[HN Gopher] If AI seems smarter, it's thanks to smarter human tr...
___________________________________________________________________
If AI seems smarter, it's thanks to smarter human trainers
Author : getwiththeprog
Score : 81 points
Date : 2024-09-28 13:28 UTC (9 hours ago)
(HTM) web link (www.reuters.com)
(TXT) w3m dump (www.reuters.com)
| GaggiX wrote:
| Let's not ignore better architectures, training techniques and
| computing power.
| kaycebasques wrote:
| Suppose you are competing to create the "best" frontier model
| and have finite R&D budget to allocate into the 4 buckets (plus
| a catchall in case we're missing something):
|
| * Data
|
| * Architecture
|
| * Training techniques
|
| * Compute
|
| * Other
|
| What allocation gives you the best chance for success? And how
| are you defining "best"?
| NitpickLawyer wrote:
| Right now I'd prioritise compute over anything, because it
| allows for more experiments, and some of those experiments
| might turn out to be the key to better models (either
| specific applications or overall generalist models).
|
| Meta did this with L3. They used L2 to pre-filter the
| training data, filtering out a lot of junk. They also used it
| to classify datasets. Then after pre-training (involving lots
| of compute) they also used almost exclusively synthetic data
| for fine-tuning (forgoing RLHF when it was surpassed). So yet
| more compute. The results are pretty good, L3, 3.1 and 3.2
| are pretty high up there in terms of open access SotA.
|
| oAI did this with their o1 models. They used lots of compute
| to have the models go over the space of generating tokens,
| analysing, correcting, and so on. They RLd the "reasoning
| traces" in a way. Lots of compute. The results seem to be
| pretty good, with impressive showings on "reasoning" tasks,
| math, code, and so on.
|
| The thing is, they weren't the first ones to propose these
| techniques! What differentiates them is the available
| compute.
|
| WizardML tried and were really successful with their RLAIF
| implementation (tho never released code afaik) about a year
| ago. And while they were connected to MS research, they
| probably didn't have as much compute available as Meta. But
| the WizardML fine-tunes on open models like Mistral and
| Mixtral were pretty much SotA when released, scoring way
| higher than the creator's own fine-tunes.
|
| In the same vein, but at lower scales is the team behind
| DeepSeek. They used RL on math problems, in their
| DeepSeekMath-7bRL model, and that model was SotA at the time
| of release as well. It took a team of multiple really
| talented folks to fine-tune a better model (in the AIMO
| kaggle competition) and everyone except the 1st place used
| the RL model. The 1st place used the base model, with
| different fine-tuning. So again, the methods were tried, just
| at much lower scales.
|
| Yeah, I think compute would be my bet in the short run.
| JCharante wrote:
| using human feedback for reinforcement learning is a training
| technique
| ben_w wrote:
| It's both. I recently saw a comparison of various models on two
| IQ tests, one of which was public and the other of which was
| carefully curated to be not directly learnable from the likely
| training sets.
|
| On public tests, LLMs vary between "just below average human"
| and "genius".
|
| On the hopefully-private test (it's difficult to be sure*), the
| best was o1, which was "merely" just below an average human,
| Claude-3 Opus which was stupid, and all the rest were "would
| need a full time caretaker".
|
| In both cases, the improvements to the models came with higher
| scores; but there's still a lot you can do by learning for the
| test -- and one thing that LLMs are definitely superhuman at is
| that.
|
| https://www.maximumtruth.org/p/massive-breakthrough-in-ai-in...
|
| * I could have said the same last year about last year's
| models, so I'm emphatically _not_ saying o1 really is as smart
| as this test claims; I 'm _only_ saying this demonstrates these
| IQ tests are a learnable skill up to at least this magnitude of
| difference.
| CamperBob2 wrote:
| Which is fine. If all AI does is represent human knowledge in a
| way that makes it explainable and transformable rather than
| merely searchable, then the hype is justified... along with
| Google's howling, terrified panic.
