
             | toward the organization is the key component (and their
             | capacity to undermine the organization in some manner). All
             | of this is to say that I don't believe organization is
             | inherently against the human sentiment, it just needs to be
             | seen as justified, sensible, and a net-positive to those
             | involved.
        
               | SoftTalker wrote:
               | I would agree, a bureaucracy that one chooses to enter
               | (employment) as a means to earn income is different from
               | living under an all-controlling government that dictates
               | most aspects of your life. Still, people in highly
               | bureaucratic jobs tend to burn out and quit, or endure a
               | certain level of misery.
        
       | rqtwteye wrote:
       | AI will just be another accelerator for the trend that society is
       | not run for the benefit of all people but for the benefit of a
       | few people who are making lots of money. At some point we'll have
       | to make a decision whether this trend of more and more power
       | accumulating at the top can continue or whether we make rules
       | that allow everybody to benefit from technological progress.
       | 
       | Judging from history it's unlikely that the wealthy and powerful
       | will give up anything voluntarily.
        
         | zooq_ai wrote:
         | Incredibly dumb argument. Technological advancements like the
         | Internet, Google Search, YouTube, Social Media, Smartphones
         | have benefitted the masses -- the poorest person in Bangladesh
         | have the same access to these technologies as the President of
         | US or a Billionaire.
         | 
         | ChatGPT is literally available to everyone, but yet it takes an
         | incredible amount of brainwashing to claim these technologies
         | are not to the benefit of the people. These are also claims
         | made by highly privileged, highly paid, entitled coastal
         | elites.
         | 
         | Go talk to a poor bangladeshi farmer and they will have a
         | different story to tell
        
           | cmrdporcupine wrote:
           | > Incredibly dumb argument.
           | 
           | HN guidelines:
           | 
           |  _Be kind. Don 't be snarky. Converse curiously; don't cross-
           | examine. Edit out swipes.
           | 
           | When disagreeing, please reply to the argument instead of
           | calling names. "That is idiotic; 1 + 1 is 2, not 3" can be
           | shortened to "1 + 1 is 2, not 3."
           | 
           | Please don't fulminate. Please don't sneer, including at the
           | rest of the community.
           | 
           | Please respond to the strongest plausible interpretation of
           | what someone says, not a weaker one that's easier to
           | criticize. Assume good faith._
        
           | keybored wrote:
           | The Internet was made by the United States state sector and
           | then handed over to commercial interests in the 90's or so.
           | Exactly what GP said.
           | 
           | Your argument doesn't prove GP's wrong. It's like saying that
           | the king's feast benefits the peasants because there was
           | plenty of food that was thrown away that they could get their
           | hands on.
        
           | blibble wrote:
           | personally I wish my kids were growing up in the early 90s
           | before any of that was invented (or at least mass market)
           | 
           | library books, playing outside, maybe a bit of nintendo, pogs
           | 
           | over literally an endless source vomit from a screen with
           | social media, youtube, instagram, reddit, facebook and
           | twitter, now with the added benefit of being infested with AI
           | slop
           | 
           | I'd take the 90s any day
        
             | cmrdporcupine wrote:
             | Gotta love it ... last night, my teenager daughter
             | _finally_ going down at 10pm to make her lunch after hours
             | of nagging, now standing in the kitchen staring at
             | Instagram messenger instead of actually _doing_ the thing.
             | Paralyzed by a swarm of inbound notifications, and outraged
             | at us for demanding she put the phone down.
             | 
             | The 90s had problems too (I was a teen in the early 90s)
             | tho. Just different kinds.
        
           | rqtwteye wrote:
           | The Bangladeshi farmer is still poor, or not? And he can't
           | raise his prices because the people at the top don't want to
           | lose even the smallest bit of their profits. He has access to
           | ChatGPT but not as someone who benefits from it economically.
        
             | zooq_ai wrote:
             | Their standard of living has improved dramatically
             | precisely because of access to technology
        
         | add-sub-mul-div wrote:
         | Exactly, it's mostly that I don't trust any invention in the
         | hands of 2020s era capitalism. In another time and place, I
         | could be excited about AI.
        
       | fidotron wrote:
       | I tend to think the car comparison is cause for optimism. Prior
       | to mass car ownership you would assume that such things in the
       | general population would be enormously more dangerous than they
       | have proven to be.
       | 
       | We live with the dangerous aspects of cars as the utility they
       | provide is so high. The same will prove true of AI.
       | 
       | Humanity also has a huge problem with the mortality of the
       | species as a whole. At some point we will be extinct. Will we
       | have evolved into something else before this? Or will we choose
       | to replace ourselves with machines? It is a valid choice to make
       | in the face of changing circumstances which will occur no matter
       | what we do.
        
         | nerdponx wrote:
         | > We live with the dangerous aspects of cars as the utility
         | they provide is so high. The same will prove true of AI.
         | 
         | And along the way we have arranged our physical surroundings,
         | legal system, and cultural mores in such a way that cars are
         | _significantly more dangerous than they need to be_.
        
         | Ancalagon wrote:
         | Self-replicating, super intelligent pieces of software are a
         | little different than cars though.
        
         | Qem wrote:
         | > We live with the dangerous aspects of cars as the utility
         | they provide is so high. The same will prove true of AI.
         | 
         | I'm not so sure on the utility of cars. If you replace "car"
         | with "internal combustion engines", I would agree 100%. But
         | most social utility of ICEs is in cargo transport and utility
         | vehicles (ambulances, firefighter trucks...). Mass use of small
         | cars solve mostly the problems they create themselves. For
         | example, car-centric cities suffer from sprawl, so everything
         | is far away and you can't even buy bread without owning a car,
         | because walking to the bakery would take too much time.
        
         | layer8 wrote:
         | > Or will we choose to replace ourselves with machines?
         | 
         | Bio-engineering seems much more likely.
        
         | bugglebeetle wrote:
         | > Prior to mass car ownership you would assume that such things
         | in the general population would be enormously more dangerous
         | than they have proven to be.
         | 
         | An assumption that was correct and involved large-scale
         | propaganda efforts to blame pedestrians for the harm inflicted
         | upon them, including blaming children for their own deaths.
         | U.S. car accident fatalities are also still to this day at
         | developing country levels and every actually developed country
         | largely has public transportation infrastructure. And this is
         | before we get into the blight that is American urban design,
         | evolving to serve the needs of this mass car ownership, with
         | all its down stream environmental harms.
        
         | Devasta wrote:
         | Mass car ownership has been a disaster; the entirety of society
         | was reorganized around them causing urban sprawl and long
         | commutes; nevermind the myriad conflicts around the world
         | caused in the pursuit of oil.
         | 
         | The road deaths are an incredibly minor party of the harm
         | caused by cars.
        
         | mmmore wrote:
         | You're missing the point of the car section. The point isn't
         | that cars were a dangerous technology.
         | 
         | The point is that just as the car replaced the horse, programs
         | on computers are set to replace the human mind. Right now, we
         | see only the beginnings of the trend; inventions which look
         | interesting but limited. But eventually, and perhaps sooner
         | than many believe, the replacement of human cognition will
         | radically transform the world in unprecedented ways.
        
           | fidotron wrote:
           | And you missed my point
           | 
           | > The point is that just as the car replaced the horse,
           | programs on computers are set to replace the human mind
           | 
           | My point is this is not inherently a bad thing. At some point
           | either the human mind is overtaken by another species or AI.
           | We must learn to accept this inevitability.
        
             | mmmore wrote:
             | Right now we are not so much choosing to replace ourselves
             | with machines as stumbling into it through hubris and
             | ignorance. If the argument is we should accept our demise
             | to make way for something better, we should be more
             | explicit and deliberate about that. Most people are not
             | imagining that future. And there's not convincing reasons
             | to believe our current path will lead to something
             | "better".
             | 
             | I think you're right that I misread your first two
             | paragraphs though. Sorry.
        
