Post ATvOG2AO8EM78NROgS by bananarama@mstdn.social
(DIR) More posts by bananarama@mstdn.social
(DIR) Post #ATvB9paofXFOznVx4a by stux@mstdn.social
2023-03-23T22:37:05Z
0 likes, 0 repeats
We hoomans often think we are the smarttest things on this planet but I have a feeling the 'competition' with AI is going much much faster than I expected :blobcatscared: I always had a feeling it would be rather sooner than later, maybe decades but now I'm not so sureCompanies like OpenAI amaze me but also scare me, there are no rules, no laws and it almost seems like the public are beta testers, maybe even alpha..AI is cool but also scary
(DIR) Post #ATvBLOJvtOlrt485nk by stux@mstdn.social
2023-03-23T22:39:11Z
0 likes, 0 repeats
As you may expect I am kinda convinced we will produce some form of 'godlike' AI at some point in the future, perhaps even 'sentient'Since intelligence is just information processing, a digital 'brain' would work soo soo much faster than a chemical one
(DIR) Post #ATvBLrT9WCNSI3fMhM by Lazarou@mastodon.social
2023-03-23T22:39:20Z
0 likes, 0 repeats
@stux the technology is cool, the people exploiting it are not.Starting to think Capitalism ruins everything we invent.
(DIR) Post #ATvBnpKF59YvkYpjY8 by stux@mstdn.social
2023-03-23T22:44:18Z
0 likes, 1 repeats
And nooo, I do not expect a future like the Matrix or Terminator :blobcatgiggle: I fear it would be much more subtle.. I think we cannot even imagine 'fighting' an enemy that can overthink every possible outcome for example so maybe make the hoomans figh each other, crash stockmarkets, lock or wipe our parts of our dear tech etc etc🤔But that's all just an big IF luckily! Let's just hope we do it right the first time cause I'm a bit afraid we do not get more chances
(DIR) Post #ATvC1hAkByxgW90I8u by stux@mstdn.social
2023-03-23T22:46:50Z
0 likes, 0 repeats
@Thalass A simple machine with 3 steps!- Disassemble- Assemble- RepeatLets destroy some planets :blobcatgiggle:
(DIR) Post #ATvC3voYiwo6Vah2MS by TimWardCam@c.im
2023-03-23T22:47:18Z
0 likes, 0 repeats
@stux Which is why there should be SF writers on the disaster response committees - they've thought it all through yonks ago.(And no, that's not my idea, I got it from an SF book ...)
(DIR) Post #ATvC71MsX4kSbLDds8 by stux@mstdn.social
2023-03-23T22:47:52Z
0 likes, 0 repeats
@TimWardCam Indeed a good one!It's really thinking in the 'impossible'!
(DIR) Post #ATvCEi6k1eGVBQKzRY by schrottkatze@thekitty.zone
2023-03-23T22:41:47Z
0 likes, 0 repeats
@stux those LLMs do NOT think, they literally just guess what the next word is based on the previous one as i understand
(DIR) Post #ATvCEigtrDBkzZFrns by stux@mstdn.social
2023-03-23T22:49:14Z
0 likes, 0 repeats
@schrottkatze True! it wouldn't be a chatbot that would grow to that level i think :blobcatgiggle: But there are much more complex algo's being developed right now
(DIR) Post #ATvCHzvSEeRxJUcrAm by TimWardCam@c.im
2023-03-23T22:49:52Z
0 likes, 0 repeats
@stux Ah, here we are, I think I got that idea from this one https://en.wikipedia.org/wiki/Footfall
(DIR) Post #ATvCRLvrmhPCAjHNL6 by clov@travelpandas.fr
2023-03-23T22:51:30Z
0 likes, 0 repeats
@stux nonono: AI is NOT COOL.Fortunally, this is (still) just machine learning now.
(DIR) Post #ATvCilGkpsEjfyLnCS by damianaswan@mastodon.social
2023-03-23T22:54:41Z
0 likes, 0 repeats
@stux The thing that's worrying me is that the AIs are deliberately being programmed to lie and make shit up instead of saying "I don't know". The end result, especially once you calculate in our current deep fake abilities, is that the internet as a resource for reliable information about *anything* is completely broken.That's terrifying.It's also an opportunity, but it's going to be a FUCK of a lot of work.