|
| The role played by humans on the training side is of little
| interest when considering the technology from a user's
| perspective.
| iwontberude wrote:
| I think the most interesting aspect of it is the human
| training. Human blindsides, dogma, ignorance, etc. All on
| demand and faster than you can validate its accuracy or
| utility. This is good.
| CamperBob2 wrote:
| Shrug... I don't know what anyone expected, once humans got
| involved. Like all of us (and all of our tools), AI is
| vulnerable to human flaws.
| ddulaney wrote:
| I think that's really important to reinforce! You probably
| know better, but lots of the less technical people I talk
| to don't think that way. It's not at all obvious to an
| observer who doesn't know how this stuff works that a
| computer could be racist or misogynist.
| CamperBob2 wrote:
| Yeah, I do think that's going to be a problem.
|
| Years ago, my GF asked me why we bother with judges and
| juries, given all the uneven sentencing practices and
| other issues with the current legal system. "Why can't
| the courts run on computers?" This was back in the pre-
| Alpha Go era, so when I answered her, I focused on
| technical reasons why Computers Can't Do That... reasons
| that are all basically obsolete now, or soon will be.
|
| The real answer lies in the original premise of her
| question: because Humans Also Can't Do That with the
| degree of accuracy and accountability that she was asking
| for. Our laws simply aren't compatible with perfect
| mechanized jurisprudence and enforcement. Code may be
| law, but law isn't code.
|
| That problem exists in a lot of areas where people will
| be looking to AI to save us from our own faults. Again,
| this has little to do with how training is conducted, or
| how humans participate in it. Just getting the racism and
| misogyny out of the training data isn't going to be
| enough.
| Terr_ wrote:
| Also: It's not just about what task can/can't can't be
| done, but what other frameworks you/can't build around
| the executor to detect errors and handle exceptional
| cases.
| jumping_frog wrote:
| The problem is my back and forth with Claude is just Claude's
| data not available to any other. Unlike stack overflow which is
| fair game for every AI.
| Stem0037 wrote:
| AI, at least in its current form, is not so much replacing human
| expertise as it is augmenting and redistributing it.
| alephnerd wrote:
| Yep. And that's the real value add that is happening right now.
|
| HN concentrates on the hype but ignores the massive growth in
| startups that are applying commoditized foundational models to
| specific domains and applications.
|
| Early Stage investments are made with a 5-7 year timeline in
| mind (either for later stage funding if successful or
| acquisition if less successful).
|
| People also seem to ignore the fact that foundational models
| are on the verge of being commoditized over the next 5-7 years,
| which decreases the overall power of foundational ML companies,
| as applications become the key differentiator, and domain
| experience is hard to build (look at how it took Google 15
| years to finally get on track in the cloud computing world)
| MostlyStable wrote:
| I notice that a lot of people seem to only focus on the
| things that AI _can 't_ do or the cases where it breaks, and
| seem unwilling or incapable of focusing on things it _can_
| do.
|
| The reality is that both things are important. It is
| necessary to know the limitations of AI (and keep up with
| them as they change), to avoid getting yourself in trouble,
| but if you ignore the things that AI can do (which are many,
| and constantly increasing), you are leaving a ton of value on
| the table.
| alephnerd wrote:
| Yep! Nuance is critical, and sadly it feels like nuance is
| dying on HN.
| aleph_minus_one wrote:
| > I notice that a lot of people seem to only focus on the
| things that AI _can 't_ do or the cases where it breaks,
| and seem unwilling or incapable of focusing on things it
| _can_ do.
|
| I might be one of these people, but in my opinion, one
| should not concentrate on things that it _can_ do, but for
| how many of the things where an AI might be of help for
| you,
|
| - it does work
|
| - it only "can" do it in a very broken way
|
| - it can't do that
|
| At least for the things that _I_ am interested in an AI
| doing for me, the record is rather bad.