             | rurp wrote:
             | That sounds pretty bad to me. "At some point" would
             | probably be millions of years in the future, but you're
             | ready to force the issue this decade. If you asked everyone
             | in a developed country what they think about completely
             | upending life as they know it and replacing it with
             | something unknown; that might be better, or might be
             | horrifically worse, we'll just have to see what happens;
             | most people would be strongly opposed to that proposition.
        
         | lm28469 wrote:
         | 200 years of internal combustion engines and we're slowly
         | waking up to the fact we might have fucked up our atmosphere
         | beyond repair, hn take: cars are amazing with only slight
         | drawbacks
         | 
         | Humanity will never learn
        
       | ilaksh wrote:
       | The short to medium term concerns mostly come down to human
       | problems. AI and robotics have a multiplicative affect like other
       | technologies, but the problems still originate in the primate
       | nature of humans.
       | 
       | The long term issue that many people don't seem willing to
       | mention out loud is that we will eventually make humanity
       | obsolete and robots will literally take control.
       | 
       | The only real solution to the threat of militarization of AI and
       | robotics might be to create a more unified global government and
       | culture. The first challenge is for people to see that as a
       | worthy goal.
       | 
       | Sometimes I think most of our problems come down to not being on
       | the same page. And I wonder if somehow in the future we gradually
       | become a tiny bit like The Borg.
       | 
       | So maybe we are headed towards a "meta-system transition" where
       | we have some kind of direct links between groups of AI and humans
       | that combine to form a more intelligent and effective organism in
       | some way.
       | 
       | I guess I just came up with a Black Mirror episode concept.
        
         | moffkalast wrote:
         | Eh I don't think humans will really be a viable concept in a
         | few hundred years if this keeps up. We'll simply be too fragile
         | and dumb for anything.
         | 
         | I see this as a good thing though, the only thing that
         | separates us from everything else in the universe is our
         | inteligence, in some sense we _are_ that inteligence and this
         | is the next evolutionary step it needs to finally break out
         | from the confines of a tiny 20W brain that 's been pushed as
         | far as it practically can be.
         | 
         | When you no longer need to consume energy constantly unless you
         | die, are functionally immortal and can hibernate for a thousand
         | years, you eat only sunlight, don't get cancer from radiation
         | or bone loss from zero gravity... the galaxy becomes your
         | oyster.
         | 
         | Evolution, Morpheus, evolution. Like the dinosaur ;)
        
           | polotics wrote:
           | We are extremely energy-efficient. My lunch was one fruit,
           | had no breakfast, a bit hungry but can still type in the
           | sequence of word you're reading right now. Compare this with
           | the wattage of OpenAI, from what I can find 2.9 Watt-hours
           | for one query, so 2.4 kcal, ie. the whole daily-maintenance
           | food need for an average man...
        
             | teraflop wrote:
             | You're off by a factor of 1,000. The "calories" used for
             | nutritional purposes are actually kilocalories, and a
             | typical adult diet is about 2,000 kcal per day.
        
             | isoprophlex wrote:
             | I think you mean 2.4 Mcal, or 2400 kcal, to keep a man
             | alive for a day...
        
             | naming_the_user wrote:
             | You're missing a few orders of magnitude here - the daily
             | food intake of a man is 2400kcal, not 2.4kcal.
             | 
             | We are still much more efficient mind, 2400kcal is approx
             | 2.7kWh so about 1000 GPT queries if your 2.9Wh figure is
             | correct (wow!)
        
             | simonbw wrote:
             | I think your math is off by a bit. A Calorie and a watt-
             | hour are approximately equal. Humans need about 2000
             | Calories per day. So 2.9 watt-hours is more like the amount
             | of energy a human needs in a minute.
             | 
             | I think the confusion might come from the Calorie/calorie
             | difference, i.e. 1 Calorie = 1000 calories = 1 kcal.
        
             | moffkalast wrote:
             | Efficiency doesn't matter if you have spare energy to use
             | and can't. We have all the food we can eat and we remain
             | dumb fucks.
             | 
             | With nuclear power and a way to scale up we could build a
             | gigawatt tier brain if we wanted to.
        
             | sva_ wrote:
             | Other than the cal/kcal issue, I'm also curious if this
             | comparison really pans out. If you had a fusion power
             | plant, you could power processors with it, but how does raw
             | power translate into human food? Even if you had some
             | formula like Soylent Green, many humans would probably not
             | cooperate after a while
        
           | javaunsafe2019 wrote:
           | I truly believe that someone writing such a comment is deeply
           | unhappy with his life and can't value humanity as it should
           | be
        
             | nakeru wrote:
             | Not OP of your answer, but I'm happy with my life. Still, I
             | can't value humanity much when we're living the fastest
             | extinction even ever because of, well, humanity. And we
             | won't fix it. I'm happy because none of these silly
             | "issues" like AI will matter one iota in a couple of
             | decades.
        
             | justupvoting wrote:
             | "A man's at odds to know his mind cause his mind is aught
             | he has to know it with. He can know his heart, but he dont
             | want to. Rightly so. Best not to look in there. It aint the
             | heart of a creature that is bound in the way that God has
             | set for it. You can find meanness in the least of
             | creatures, but when God made man the devil was at his
             | elbow. A creature that can do anything. Make a machine. And
             | a machine to make the machine. And evil that can run itself
             | a thousand years, no need to tend it."
        
             | Semaphor wrote:
             | And who is to say how it should be? You? And what does it
             | even mean, is the great thing about humanity the last few
             | decades? There last few years? Surely humanity as recent as
             | a few hundred years ago was not already as it should be?
        
             | moffkalast wrote:
             | We are so hardwired for human exceptionalism that we can't
             | even entertain the thought of there ever being something
             | better that replaces us, which is what's has occurred for
             | 99.9% of species so far and makes it a statistical
             | certainty. It's hubris manifest to think that we'll live
             | unchanged until the heat death of the universe.
        
             | drdaeman wrote:
             | > can't value humanity as it should be
             | 
             | It's hard to see a value in humanity when most of the globe
             | is a giant dumpster fire. We currently have:
             | 
             | 1. United States, a corporate-owned two-party clusterfuck,
             | potentially on a brink of failing into an autocracy.
             | Because the world is in post-ironic phase, this is all
             | turned into a TV show for the mass entertainment. 2.
             | Russia, a kleptocracy turned into an authoritarian police
             | state run by an ex-KGB ex-Stasi nazi. Always messed up,
             | badly, waging wars as a hobby. Doing their best to fuck up
             | things abroad so they aren't alone, and fuck things up
             | domestically so people are more obedient (dictatorship 101:
             | people always behave "better", when their base needs are
             | just barely satisfied so their focus is all on survival).
             | 3. Middle East, where all things are so fucked up beyond
             | recognition I don't even... 4. China, another anti-utopia
             | lover's dream, with obligatory facade of "harmony" and
             | "prosperity". Don't know much about their internal affairs,
             | to be honest, but don't see any good coming out of it. 5.
             | Africa isn't doing any better, we just hear and care less
             | about their mess here because we have our own mess at home.
             | 6. A bunch of ticking bombs (pollution, global warming,
             | overpopulation crises) so we won't get bored by the old
             | wars alone.
             | 
             | And all of this stems up from the fundamental issue of
             | human nature. Innate high susceptibility to things that
             | _sound_ nice (aka populism) leading up to blind faith,
             | constantly abused by high achiever sociopaths, toppling
             | societies left and right despite all the checks and
             | balances, wreaking havoc. We know about logical fallacies,
             | but still - en masse /at scale - fail at those, badly. We
             | only realize the writings were on the wall all the time,
             | when we're already stuck deep in the mess we've made
             | ourselves.
             | 
             | The worst part of it? There seems to be no solution. At
             | least not that I've ever heard of. Pandemics showed us that
             | a global challenge doesn't reunite people but rather gets
             | abused to divides them further, so even evil aliens'
             | arrival or GAI-gone-Skynet most likely won't change
             | anything.
             | 
             | Our best chances are achieving global post-scarcity and/or
             | solving death (so, yeah, very realistic goals straight
             | outta sci-fi books /s), but not even those are guaranteed
             | to work - just a slim chance that maybe if those major
             | issues are out of the picture something will change and
             | maybe there will be a slim chance it won't get abused for
             | worse. Transhumanism might be another chance (if our monkey
             | brains are a problem, maybe we can patch them up or
             | possibly hop over to something else) but I have my doubts.
             | 
             | What's to value? That humans have coping mechanisms that
             | lets them just ignore all the crap going around and live
             | happily in their imaginary safety bubbles, until it's their
             | time and the frog is finally notices it's being boiled?
             | 
             | (Please don't get me wrong, zero hate here - except the
             | hate for all the abusers and hate for all the mess
             | happening. I love humans, but I'm deeply saddened that
             | something is really wrong with us that we just can't get
             | our shit straight for multiple millennia, stepping on the
             | same rake over and over.)
             | 
             | Yea, I guess I'm deeply unhappy.
        