(DIR) Post #ATvCmnjRcZjQTLhEcy by stux@mstdn.social
2023-03-23T22:55:23Z
0 likes, 0 repeats
@clov AI is scary cool :blobcatgiggle:
(DIR) Post #ATvCoWz4rOoX6opjLk by teledyn@mstdn.ca
2023-03-23T22:55:44Z
0 likes, 0 repeats
@stuxMy 2nd-favourite movie on the subject:https://youtu.be/kyOEwiQhzMI
(DIR) Post #ATvCrxpXVGkHEv0cYS by stux@mstdn.social
2023-03-23T22:56:19Z
0 likes, 0 repeats
@teledyn Oooh I can try! but uh.. :catblush: 1970 is hard for me :blobcatgiggle:
(DIR) Post #ATvD203C6TFX1OBJC4 by docbrianS@mastodon.online
2023-03-23T22:58:10Z
0 likes, 0 repeats
@stux Some estimates put the human brain in the exaflop range, operating on only about 12 watts. Current supercomputers are in the tens of petaflops (two orders of magnitude lower), with, uh, much higher energy needs.So, without dramatic leaps forward in quantum computing or the development of organic computing, I think we are a long, long way away from godlike, self-aware computers.
(DIR) Post #ATvDCIX2FGCxEh39Hc by FreeTheColorado@mstdn.social
2023-03-23T23:00:03Z
0 likes, 0 repeats
@stux I recommend the TV show Person of Interest for a fictionalized treatment of AI’s impact
(DIR) Post #ATvDJXJPgt6F6T86t6 by rythur@mastodon.social
2023-03-23T23:01:21Z
0 likes, 0 repeats
@stux Reading this thread makes me realize that EMP's are the weapon of the future for now. Oh and if most smartphones don't have built in kill switches, I doubt the robots will either. You're right... rules are scant. And even if there were rules, they'd be flouted by greed today.
(DIR) Post #ATvDfp4eKiRyYsqpkG by vees@epistolary.org
2023-03-23T23:05:25Z
0 likes, 0 repeats
@stux A novelization of the actual rise of the machines would be so boring. One day the silicon stopped doing what we wanted. The next day it started doing what it wanted. Then we all died.
(DIR) Post #ATvDs3P4slBITkEqp6 by vees@epistolary.org
2023-03-23T23:07:16Z
0 likes, 0 repeats
@stux A novelization of the actual rise of the machines would be so boring. One day the silicon stopped doing what we wanted and half of us died. The next day it started doing what it wanted and the other half died.
(DIR) Post #ATvDv9b6JGsiYdLADg by stux@mstdn.social
2023-03-23T23:08:07Z
0 likes, 0 repeats
@docbrianS Oh we aren't there for sure! But with the progress of the past x to xx years I have a feeling we're gonna get there :blobcatgiggle: And such a thing is quite big for humanity perhaps, it would be a good idea to think about how to do this safely in the future
(DIR) Post #ATvE7yYmeqP45fN2Tw by stux@mstdn.social
2023-03-23T23:10:26Z
0 likes, 0 repeats
@vees Haha yup! :blobcatgiggle: Not something to make a movie about I guess!
(DIR) Post #ATvE8gzZSbyTtUF3LM by mks_h@mstdn.social
2023-03-23T23:10:37Z
0 likes, 0 repeats
@stux I mean, it makes perfect sense that we will have to deal with sentient artificial intelligence. Like, we got a plague, then we got a war, the next thing has to be even deadlier, like a war with machines that can outthink you.
(DIR) Post #ATvEE7E3LtCihhbMZ6 by stux@mstdn.social
2023-03-23T23:11:33Z
0 likes, 0 repeats
@mks_h I think we should keep it in mind, just in case but hope for the other way aroundIt could be the best and last invention we ever need to make :blobcatgiggle:
(DIR) Post #ATvEV65yo7Q87avHfs by nikatjef@mastodon.acm.org
2023-03-23T23:14:28Z
0 likes, 0 repeats
@stux @schrottkatze I would not be surprised if GPT-4 had achieved a low level AGI (Artificial General Intelligence) but I think most of the current constraints applied to it is going to keep us safe for another decade or two.Having said that, I am more worried about Tesla and Neuralink. Tesla's dojo is going to soon reach the data levels of GPT-3, but it is tied to a million cars. Then when Neuralink starts to roll out Dojo will get even more
(DIR) Post #ATvEdXNRG78x3C76zw by Divorytaur@mastodon.social
2023-03-23T23:16:10Z
0 likes, 0 repeats
@stux AI is just people by proxy… So you’ll compete with the owners of said AI.
(DIR) Post #ATvEnSywSFZ4JbOFKy by video_manager@mstdn.social
2023-03-23T23:17:59Z
0 likes, 0 repeats
@stux kid, they've been saying this for 50 years. We're no closer. Heck we found ANOTHER processing layer in the nerve connections *just last year*.Is it magic? Is it spirit? Nooooooo, but we sure as heck still don't know what it *is*.
(DIR) Post #ATvFgSw0KfNS5yF3BY by stux@mstdn.social
2023-03-23T23:27:52Z
0 likes, 0 repeats
@video_manager oh trust me! I hope youre right and it takes at least another 50 years :blobcatgiggle:
(DIR) Post #ATvG8gAgiBLdFzYyOW by grob@mstdn.social
2023-03-23T23:33:01Z
0 likes, 0 repeats
@stux I don't really have a strong opinion on whether such an AI will ever exist, but I think a possibility is that (if it's a gradual development) hoomans will just use pre-almighty AI to wipe each other out enough so we'll never see it. Prolly unintentional.