| vladms wrote:
| How do you define "can do" ? Would answering correctly 9
| out of 10 questions correctly for a type of question (like
| give directions knowing a map) mean it "can do" or that it
| "can't do" ?
|
| Considering it works for so many cases, I think it is
| naturally to point out the examples where it does not work
| - to better understand the limit.
|
| Not to mention that practically, I did not see anything
| proving that it will always "be able" to do something .
| Yes, it works most of the times for many things, but it's
| important to remember it can (randomly?) fail and we don't
| seem to be able to fix that (humans do that too, but having
| computers fail randomly is something new). Other software
| lets say a numerical solver or a compiler, are more stable
| and predictable (and if they don't work there is a clear
| bug-fix that can be implemented).
| Workaccount2 wrote:
| ...and then being blown up when the AI company integrates
| their idea.
| alephnerd wrote:
| Not exactly.
|
| At least in the cybersecurity space, most startups have 3-5
| year plans to build their own foundational models and/or
| work with foundational model companies to not directly
| compete with each other.
|
| Furthermore, GTM is relationship and solution, and an
| "everything" company has a difficult time sympathizing or
| understanding GTM on a sector to sector basis.
|
| Instead, the foundational ML companies like OpenAI have
| worked to instead give seed/pre-seed funding to startups
| applying foundational MLs per domain.
| danielbln wrote:
| Same with consultancy. There is a huge amount of automation
| that can be done with current gen LLMs, as long as you keep
| their shortcomings in mind. The "stochastic parrot" crowd
| seems an over correction to the hype bros.
| alephnerd wrote:
| It's because the kind of person who understands nuance
| isn't the kind of person to post in HN flame wars.
|
| The industry is still in it's infancy right now, and stuff
| can change in 3-5 years.
|
| Heck, 5 years ago models like GPT-4o were considered
| unrealistic in scale, and funding in the AI/ML space was
| drying up at the expense of crypto and cybersecurity. Yet
| look at the industry today.
|
| We're still very early and there are a lot of opportunities
| that are going to be discovered or are in the process of
| being discovered.
| skybrian wrote:
| It would be nice to have more examples. Without specifics,
| "massive growth in startups" isn't easily distinguishable
| from hype.
|
| A trend towards domain-specific tools makes sense, though.
| alephnerd wrote:
| DevTools/Configuration Management and Automated SOC are two
| fairly significant example.
| skybrian wrote:
| Those are more like broad categories than examples of
| startups, though.
| jayd16 wrote:
| Am I the only one unimpressed by the dev tool situation?
| Debugging and verifying the generated code is more work
| than simply writing it.
|
| I'm much more impressed with the advances in computer
| vision and image generation.
|
| Either way, what are the startups that I should be
| looking at?
| Terr_ wrote:
| And even when the output is perfect, it may be that the
| tool is helping you write the same thing a hundred times
| instead of abstracting it into a better library or helper
| function.
|
| Search/Replace as a service.
| hanniabu wrote:
| Yes, it should really be called collective intelligence not
| artificial intelligence
| ysofunny wrote:
| I feel weird being stubborn against free tier google gemini
|
| I feel as though it 'extracts' some sort of "smartness" out of me
| (if any) and then whatever intelligence from me becomes part of
| google gemini
|
| this is why I would never want to pay for using these tools,
| anything good that comes from me in the chat becomes google's by
| AI training, which is ok so long as it's free to use
|
| i.e. I won't pay to make their stuff better through my own work
| buzzerbetrayed wrote:
| I totally sympathize with the sentiment. But how long until
| people who are taking a moral stand against AI are simply
| obsoleted by the people who don't? Today it's easy to code
| effectively without relying on AI. But in 10 years will you
| simply be too slow? Same argument can be made with nearly any
| industry.
| pixl97 wrote:
| Pretty much like the people that don't care about privacy.