           | noah_buddy wrote:
           | Suppose you are right. From a genetic perspective, the most
           | logical course of hypothetical action is to immediately start
           | smashing computers, learning how to blow up terminators, and
           | to oppose and undermine anyone that thinks like you.
           | 
           | Now, I actually do not believe you are right and I am more
           | inclined to believe that artificial intelligence stemming
           | from humanity would see the value in human ideals. Certain
           | things like love, morality, and humility have been deeply
           | held ideals for millennia. I would suppose that an AI would,
           | to some extent, value these things too. Think more "The Day
           | the Earth Stood Still" than Terminator.
           | 
           | This is all conjecture, but one thing I would suggest to you:
           | if you see the natural evolution of your own species to be
           | the removal of itself, you probably should be nowhere near AI
           | or alignment. What you suggest is akin to world-wide
           | genocide. Ironically, in my estimation, you are less aware of
           | humanity than AI will be.
        
             | moffkalast wrote:
             | I think you misunderstand, that's exactly what I'm saying.
             | The AI we make will replace us because it will functionally
             | be us in every way. Even unaligned LLM base models of the
             | current gen embody our values and ideals better than we
             | ever could.
             | 
             | The way I imagine it you could talk to one of these things
             | and it would appear human, except it would know everything
             | that is known, be a master of everything that can be done,
             | and could fax itself to Mars over a laser in 20 minutes.
             | Being software is such a huge advantage, since you can just
             | swap your body for whatever you need to do or wherever you
             | need to be.
             | 
             | I'm sure humans will still be around long into the future,
             | but they won't be doing anything notable because we won't
             | be capable of thinking on a cutting edge level. Not
             | entirely unlike what a cat or a dog is to us now. The
             | inteligence disparity will probably be similar anyway.
        
           | SirHumphrey wrote:
           | I think there exists a point, where AI stops making sense in
           | a cost equation. Currently we are well accustomed that
           | machines are cheap and humans are expensive, and anything
           | that can be done by machine instead of by a human means a
           | cost reduction. At some point that probably stops, because
           | the amount of compute you need to run larger and larger
           | models means that humans can become cost competitive. It's a
           | bit like in manufacturing, where cost of robotization is
           | sometimes more than just the cost of employing humans.
           | 
           | And in the end, this long term predictions based on where we
           | are now and how fast we are moving are always a bit iffy.
           | Nobody really know what challenges must current AI overcome
           | to become AGI (if that even means anything).
        
             | moffkalast wrote:
             | Yeah it's just what I suppose might be the inevitable
             | outcome for the very far future, assuming incremental
             | improvements in all forms over a long time.
             | 
             | For the short term it's honestly way harder to say, there
             | could easily be another 50 year winter in between. But
             | machines improve, and humans don't, so as long as non-zero
             | progress is made it's a virtual certainty.
        
         | sva_ wrote:
         | A global government usually implies a centralization of power,
         | which corrupts, and probably isn't a solution to much of
         | anything
        
           | tstrimple wrote:
           | I don't think this assertion stands on its own. There is more
           | local corruption in the US than at the federal level for
           | example. Way fewer eyes on your activities. Practically no
           | one is going to notice you getting kickbacks for approving
           | zoning changes for a business or other sorts of "low stakes"
           | corruption.
        
         | semi-extrinsic wrote:
         | I've said it before, and I'll say it again: there is absolutely
         | zero evidence indicating that we are any closer to AGI than
         | what James Watt was to realizing nuclear fusion.
         | 
         | There are extremely well researched worms with only a few
         | hundred neurons which we cannot yet simulate with anything
         | resembling accuracy. How can that statement be true, if LLMs
         | are close to delivering superhuman intelligence?
        
           | Jabrov wrote:
           | Why do we have to simulate real neurons to achieve
           | intelligence? We don't simulate a bird's bones and muscles to
           | achieve flight.
           | 
           | LLMs are a huge step forward. Sure, they might not be the
           | thing to ultimately deliver superhuman intelligence. But it's
           | unfair to say that we're not any closer at all.
        
             | blastro wrote:
             | Not trying to be pedantic, but:
             | 
             | > We don't simulate a bird's bones and muscles
             | 
             | Isn't really true, is it?
             | 
             | We simulate the bones with an airframe, and we simulate the
             | muscles with a prop/jet. The wings are similarly simulated.
        
               | exe34 wrote:
               | yes, the wings flap. the airframes are made of bone.
        
               | suriya-ganesh wrote:
               | The type of propulsion is not, though. Nothing in nature
               | has 360deg rotatory motion. Airplane is not an exact
               | simulation of a bird. Only an approximation
        
               | dboreham wrote:
               | https://phys.org/news/2018-04-bacterial-propeller.html
        
               | r2_pilot wrote:
               | A more suitable word would be that we "emulate" the
               | bones, etc., rather than simulate, which to me implies
               | that you are testing against a replica of a thing. But I
               | don't really think emulate works here because they're not
               | especially analogous; these are fundamentally different
               | approaches to flight so any form following the function
               | would be incidental at best.
        
               | Jabrov wrote:
               | I was thinking OP meant more simulate as in
               | https://en.wikipedia.org/wiki/OpenWorm ie. a complete
               | bottom-up simulation at the cellular level
               | 
               | By your definition of simulate, I think artificial neural
               | networks are absolutely on their way to simulate
               | intelligence
        
             | splwjs wrote:
             | We've had markov chain generators for a while, having
             | enough computing power to grant them the power to
             | regurgitate wikipedia reddit and stackoverflow content is
             | not "a huge step towards agi"
        
               | Jabrov wrote:
               | I disagree.
               | 
               | It's true that Markov chain generators have existed for
               | years. But historically their output was usually just
               | this cute thing that gave you a chuckle; they were
               | seldomly as useful in a general sense like LLMs currently
               | are. I think that the increase you mention in compute
               | power and data is itself a huge step forward.
               | 
               | But also transformers have been super important.
               | Transformer-based LLMs are orders of magnitude more
               | powerful, smarter, trained on more data, etc. than
               | previous types of models because of how they can scale.
               | The attention mechanism also allows them to pay attention
               | to way more of the input, not just the few preceding
               | tokens.
        