| You still get captured and tagged in their information and
| uploaded to the web. As an individual it's difficult to do
| much about it.
| croes wrote:
| That's same logic as for frameworks like react.
|
| With react you are more productive, my web experience is
| worse than without a those frameworks.
|
| And LLMs get worse if they are trained on AI generated text.
| At the current speed I don't know if in 10 years AI is still
| worse the high costs.
| joshstrange wrote:
| > With react you are more productive, my web experience is
| worse than without a those frameworks.
|
| You cannot begin to know that for sure and really makes
| little to no sense if you think about it.
|
| As with the anti-electron crowd the options are not:
|
| * Electron app
|
| or
|
| * Bespoke, hand-crafted, made with love, native app
|
| The options are normally "electron app" or "nothing".
|
| Same deal here. Taking away React/Angular/Vue won't
| magically make people write more performant websites. I'm
| sure people bitched about (and continue to) PHP for making
| it easy for people to create websites that aren't
| performant or Wordpress for all its flaws. It's the same
| story that's repeated over and over in tech circles and I
| find it both silly and incredibly short-sighted. Actually I
| find it tiring because you can always go one level deeper
| to one-up these absurd statements. It's No True Scotsman
| all the way down.
| emptiestplace wrote:
| I feel like I (probably?) agree with what you are saying,
| but this is a very confusing comment. You started out
| with an epistemological argument, and then jumped into an
| analogy that's so close to what is being discussed that
| on first read I thought you were just confused. I'm not
| sure anyone can continue the discussion in a meaningful
| way from what you've written because so many aspects of
| your comment are ambiguous or contradictory.
| croes wrote:
| I mean retrospectively.
|
| In the time before all those framework like react the UX
| was better for me than now.
|
| Less flashy, animated but faster.
| smileson2 wrote:
| I hate this analogy, even things from the rad days like
| vb were better than electron
| simonw wrote:
| Several LLM providers have solid promises that they won't train
| on your inputs to them. OpenAI have this if you are using their
| paid API (though frustratingly not for their paid ChatGPT
| users, at least to my knowledge), and Anthropic have that for
| input to their free apps as well:
| https://support.anthropic.com/en/articles/7996885-how-do-you...
|
| I was hoping I could say the same for Gemini, but unfortunately
| their policy at
| https://support.google.com/gemini/answer/13594961?visit_id=6...
| says "Google uses this data, consistent with our Privacy
| Policy, to provide, improve, and develop Google products and
| services and machine-learning technologies"
|
| My intuition is that Google don't directly train on user
| conversations (because user conversations are full of both junk
| and sensitive information that no model would want to train
| on), but I can't state that with any credibility.
| fhdsgbbcaA wrote:
| I'm sure there's absolutely zero chance that Sam Altman would
| lie about that, especially now that he's gutted all oversight
| and senior-level opposition.
| light_hue_1 wrote:
| Ah yes. Solid promises you can never verify. That companies
| would benefit massively from violating.
|
| That's worth literally nothing.
| choilive wrote:
| It would also destroy these companies if they were ever
| caught lying.
| atq2119 wrote:
| That seems awfully optimistic, given what Sam Altman is
| getting away with transforming the governing structure of
| OpenAI.
| jart wrote:
| Not if the government required them to do it.
| Workaccount2 wrote:
| I know this sounds heretical, but companies generally do
| not go against what they say they are doing. They might use
| clever language or do slimey things, but it's very rare
| that they will say "We do not do xyz" while they are in
| fact doing xyz. Especially for big companies.
|
| Reputation has far more value than whatever they gain by
| lying. Besides, they can just say "We do xyz" because <1%
| of users read the TOS and less than <0.1% care enough to
| not use the service.
| pton_xd wrote:
| This is supremely naive, in my opinion.
|
| Big companies not only lie, some of them do so routinely,
| including breaking the law. Look at the banking industry:
| Wells Fargo fraudulent / fake account scandal, JPMorgan
| Chase UST and precious metals future fraud. Standard
| Charter bank caught money laundering for Iran, twice.
| Deutsche Bank caught laundering for Russia, twice. UBS
| laundering and tax evasion. Credit Suisse caught
| laundering for Iran. And so on.
|
| Really it comes down to what a company believes it can
| get away with, and what the consequences will be. If
| there are minimal consequences they'd be dumb not to try.
|
| Oh I just remembered a funny one: remember when it came
| out that Uber employees were using "God view" to spy on
| ex-partners, etc? For years. Yeah I'm pretty sure the TOS
| didn't have a section "Our employees may, from time to
| time, spy on you at their discretion." Actually the
| opposite, Uber explicitly said they couldn't access ride
| information for its users.
| startupsfail wrote:
| The company can certainly make a calculated risk of going
| against their TOS and their promise to the customers at
| the cost of potential risk of their reputation.
|
| Note that such reputation risks are external and
| internal. The reputation reflects on the executive team
| and there is a risk that the executive team members may
| leave or attempt to get the unscrupulous employee fired.
| blooalien wrote:
| > Google: "Don't Be Evil" is our motto!
|
| > Also Google: "Let's do _all_ the evil things... " ~
| heavily "paraphrased" ;)
|
| My "tongue-in-cheek point" is that it seems like
| corporations beyond a certain point of "filthy-richness"
| just do as they please, and say what they please, and
| mostly neither thing has to agree with the other, nor
| does either one need affect their profits "bottom line"
| all that seriously much. Most of your typical "mega-
| corps" are really only able to be affected much by the
| laws and legal system, which they've been increasingly
| "capturing" in various ways so that happens very rarely
| anymore these days, and when it does it's most often a
| "slap on the wrist" and "don't do that!" sorta thing,
| followed by more business-as-usual.
|
| You know the old worry about the "paperclip production
| maximizer AI" eating everything to create paperclips?
| That's kinda where we're pretty-much _already_ at with
| mega-corps. They 're so utterly laser-focused on
| maximizing to extract every last dime of profit out of
| _everything_ that they 're gonna end up literally
| consuming all matter in the universe if they don't just
| destroy us all in the process of trying to get there.
| JCharante wrote:
| tbh your data would be too unstructured, it's not really being
| used to train unless you flag it deliberately with a feedback
| mechanism
| JCharante wrote:
| > AI models now require trainers with advanced degrees
|
| Companies that create data for FM (foundational model) companies
| have been hiring people with degrees for years
|
| > Invisible Tech employs 5,000 specialized trainers globally
|
| Some of those companies have almost a million freelancers on
| their platforms, so 5k is honestly kinda medium sized.
|
| > It takes smart humans to avoid hallucinations in AI
|
| Many smart humans fail at critical thinking. I've seen people
| with masters fail at spotting hallucinations in elementary level
| word problems.
| aleph_minus_one wrote:
| > Many smart humans fail at critical thinking. I've seen people
| with masters fail at spotting hallucinations in elementary
| level word problems.
|
| This is like lamenting that a person who has a doctoral degree,
| say, in mathematics or physics often don't have a more than
| basic knowledge about, for example, medicine or pharmacy.
| JCharante wrote:
| It depends on your definition of smart. I think that holding
| a degree != smart.
| aniviacat wrote:
| > I think that holding a degree != smart.
|
| Does this mean that these two sentences are completely
| unrelated and only happen to exist within the same
| paragraph?
|
| > Many smart humans fail at critical thinking. I've seen
| people with masters [...]
|
| I've understood you to use "people with masters" as an
| example of "smart humans".
| JCharante wrote:
| Well the article began by talking about how before these
| data training companies would just hire generalists for
| $2/hr, but now they're hiring degree holders. And it
| mentions that smart people will be necessary. I'm just
| saying that degree holding != smart and it's a trap that
| those data training companies have to avoid.