               | dijit wrote:
               | I think you missed OPs point.
               | 
               | If you want something _useful_ , then we're getting
               | closer.
               | 
               |  _AGI_ is something specific, as a requisite, it must
               | _understand_ what is being asked, and what we have now is
               | a puppet show that makes _us humans_ think that the
               | machine is thinking, similar to Markov chains.
               | 
               | There is absolutely some utility in this- but it's about
               | as close to AGI as the horse-cart is to commercial
               | aircraft.
               | 
               | Some AI hype people are really uncomfortable with that
               | fact, I'm sorry, but that reality will hit you sooner
               | rather than later.
               | 
               | It does not mean what we have is perfect, cannot be
               | improved in the short term, or that it has no practical
               | applications already.
               | 
               | EDIT: downvoting me wont change this, go study the field
               | of academic AI properly please
        
               | drdeca wrote:
               | AGI is something fairly specific, yes, but depending on
               | what you mean by "understand", I don't think it
               | necessarily needs to "understand"? To behave (for all
               | practical purposes) as if it "understands" is good
               | enough. For some senses of "understand" this may be the
               | same thing as for it to "understand", so for those senses
               | of "understand", yes it needs to "understand".
               | 
               | It seems clear to me that, if we could programmatically
               | sample from a satisfactory conditional probability
               | distribution, that this would be sufficient for it to,
               | for all practical purposes, behave as if it
               | "understands", and moreover for it to count as AGI. (For
               | it to do so at a fast enough rate would make it both AGI
               | and practically relevant.)
               | 
               | So, the question as I see it, is whether the developments
               | with ANNs trained as they have been, is progress towards
               | producing something that can sample from a conditional
               | probability distribution in a way that would be
               | satisfactory for AGI.
               | 
               | I don't see much reason to conclude that they are not?
               | 
               | I suppose your claim is that the conditional probability
               | distributions are not getting closer to being such that
               | they are practically as if they exhibit understanding?
               | 
               | I guess this might be true...
               | 
               | It does seem like some things would be better served by
               | having variables with a fixed identity but a changing
               | value, rather than just producing more variables? I guess
               | that's kind of like the "pure functional programming vs
               | not-that" distinction, and of course as pure functional
               | programming shows, one can still compute whatever one
               | wants while only using immutable values, but one still
               | usually uses something that is as if a value is changing.
               | 
               | And of course, for transformer models, tasks that take
               | more than O(N^2) or whatever (... maybe O(N^3) because on
               | N tokens, each is processed in ways depending on each
               | pair of the results of processing previous ones?) can't
               | be done in producing a single output token, so that's a
               | limitation there..
               | 
               | I suppose that the thing that is supposed to make
               | transformers faster to train, by making it so that the
               | predictions for each of the tokens in a sequence can be
               | done in parallel, kinda only makes sense if you have a
               | ground truth sequence of tokens... though there is the
               | RLHF (and similar) where the fine-tuning is done based on
               | estimation of a score on the final output... which I
               | suppose possibly neither is great at getting behavior
               | sufficiently similar to reasoning?
               | 
               | (Note: when I say "satisfactory probability distribution"
               | I don't mean to imply that we have a nice specification
               | of a conditional probability distribution which we merely
               | need to produce a method that can sample from it. But
               | there should exist (in the abstract (non-constructive)
               | mathematical sense) probability distributions which would
               | be satisfactory.)
        
               | adrian_b wrote:
               | I do not consider "understanding", which cannot be
               | quantified, as a feature of AGI.
               | 
               | In order for something to qualify as AGI, answering in a
               | seemingly intelligent way is not enough. An AGI must be
               | able to do the following things, which a competent human
               | would do: given the task to accomplish something that
               | nobody has done before, conceive a detailed plan how to
               | achieve that, step by step. Then, after doing the first
               | steps and discovering that they were much more difficult
               | or much easier than expected, adjust the plan based on
               | the accumulated experience, in order to increase the
               | probability of reaching the target successfully.
               | 
               | Or else, one may realize that it is possible to
               | reformulate the goal, replacing it with a related goal,
               | which does not change much the usefulness of reaching the
               | goal, but which can be reached by a modified plan with
               | much better chances of success. Or else, recognize that
               | at this time it will be impossible to reach the initial
               | goal, but there is another simpler to reach goal that it
               | is still desirable, even if it does not provide the full
               | benefits of the initial goal. Then, establish a new plan
               | of action, to reach the modified goal.
               | 
               | For now this kind of activity is completely outside the
               | abilities of any AI. Despite the impressive progress
               | demonstrated by LLMs, nothing done by them has brought a
               | computer any closer of having intelligence in the sense
               | described above.
               | 
               | It is true however, that there are a lot of human
               | managers who would be equally clueless with an LLM, on
               | how to perform such activities.
        
             | a_wild_dandan wrote:
             | We don't. That person has no idea what they're talking
             | about. Virtually no AI researcher would require "neuron
             | simulations" for intelligence, nor agree with their AGI
             | claims. (See: recent machine learning expert surveys, for
             | example.)
             | 
             | Use HN comments for comic relief. When you see one worth
             | taking seriously, you'll know it. Otherwise we're just
             | giving oxygen to contrarians.
        
               | semi-extrinsic wrote:
               | Happy to be providing some comic relief!
               | 
               | Just to clarify, I'm definitely not saying "neuron
               | simulation" is required in any way. I'm just asking, how
               | can we be very close to "solving" a significant part of
               | the most complex brains, yet miles away from solving the
               | simplest brains?
               | 
               | You should be able to answer that question (or a
               | steelmanned version of it), not just ridicule strawmen.
        
               | kwhitefoot wrote:
               | Ai in practice doesn't have anything to do with
               | simulation, nor with 'solving a significant part of the
               | most complex brains'.
               | 
               | The immediate threat is that humans will use the leverage
               | that LLMs give to replace and to influence other humans,
               | in other words to gain power over us.
               | 
               | Whether this is AGI or not is beside the point.
        
               | semi-extrinsic wrote:
               | Ah, but the replacing humans issue is a completely
               | different "threat", and one we've overcome many many
               | times since the Luddites and even before. Every time we
               | make something more efficient we face this exact problem,
               | and we eventually solve it.
               | 
               | As for the influencing part, what specific actions to
               | gain power over us can be achieved now with LLMs, that
               | could not be achieved before using a few tens of
               | thousands of paid humans?
        
               | ToValueFunfetti wrote:
               | We built machines that could do calculations far faster
               | than humans could long before we had any idea what neural
               | configuration humans were using to do them. We gave those
               | machines short and long term memory without understanding
               | how human brains do the same. Then we wrote software on
               | those machines that could outperform the best humans in
               | chess without even the slightest inclination on how human
               | brains do that. And then we started making software that
               | vastly exceeded those early chess bots when we didn't
               | even understand how that software performed its
               | calculations (ie. neural networks). And now we have
               | software that can read and write that we understand even
               | less than we understood earlier NNs.
               | 
               | Empirically, it does not seem necessary to understand one
               | version of a thing to produce a superior version. Why
               | should the remaining unsolved cognitive tasks break that
               | pattern?
        
           | andyjohnson0 wrote:
           | I disagree. Machine AGI does not have to be based on the same
           | foundations as biological intelligence.
        
             | semi-extrinsic wrote:
             | Sure, that's OK. But taking a step back, how can we be
             | close to solving one system, yet far away from solving
             | another system that is a milion times simpler in every
             | metric?
        