| aleph_minus_one wrote:
| > It depends on your definition of smart. I think that
| holding a degree != smart.
|
| You wrote:
|
| > I've seen people with masters fail at spotting
| hallucinations in elementary level word problems.
|
| I wanted to express that having a master in some (even
| complicated) subject does not make you a master at [pun
| intended] spotting hallucinations. To give evidence for
| this statement, I gave a different, more down-to-earth
| example of a similar situation.
| JCharante wrote:
| It was a math problem, like this.
|
| Q: A farmer has 72 chickens. He sells 15 chickens at the
| market and buys 8 new chicks. Later that week, a fox
| sneaks into the coop and eats 6 chickens. How many
| chickens could the farmer sell at the market tomorrow?
|
| AI Answer: The farmer started with 72 chickens. After
| selling 15, he had 57 chickens left. Then he bought 8 new
| chicks, bringing the total to 65. Finally, the fox ate 6
| chickens, so we subtract 6 from 65. This gives us 59
| chickens. Therefore, the farmer now has 59 chickens that
| he could sell at the market tomorrow.
|
| --
|
| You'd expect someone who can read/understand proofs to be
| able to spot a a flow in the logic that it takes longer
| than 1 week for chicks to turn into chickens.
| aleph_minus_one wrote:
| > You'd expect someone who can read/understand proofs to
| be able to spot a a flow in the logic that it takes
| longer than 1 week for chicks to turn into chickens.
|
| Rather, I'd assume that someone who is capable of
| spotting the flow in the logic has a decent knowledge of
| the English language (in this case referring to the
| difference in meaning between "chick" and "chicken").
|
| Many people who are good mathematicians (i.e. capable of
| "reading/understanding proofs" as you expressed it) are
| not native English speakers or have a great L2 level of
| English.
| Viliam1234 wrote:
| But I was told that humans have this thing called
| "general intelligence", which means they should be
| capable to do both math _and_ English!
|
| If an AI made a similar mistake, people would laugh at
| it.
| aleph_minus_one wrote:
| > But I was told that humans have this thing called
| "general intelligence", which means they should be
| capable to do both math and English!
|
| You confuse "intelligence" with "knowledge". To keep to
| your example: there exist quite a lot of highly
| intelligent people on earth who don't or barely know
| English.
| jonahx wrote:
| When an educated person misses this question, it's not
| because the temporal logic is out of their reach. It's
| because they scanned the problem and answered quickly.
| They're pattern matching to a type of problem that
| wouldn't include the "tomorrow/next week" trick, and then
| giving the correct answer to that.
|
| Imo it's evidence that humans make assumptions and aren't
| always thorough more than evidence of smart people being
| unable to perform elementary logic.
| echoangle wrote:
| As a layman, i have no clue at what point a chick turns
| into a chicken. I also think this isn't even answerable,
| because ,,new chick" doesn't really imply ,,newborn" but
| only means ,,new to the farmer", so the chicks could be
| at an age where they would be chickens a week later, no?
| visarga wrote:
| > This is like lamenting that a person who has a doctoral
| degree, say, in mathematics or physics often don't have a
| more than basic knowledge in, for example, medicine or
| pharmacy.
|
| It was word problems not rocket science. That tells a lot
| about human intelligence. We're much less smart than we
| imagine, and most of our intelligence is based on book
| learning, not original discovery. Causal reasoning is based
| on learning and checking exceptions to rules. Truly novel
| ideation is actually rare.
|
| We spent years implementing transformers in a naive way until
| someone figured out you can do it with much less memory
| (FlashAttention). That was such a face palm, it was a trivial
| idea thousands of PhDs missed. And the code is just 3 for
| loops, with a multiplication, a sum and an exponential. An
| algorithm that fits on a napkin in its abstract form.