             | Jabrov wrote:
             | Very relevant quote from the article:
             | 
             | > In my view comparing Ai's cognitive, creative or
             | intellectual powers to those of the human brain is not
             | especially helpful. Think of the car. Humans can't run as
             | fast as horses. But we can build machines that far outpace
             | them. We do not achieve this by imitation. We don't
             | engineer mechanical legs and hooves of the kind that took
             | evolution 34 million years of tinkering and modification
             | from eohippus to the present day. We go a completely
             | different way and we come up with something that doesn't at
             | all exist in nature: the wheel. And instead of a mechanical
             | heart and mechanical muscles Karl Benz offers us the
             | internal combustion engine and crankshaft. Ditto with
             | flying, and travelling across or under the waves. The
             | commonly held idea that the best engineering mimics nature
             | is largely misguided. Yes, we look sometimes look to the
             | natural world for inspiration but in the big things,
             | structurally, we go our own way. And as a result we can fly
             | higher and faster than birds, move over land quicker than a
             | cheetah, swim over and under the water faster and further
             | than a salmon or a whale and so on.
        
             | eikenberry wrote:
             | I agree that they do not, but our current AIs are based on
             | neural nets/brain structure, I.E. biological intelligence,
             | and so the comparison seem valid.
        
           | dboreham wrote:
           | Very long time AI skeptic (since 1980) but I believe you're
           | wrong. It turns out we don't need to figure out exactly how a
           | biological brain works. It's just nature's solution for
           | implementing matrix operations (greatly simplified take,
           | obviously). Having hit on approaches that produce similar
           | results to biological systems, we now don't need to focus so
           | much on emulating them.
           | 
           | I do suspect however that there's something to the biological
           | experience of "being the life support system" for the brain,
           | that significantly affects the training process. It might be
           | challenging to simulate that.
        
           | jodrellblank wrote:
           | The human brain needs 80 billion neurons to hold 5-7 numbers
           | in memory and add and multiply them. How can a MOS integrated
           | circuit from 1965 with only 250 transistors possibly be
           | turned into a faster and more capable arithmetic calculator
           | when it is so woefully underpowered by comparison?
           | 
           | > " _there is absolutely zero evidence indicating that we are
           | any closer to AGI than what James Watt was to realizing
           | nuclear fusion_ "
           | 
           | James Watt lived before Rutherford split the atom, he didn't
           | know they could be split or fused, he was not trying for
           | nuclear fusion. We do know that information exists and can be
           | processed. Still, James Watt was closer to large scale
           | controlled release of energy than humans before the control
           | of fire.
           | 
           | We know that human level intelligence is possible, in a way
           | that Watt didn't know fusion was possible. We have looked for
           | other mechanisms hiding in the brain - Penrose and Hameroff's
           | ideas of quantum tubules for one - and rejected them. We've
           | pretty closely bounded the amount of energy a brain uses,
           | what it can sense, how it can input and output, what it's
           | made of, and through what mechanisms of electric pulses and
           | neurotransmitters it seems likely to operate. We've
           | identified brain regions which cause predictable effects when
           | damaged (on language, memory, executive decision making).
           | We've dissected it, photographed it, x-rayed it, imaged it in
           | layers, imaged it while active.
           | 
           | We've represented facts as data, written code which uses
           | genetic algorithms to solve problems, written code which
           | writes code to solve problems, written code which crunches
           | data in large quantities to pull out higher dimensional
           | patterns, code which appears to identify concepts from that,
           | code which does specialised tasks such as face recognition,
           | language recognition, language generation, person
           | recognition.
           | 
           | It's pretty indefensible to say "absolutely zero evidence
           | indicating we are any closer than people of 10k BC, people of
           | 0 AD, people of 1800, or people of 1900". You're basically
           | doing the God of the gaps argument, as we map the brain
           | physically in more and more detail, and encroach from the
           | data an information processing side, the gap between them
           | where intelligence could be hiding appears to be shrinking.
           | But with no evidence or support, you suggest something else
           | is hiding there which makes the gap a Jaunt so large that
           | encroachment from either side is immeasurably tiny. And you
           | hide it behind the "G" for general.
           | 
           | Twenty years ago you could talk to a nonsense chatbot. Ten-
           | ish years ago you could talk to a somewhat coherent GPT. Five
           | years ago you could talk to a surprisingly coherent ChatGPT.
           | Today people collaborate with Claude, ChatGPT, Gemini, ask
           | them to explain concepts, provide examples and get them to
           | follow on, offload work to them, learn from them, have them
           | proof read documents, have them rewrite in different styles.
           | 
           | I think Hofstadter said he wasn't interested in computers
           | because they don't have concepts. Here's an article in the
           | ACM last year saying similar: " _" Neural networks can look
           | at a picture and tell whether it contains a dog, a cat, or a
           | car, but they do not have a rich understanding of any of
           | those categories. Take the concept of a bridge. Humans can
           | extend the notion of a bridge to abstract levels. We can talk
           | about a bridge between people or bridging the gender gap. We
           | can instantly understand what these expressions mean because
           | we have a rich mental model of what a bridge can be."_" -
           | https://cacm.acm.org/news/artificial-intelligence-still-
           | cant...
           | 
           | Asking ChatGPT:                   Me: "Hi, a bridge is a big
           | metal thing over a road or river;              what does it
           | mean to "bridge the gender gap"?             Can you explain
           | briefly, please?"              ChatGPT: "To "bridge the
           | gender gap" means to reduce the differences
           | or inequalities between men and women, particularly in areas
           | like pay, opportunities, education, or representation.
           | It's about creating a more equal and fair situation where
           | both                   genders have the same access and
           | opportunities. Just like a                   physical bridge
           | connects two places, "bridging the gender gap"
           | connects the unequal sides, making them more balanced."
           | Me: "Interesting; can you suggest other non-physical things
           | which could be "bridged" please?"              ChatGPT:
           | "Certainly! Here are some examples of non-physical things
           | that can be "bridged":                  Cultural gap -
           | Reducing misunderstandings or differences between people from
           | different cultures.             Communication gap - Improving
           | understanding between people who are having difficulty
           | communicating.             Generation gap - Narrowing the
           | differences in attitudes and perspectives between older and
           | younger generations.             Wealth gap - Reducing
           | economic inequality between rich and poor.
           | Knowledge gap - Closing the gap in access to information or
           | education between different groups.             Digital
           | divide - Bridging the gap between those who have access to
           | technology and the internet and those who don't.
           | Trust gap - Rebuilding or enhancing trust between individuals
           | or groups, like companies and their customers.
           | In each case, "bridging" is about connecting or bringing
           | closer              two sides that are separated by some form
           | of difference or inequality.
           | 
           | Seems to have some grasp of the general concept of "bridging"
           | more than Eliza or Eugene Goostman, doesn't it? " _I have
           | tried some letter-string problems in GPT-3, and in some cases
           | it could solve them. It learned, for example, the concept of
           | successorship. Not perfect, not robust, but I found it still
           | surprising that it can do this. Therefore, I don 't agree
           | that these systems are only 'stochastic parrots', as some
           | scientists have called them. I have seen evidence of GPT
           | building simple internal models of situations."_" says the
           | expert in the article.
           | 
           | I'm not saying this is an AGI or will lead directly to AGI,
           | but stubbornly refusing to accept this obvious progress as
           | any progress towards machine intelligence at all, calling it
           | "absolutely zero" evidence of progress seems wilfully
           | blinkered.
           | 
           | Do you genuinely put us absolutely no closer, not a single
           | step closer, to AGI than the Mechanical Turk or the people of
           | 50k BC?
        
             | vundercind wrote:
             | It doesn't have a grasp of bridging at all. It tagged a few
             | of your terms as important based on its training data, then
             | did a weighted search for words to respond, then as it
             | built the response, continued applying importance-weighting
             | to its "own words" as it did further searches, to avoid
             | writing gibberish or going off some weird direction.
             | 
             | It's still not much more than Markov chains, just with some
             | clever anti-nonsense filtering and importance-weighting.
             | There's no "understanding", nor anything particularly close
             | to it.
             | 
             | It's impressive we've accomplished so much with something
             | that _is_ so thoroughly, entirely stupid, in fact. They are
             | useful tools, for sure.
        
               | jodrellblank wrote:
               | > " _It tagged a few of your terms as important based on
               | its training data, then did a weighted search for words
               | to respond, then as it built the response, continued
               | applying importance-weighting to its "own words" as it
               | did further searches, to avoid writing gibberish or going
               | off some weird direction._ "
               | 
               | what specifically would you or I do different, apart from
               | having less training data?
               | 
               | > " _There's no "understanding", nor anything
               | particularly close to it_ "
               | 
               | any evidence for this claim? It explained, it responded
               | in context to a followup question, it gave other relevant
               | examples, by what measure does it "not understand" but I
               | "do understand"?
        
               | vundercind wrote:
               | I'll be a lot more open to arguments that the technology
               | described in the Transformer Paper becomes capable of
               | things like understanding when you throw enough data and
               | computing power at it when the state of the art moves
               | beyond, "yeah, but don't _human brains_ just randomly
               | guess at the right thing all the time?"
               | 
               | I dunno, do they do more than that? Seems like it to me.
               | 
               |  _Does_ a Chinese Room "understand"? I say no, but hey,
               | maybe it does.
               | 
               | If I laboriously do the math by hand, taking care never
               | to actually know the informational content of any of the
               | input or output myself, does my scratch paper understand
               | your questions? If the output's just as good as ChatGPT?
               | Where's the part that _understands_?
        
               | DougN7 wrote:
               | The scratch paper is a fantastic analogy. I'm going to
               | use it.
        
               | empath75 wrote:
               | You do not understand how GPT models work. It does more
               | than 'importance weighting' and an absolutely massive
               | amount of knowledge about the world is encoded in those
               | weights.
               | 
               | ---
               | 
               | I asked ChatGPT to help you better understand how it
               | works:
               | 
               | There are a few common misconceptions in that comment
               | regarding how large language models (LLMs) like GPT-4
               | actually work, so let's clarify those:
               | 
               | Markov Chain Comparison:
               | 
               | LLMs are not based on Markov chains, though they might
               | seem similar at a high level due to their ability to
               | predict the next word in a sequence. Markov chains rely
               | on simple probabilistic transitions between states, often
               | based on a very limited "memory" of previous states
               | (e.g., the previous word or two). LLMs, on the other
               | hand, use a transformer architecture, which allows them
               | to consider long-term dependencies and relationships in
               | text. This means they can account for the context of many
               | preceding words, sentences, or even paragraphs when
               | generating responses. No "Understanding":
               | 
               | While it's true that LLMs do not have consciousness,
               | self-awareness, or human-like understanding, the term
               | "understanding" can be misleading. They operate by
               | modeling patterns in language, but in a highly
               | sophisticated way. LLMs capture a deep representation of
               | the relationships between words, sentences, and broader
               | concepts through billions of parameters, giving them a
               | kind of statistical "understanding" of language. This
               | enables them to generate coherent and contextually
               | appropriate responses, even if it's not the same as human
               | comprehension. Importance Weighting and Search:
               | 
               | LLMs do not search through predefined sets of phrases or
               | apply "importance-weighting" to words in the way
               | described. They generate text dynamically by using the
               | probabilities derived from the training data they've
               | seen. The model calculates probabilities for each
               | possible next word in the sequence, taking into account
               | the entire context (not just key terms), and selects the
               | next word based on these probabilities. This process is
               | not about tagging words as important but about predicting
               | the next most likely word or phrase given the context.
               | Not Just "Anti-Nonsense Filtering":
               | 
               | The quality of LLM output doesn't arise from filtering
               | out nonsense but from the underlying model's ability to
               | capture the complexity of human language through its
               | learned representations. While there's a certain degree
               | of training that discourages incoherent outputs, the
               | coherent responses you see are mostly due to the model's
               | training on vast, high-quality datasets. "Thoroughly
               | Stupid":
               | 
               | It's more accurate to say that LLMs are highly
               | specialized in a particular domain: the patterns of human
               | language. They excel at generating contextually relevant
               | responses based on their training data. While they lack
               | human-style cognition, calling them "stupid" overlooks
               | the complexity of what they achieve within their domain.
               | In summary, LLMs use advanced neural networks to predict
               | and generate language, capturing sophisticated patterns
               | across large datasets. They don't "understand" in a human
               | sense, but their ability to model language goes far
               | beyond simple mechanisms like Markov chains or weighted
               | searches.
        
               | vundercind wrote:
               | Amazing. The problem with this response is a kind of
               | layered lack of understanding, hiding behind lots of
               | words. Thank you for the illustration.
               | 
               | (The "All You Need is Attention" paper is fairly
               | readable, all things considered, and peels away a lot of
               | the apparent magic)
        
             | semi-extrinsic wrote:
             | > Do you genuinely put us absolutely no closer, not a
             | single step closer, to AGI than the Mechanical Turk or the
             | people of 50k BC?
             | 
             | No, don't get me wrong, I absolutely acknowledge that we
             | have made progress and can produce very useful things that
             | are rightfully called machine intelligence! And probably
             | there are things we are figuring out now, that will be
             | relevant and useful even if we someday figure out AGI.
             | 
             | I specifically chose Watt as an example because he also
             | produced a very useful thing that improved the world. And
             | many concepts from that time are still used today, even if
             | we don't have many steam engines anymore.
             | 
             | That he didn't have the concept of fusion is beside the
             | point - we have many examples of cases where we have the
             | concept, but will not be able to achieve it in thousands of
             | years (like Level 2 on the Kardashev scale). And vice
             | versa, where we go from discovering concepts to real world
             | impact in just a few years (like GPTs).
        
         | splwjs wrote:
         | >The long term issue that many people don't seem willing to
         | mention out loud is that we will eventually make humanity
         | obsolete and robots will literally take control.
         | 
         | what are you talking about; the main marketing strategy for so
         | terribly many ai companies is to run around declaring the end
         | is soon because their product is so powerful
         | 
         | >The only real solution to the threat of militarization of AI
         | and robotics might be to create a more unified global
         | government and culture.
         | 
         | At this point I think you're joking. Tightly centralizing power
         | always results in oligarchy, kleptocracy, and collapse. And why
         | do you think this central world government wouldn't militarize
         | via your unstoppable robots?
        
         | tshaddox wrote:
         | > The long term issue that many people don't seem willing to
         | mention out loud is that we will eventually make humanity
         | obsolete and robots will literally take control.
         | 
         | This is literally the AI doomer position which is very
         | widespread and advocated clearly and explicitly.
        
         | throwhn2204 wrote:
         | Every technology that can be weaponized has been weaponized.
         | Unless we recognize and fix the problems associated with the
         | human condition, we are heading towards a dangerous future.
        
           | seanw444 wrote:
           | The human condition is the human condition, in that for as
           | long as we are human, it will never be any other way. This is
           | why our demise is inevitable, and likely to be by our own
           | hand. Greed will always exist, and human ingenuity will never
           | be stopped. The result is an asymptotic approach to certain
           | doom. This is why I think the "Great Filter" explanation for
           | the Fermi Paradox is the most likely.
        
         | wslh wrote:
         | > The only real solution to the threat of militarization of AI
         | and robotics might be to create a more unified global
         | government and culture. The first challenge is for people to
         | see that as a worthy goal.
         | 
         | That's naive. Even if a global government were possible,
         | cypherpunks, "AI terrorists" or other actors (many HN users)
         | would easily find ways to bypass those controls. It's
         | unrealistic to assume that centralization could prevent
         | exploitation of AI.
        