| beepbooptheory wrote:
| Doesn't this lead you to, perhaps, question the category
| and measure of "intelligence" in general, especially how it
| is mobilized in this kind of context? Like this very angle
| does a lot to point out the contradictions in some
| speculative metaphysical category of "intelligence" or
| "being smart," but then you just seem to accept it in this
| particular kind of fatalism.
|
| Why not take away from this that "intelligence" is a word
| that obtains something relative to a particular society,
| namely, one which values some kind of behavior and speech
| over others. "Intelligence" is something important to
| society, its the individual who negotiates (or not) the way
| they think and learn with what this particular signifier
| connects with at a given place and time.
|
| Like I assume you don't agree, but just perhaps if we use
| our "intelligence" here we could maybe come to some
| different conclusions here! Everyone is just dying to be
| like mid-20th century behaviorist now, I just don't
| understand!
| klabb3 wrote:
| > And the code is just 3 for loops, with a multiplication,
| a sum and an exponential.
|
| All invented/discovered and formalized by humans. That we
| found so much (unexpected) power in such simple
| abstractions is not a failure but a testament to the
| absolute ingenuity of human pursuit of knowledge.
|
| The mistake is we're over-estimating isolated discoveries
| and underestimating their second order effects.
| dilawar wrote:
| I think many people like to believe that solving puzzles will
| somehow make them better at combinatorics. Lateral skill
| transfer in non-motor skills e.g. office works, academics
| works etc may not be any better than motor skills. It's
| easier to convince people that playing soccer everyday
| wouldn't make them any better at cricket, or even hockey.
| sudosysgen wrote:
| But motor skills transfer extremely well. It's not uncommon
| for professional athletes to switch sports, some even
| repeatedly.
| Der_Einzige wrote:
| There's some famous ass basketball players with mediocre
| but still existent MLB careers.
| jononor wrote:
| Wealth, network and fame transfers incredibly well
| between fields. Possibly better than anything else. It
| should be accounted for when reasoning about success in
| disparate fields. In addition to luck, of course.
| thatcat wrote:
| Kobe Bryant played soccer, Michael Jordan played baseball,
| Lebron played football.. it actually makes you even better
| because you learn non traditional strategies to apply to
| the other sport you're playing.
| 39896880 wrote:
| All the models do is hallucinate. They just sometimes
| hallucinate the truth.
| SamGyamfi wrote:
| There is a cost-quality tradeoff companies are willing to make
| for AI model training using synthetic data. It shows up fairly
| often with AI research labs and their papers. There are also
| upcoming tools that remove the noise that would trip up some
| advanced models during annotation. Knowing this, I don't think
| the "human-labeled data is better" argument will last that long.
| yawnxyz wrote:
| "raw dogging" non-RLHF'd language models (and getting good and
| unique output) is going to be a rare and sought-after skill soon.
| It's going to be a new art form
|
| someone should write a story about that!
| theptip wrote:
| I feel this is one of the major ways that most pundits failed
| with their "the data is going to run out" predictions.
|
| First and foremost a chatbot generates plenty of new data (plus
| feedback!), but you can also commission new high-quality content.
|
| Karpathy recently commented that GPT-3 needs so many parameters
| because most of the training set is garbage, and that he expects
| eventually a GPT-2 sized model could reach GPT-3 level, if
| trained exclusively on high-quality textbooks.
|
| This is one of the ways you get textbooks to push the frontier
| capabilities.
| from-nibly wrote:
| At a good cost though? Last time I checked generating good data
| costs a tiny bit more than an http request to somebody elses
| website.
| recursive wrote:
| It kind of seems like it got dumber to me. Maybe because my first
| exposure to it was so magical. But now, I just notice all the
| ways it's wrong.
| bdjsiqoocwk wrote:
| Submarine article placed by Cohere. Wtf is cohere.
| wlindley wrote:
| a/k/a It is all a clever scam. True, or true?
___________________________________________________________________
(page generated 2024-09-28 23:00 UTC)