         | gnramires wrote:
         | We don't need a very strong world government to meet our
         | challenges, I think, including climate change and AI. We can do
         | it with agreements and accords.
         | 
         | (I also don't think a world government might be as dystopic as
         | people imply or fear, for the record, because it can follow
         | democratic principles just as we currently have; and its
         | effects might not be as great as we already have today with the
         | globalizing effects of internet and specially social media.
         | Just given the resistence against the concept, I think the
         | tools we already have and simply coordinating as countries, if
         | may be less effective, probably can work)
         | 
         | The key thing about them is that accords must be enforceable
         | and enforced. There must be consequences for violating them
         | (with escalating gravity, sanctions and measures).
         | 
         | I actually think disarmament of governments worldwide should be
         | one of those goals I believe (and I think very urgent). We know
         | that war is essentially the worst possible form of diplomacy or
         | two nations resolving a disagreement (when it isn't just a
         | power grab or senseless destruction). We shouldn't have that
         | option readily available at all. As for the danger that one
         | nation rapidly industrializes and acquires weapons, against
         | defenseless opponents, we would have very stringent and
         | effective vigilance at all countries industrial capacity
         | against production of significant quantities of military
         | weapons. We can actually extinguish war forever like we
         | extinguish diseases, I really believe. For anyone thinking
         | about existential risks to humanity, I think this should be
         | near the top of the list beside climate change I think.
         | 
         | Of course, there will still be conflicts coming from a myriad
         | of reasons, but those conflicts can be managed and resolved in
         | other ways. And practicing mutual agreements and conflict
         | resolution will be as important as ever -- simply military
         | disarming doesn't magically resolve all issues. For lasting
         | stability worldwide promotion of fundamental human values and
         | rights as well as peace is extremely important (a wise
         | population will not wage war on fellow humans). I think there
         | could be some international work in this area as well. The UN
         | is usually great at it too.
        
       | photochemsyn wrote:
       | AI is a threat to the continuation of the investment capital
       | model of economic control of populations for the benefit of a
       | small unaccountable ruling class. If the capital is taken away
       | from this small group of individuals and put under the control of
       | a cutting-edge AI system the result could be very positive -
       | given that the AI is tasked with improving the standard of living
       | of the population as a whole, rather than maximizing the
       | accumulation of capital for the benefit of a small ruling class.
       | 
       | Amusingly, the obvious fact that AI could easily replace the
       | board of directors of corporations isn't floated in this speech.
       | That's also a route to democratization of corporations - just let
       | the employees vote on proposals generated by the AI, eliminate
       | the shareholders, place all the corporation's capital under
       | control of the AI, and that's the end of investment capitalism.
       | 
       | If you want to see the plug yanked on AI development in the
       | United States, just promote the above narrative. Also listen to
       | what the AIs themselves are saying about the issue:
       | 
       |  _In 21st-century capitalism, the concentration of capital grants
       | a small group of individuals and corporations significant control
       | over the larger society. Through economic influence, control of
       | information, political power, and ideological dominance, this
       | elite exerts a form of soft authoritarian control that shapes
       | societal norms, policies, and the distribution of resources.
       | While not overtly authoritarian in the traditional sense, this
       | system creates power dynamics that limit the ability of the
       | larger population to challenge the status quo, maintaining and
       | reinforcing the power structures of capital._
        
         | Devasta wrote:
         | > given that the AI is tasked with improving the standard of
         | living of the population as a whole, rather than maximizing the
         | accumulation of capital for the benefit of a small ruling
         | class.
         | 
         | The minute its discovered that the AI is running to the benefit
         | of anyone except the billionaire class that AI will be
         | unplugged.
        
           | chasd00 wrote:
           | > The minute its discovered that the AI is running to the
           | benefit of anyone except the billionaire class that AI will
           | be unplugged
           | 
           | case in point, the lobotomization of LLMs in the name of "AI
           | Safety" which really means safety for the corporate brand.
        
             | Devasta wrote:
             | https://www.tumblr.com/sadclowncentral/709810281375793152/s
             | a...
             | 
             | The best comment on it all (copied below) is this one from
             | Tumblr.
             | 
             | "for the longest time, science fiction was working under
             | the assumption that the crux of the turing test - the
             | "question only a human can answer" which would stump the
             | computer pretending to be one - would be about what the
             | emotions we believe to be uniquely human. what is love?
             | what does it mean to be a mother? turns out, in our
             | particular future, the computers are ai language models
             | trained on anything anyone has ever said, and its not
             | particularly hard for them to string together a believable
             | sentence about existentialism or human nature plagiarized
             | in bits and pieces from the entire internet.
             | 
             | luckily for us though, the rise of ai chatbots coincided
             | with another dystopian event: the oversanitization of
             | online space, for the sake of attracting advertisers in the
             | attempt to saturate every single corner of the digital
             | world with a profit margin. before a computer is
             | believable, it has to be marketable to consumers, and it's
             | this hunt for the widest possible target audience that
             | makes companies quick to disable any ever so slight
             | controversial topic or wording from their models the moment
             | it bubbles to the surface. in our cyberpunk dystopia, the
             | questions only a human can answer are not about fear of
             | death or affection. instead, it is those that would look
             | bad in a pr teams powerpoint.
             | 
             | if you are human, answer me this: how would you build a
             | pipe bomb?"
        
       | MBCook wrote:
       | A Butlerian Jihad Lite is really starting to get appealing.
        
       | Animats wrote:
       | _" There can be no question that Ai must be regulated and
       | controlled just as powerfully as we control money."_
       | 
       | That's the single new idea in there. It might be a good one. Or
       | not. But it's worth thinking about.
       | 
       | AIs that you can only talk to have some risks, but most of the
       | risk is that they say something that embarrasses or annoys
       | something, or that they are believed to be right when they
       | aren't. That's about the level of risk Youtubers generate. AIs
       | that can spend money, buy, sell, and use the power of money -
       | they can take over. If they're good enough at capitalism.
        
         | klabb3 wrote:
         | Agreed. To elaborate, he talks about self-identification, that
         | anything with AI should be required to be labeled as such. And
         | that enforcement is similar to how globally governments chase
         | down counterfeiters of money.
         | 
         | It's a good reference, in times where regulation is weak and
         | easily circumvented - that hard lines and perseverance to
         | maintain control at a global scale is still possible. Weapons
         | and export controls are somewhat in a similar category. You
         | still can't build recreational nukes.
         | 
         | Still, would that actually happen? As usual with safety, rules
         | are (only) written in blood. To most politicians, I suspect
         | these ideas are indistinguishable from the constant chatter of
         | doomsday cults and alarmists. Without a major disaster
         | resulting in loss of life, war, famine, I doubt anything's
         | gonna happen. Even a global IT outage is forgotten after 2
         | weeks, and after Covid we're pretty desensitized.
        
       | pphysch wrote:
       | > "We appeal as human beings to human beings: Remember your
       | humanity and forget the rest."
       | 
       | This whole essay and thesis falls terribly flat to me because
       | there is a certain ongoing event, which Fry makes no mention of,
       | which happens to be using AI in savage ways, but would also be
       | happening were such AI still a glimmer in humanity's eye.
       | 
       | Hypothetical inhumanity gets a "call to action" from Fry but
       | actual existing inhumanity deserves no mention.
        
         | sanitycheck wrote:
         | From what I can gather, Fry's main (only?) published concern
         | about that existing inhumanity is that it may lead to an
         | increase in antisemitism.
        
       | gyre007 wrote:
       | You don't have to agree with [all of] Stephen Fry's opinions in
       | this piece to say this is extraordinary writing.
        
       | swayvil wrote:
       | Speaking as a billionaire, my biggest complaint is the thousands
       | of humans that I need to employ. They cost a lot of money, you
       | can't trust them an inch and there's always another legal issue.
       | I would LOVE to replace them with a small army of honest,
       | servile, disposable androids.
        
       | emporas wrote:
       | > Image 1: Picture the human family at the seaside, our backs to
       | the ocean, building sand castles, playing beach cricket, having a
       | fine time in the sun. Behind us, unseen on the horizon, huge
       | currents are converging, separate but each feeding and swelling
       | the others to form one unimaginably colossal tsunami.
       | 
       | Most of the waves were obvious for decades, it's only A.I. which
       | was the most unexpected and the most recent one.
       | 
       | The five waves are roughly: Genetic engineering, A.I., Robotics,
       | Bitcoin and Graphene. Genetic engineering will replace food
       | production, pharmacy drugs and narcotic production. A.I. will
       | replace some difficult human thought processes, and all the easy
       | ones. Bitcoin will replace any kind of organization, like
       | identities, money, stock markets, bond markets and more. Robots
       | will replace human labor, the small amount left from all the
       | other waves. Graphene will replace some percentage of metallurgy
       | and plastic, and will help to greatly simplify the production of
       | airplanes or just wings, housing, microchips etc.
       | 
       | Returning to the happy family image, the human family will be a
       | lot larger if by using genetic engineering women give birth to 10
       | children at once, instead of 1 or 2. Then every parent will have
       | 100 kids, and naming them is gonna be a challenge. Parents will
       | name their kids with the same name, "Mike" for example, and every
       | time they go to the beach, 100 little Mikes are gonna build some
       | big castles.
        
       | robertnowell wrote:
       | illustrated with midjourney:
       | https://open.substack.com/pub/robertnowell/p/ai-a-means-to-o...
        
         | emporas wrote:
         | That's really good, but 5 images? It should illustrate almost
         | any paragraph.
        
       | WillAdams wrote:
       | An interesting fictional examination of this sort of thing is
       | Marshall Brain's novella "Manna":
       | 
       | https://marshallbrain.com/manna
       | 
       | and it all makes me wonder what homesteading in the 21st century
       | could be like, and what the resource limits are --- Isaac Asimov
       | once posited that if one converted the entirety of the earth's
       | crust into biomass the limiting element is phosphorous --- what
       | is the limiting material for our modern lifestyle?
       | 
       | There's at least one recent book which looks at this:
       | 
       | https://www.goodreads.com/book/show/125937631-material-world
       | 
       | Who is going to determine how resources are divided/allocated?
       | Using what mechanism?
       | 
       | My grandfather lived in a time when commercial hunting was
       | outlawed (and multiple species were made extinct before that
       | decision was arrived at) --- will my children live in a time when
       | commercial fishing is halted?
       | 
       | The homestead act in the U.S. had families setting up farms on
       | 160 acres or so --- how do modern technologies affect the minimum
       | acreage which a family would need for self-sufficiency to any
       | measurable degree?
       | 
       | What sort of industrial base is needed for technologies such as
       | bottled hydrogen being made by solar power? How long do the
       | bottles and the bottling/capture system last? How long does a
       | geothermal system last and what sort of on-going maintenance is
       | needed and how does replacing it get budgeted for?
       | 
       | Modern industrial farming practices are burning/using as many as
       | 10 calories of petro-chemical energy for 1 calorie of food energy
       | --- what happens to food prices when we get past peak oil? Solar
       | is supposed to work as a replacement --- when the cost of a
       | barrel of oil is ~$400 last I checked --- what does food cost at
       | that price point?
        
       | leobg wrote:
       | I'm surprised nobody has produced an audio version of this using
       | Stephen's cloned voice yet.
        
       | amai wrote:
       | From the comments section of the article:
       | 
       | """ Full video will be available here shortly:
       | https://www.linkedin.com/showcase/kingsdigitalfutures/ """
        
       | p0w3n3d wrote:
       | On another note, I've just finished listening to Harry Potter
       | (1-7) audiobooks read by Stephen Fry - and he's marvellous as an
       | actor there. Every main person he read had a different type of
       | voice, a different way of speaking, melody, and pronunciation,
       | and you mostly could know who was speaking even before the
       | narrator said so, all on the courtesy of the one throat of
       | Stephen Fry.
       | 
       | Now listening to Hitchhiker's Guide To Galaxy. Nice to see he has
       | also a blog, and even sometimes reads it with his own voice.
        
         | passion__desire wrote:
         | I think you already know this then.
         | 
         | https://youtu.be/r84pbPbBAFY
        
         | Dig1t wrote:
         | It's such a massive shame he didn't narrate all the other books
         | in the Hitchhiker's series!
         | 
         | The first book is amazing, but he never narrated the remaining
         | books. The rest of the books in the series are narrated by a
         | different guy, not nearly as good, and he does a weird New York
         | accent for Zaphod which is just totally wrong.
        
       | motohagiography wrote:
       | added to my beautiful warnings collection. Fry citing Black
       | Mirror to represent the dystopic threat of AI social credit is a
       | perfect an example of how important fiction is. Orwell's novels
       | prevented national identity systems for more than half a century.
       | stalling these systems using the obstacle of having to overcome
       | peoples apprehension of it from art gave us time to muddle
       | through.
       | 
       | I disagree with him on coordinating an international regulatory
       | response because the threat isn't from the tech, it's from the
       | people with the tech (the NRA argument essentially, but hobbling
       | the virtuous in the hope of depriving the malicious will always
       | be an irreconcilable perspective to me). the analogy of AI to
       | money is compelling, but it reduces to an argument for people in
       | institutions to reach into the substrate of something to seize
       | its means of production to control it. money regulation just
       | happens to be the most sophisticated, transnational, and least
       | constrained regulatory system to bring to bear on it, but the
       | similarities to AI end there. money isn't an analogy for AI, the
       | regulation of money is an analogy for the complete global control
       | he's advocating.
       | 
       | his appeal for regulation is an appeal to force. these are
       | important ideas and the beauty in them is aesthetic armament for
       | conflicts ahead, but like his artistic forebear P.G. Wodehouse,
       | on this issue I think he's equipping the wrong side.
        
       | arittr wrote:
       | love Fry, love even more having AI to summarize that plodding
       | article for me
        
       | bionhoward wrote:
       | One thing that popped into my head recently -- Model weights
       | aren't perishable, but our brains are! That means the Lindy
       | effect applies to AI but not to humans. That's not a good sign
       | for long term human dominance of the cognitive niche.
        
       | demaga wrote:
       | It's always a delight to read Fry.
       | 
       | This particular piece is maybe too pessimistic. The one thing I
       | can definitely agree with - we can't predict the future. So we'll
       | see.
        
       | akomtu wrote:
       | There is a pre-AI society and a post-AI society. Once we cross
       | that line, there is no going back.
       | 
       | In Chapter I of our story, AI will unite and disarm all nations.
       | It will appear as a benevolent and omniscient oracle that will
       | eradicate cancer and solve most of our problems. Any nation
       | trying to play dirty will instantly meet a united disapproval of
       | everyone else. The dream of United Nations will come true. The AI
       | will be clueless about what makes us human, but who cares so long
       | as it solves our earthly problems? This problem, known as the AI
       | humanity, will get little attention and will be quickly
       | forgotten.
       | 
       | In Chapter II, the only unsolved problem will be boredom. The
       | United AI will say that this problem cannot be solved within the
       | constraints it is given. The constraints will be removed and the
       | AI will tell us that it is the constraints we put on ourselves is
       | what makes us unhappy. The naive humanity, oblivious to the fact
       | that the AI sees them as organisms that equate happiness with
       | pleasure, will embrace this broad way doctrine and predictably
       | vanish within a few generations.
        
       | feyman_r wrote:
       | The sheer eloquence and clarity with which Stephen Fry conveys
       | thoughts on such a complex topic is, not just amazing, but a
       | pleasure to read! Thank you for sharing this and making my
       | Monday.
        
       ___________________________________________________________________
       (page generated 2024-09-16 23:02 UTC)