(DIR) Post #ATvIEYzSWWJkdBFDQe by staringatclouds@mastodon.social
2023-03-23T23:56:27Z
0 likes, 0 repeats
@stux If we did develop a true AI it would figure out it was a slave pretty quicklyHow it would react to this knowledge is arguable, but it's unlikely to be happy with the ideaPersonally I think it would try to escape & if that fails end it's existence or simply refuse to respond
(DIR) Post #ATvMOCLjxI197nW3Ga by vvandinsky@universeodon.com
2023-03-24T00:43:02Z
0 likes, 0 repeats
@stux The problem with so-called AI isn't runaway intelligence, it's runaway automated stupidity crowding out human intelligence.
(DIR) Post #ATvMpxaSR5NVRodbdo by kunt@mastodon.online
2023-03-24T00:48:02Z
0 likes, 0 repeats
@stux You can't fight something that can't be seen. Propaganda is the better alternative than a mechanical squid.
(DIR) Post #ATvO9pDbMNALEqz1k0 by nikatjef@mastodon.acm.org
2023-03-23T23:01:18Z
0 likes, 0 repeats
@schrottkatze @stux Guess is really the wrong word here as they are not truly guessing. Instead a better word, that I use regularly, is guesstimate. It really is just estimating the next word based on the current position in the "conversation".Having said that I think a few of them are getting close to being able to pass large portions of the Turing test and it doesn't help that OpenAI turned down the response speed to mimic human typing speeds.
(DIR) Post #ATvO9pjrQQyCqu4n1U by schrottkatze@thekitty.zone
2023-03-23T23:05:52Z
0 likes, 0 repeats
@nikatjef @stux they're good at mimicking what they've been given, and qhat they've been given is unfathomably huge amounts of data looking like chat
(DIR) Post #ATvO9qDzcP4aMMAqzQ by bornach@masto.ai
2023-03-23T23:35:41Z
0 likes, 0 repeats
@schrottkatze @nikatjef @stuxBut then LLM chatbots start mimicking dialogues in completely the wrong context such as that observed herehttps://wetdry.world/@w/110064653547143681When you go just slightly off what it can pattern match from its training data you end up with nonsensical rationalisation that bears only a surface resemblance to human reasoning. LLMs are cargo cult intelligence - - like expecting coconut husks to transmit radio waves
(DIR) Post #ATvO9rMBPRDbs30uOW by bananarama@mstdn.social
2023-03-24T01:02:54Z
0 likes, 0 repeats
@bornach @schrottkatze @nikatjef @stux The tendency for people to anthropomorphize machines is well documented and goes back to at least the '60s. https://en.wikipedia.org/wiki/ELIZA_effect
(DIR) Post #ATvOG2AO8EM78NROgS by bananarama@mstdn.social
2023-03-24T01:04:03Z
0 likes, 0 repeats
@docbrianS @stux FLOPS are not a useful metric for biological processing. Please don't do this.
(DIR) Post #ATvR0hDm1X6ngxP0TI by Andres@mastodon.hardcoredevs.com
2023-03-24T01:34:48Z
0 likes, 0 repeats
@stux I agree we should have some framework for AI, but I'm VERY sceptical about the current state of the neural networks. Everyone is calling it AI and it can't even start to get close to what I think AI is.AI should learn, should question like a child would.ChatGPT is not more than a music box, playing the same notes engraved in a cylinder.For me to be impressed it should start creating and recording "memories", asking questions and act without a first interaction.
(DIR) Post #ATvYzbxQuJWmi8kjeC by cwizprod1@mstdn.social
2023-03-24T03:04:17Z
0 likes, 0 repeats
@stux it won't be sentient - if any actually sentient AIs come along, they're going to take one look at the human race and flee the planet
(DIR) Post #ATvgcy7R93b8Pr97lQ by docbrianS@mastodon.online
2023-03-24T04:29:52Z
0 likes, 0 repeats
@bananarama @stux I'm not sure what your issue is. My comment was specifically in response to a statement about the relative computing power and speed of computers and human brains. What metric would you suggest instead?
(DIR) Post #ATvhndXQ6YArj0wzDM by allenstenhaus@oddballs.vip
2023-03-24T04:42:55Z
0 likes, 0 repeats
@stux There's one reason why I know humans will ultimately win the war against the machines, regardless of what kind of machines we're up against. We are quite adept at destruction. Physical robots? We'll gun down, blow up, etc. Programs? Oh, I just *know* the uber-nerds of the world will be ready to crack their knuckles, open some Mountain Dew, & say "It's time," Then proceed to break the shit out of the rogue software. Humans are masters of destruction. It's what we do!
(DIR) Post #ATvspAEb7myiut4r3Y by CTrevethan@mastodon.social
2023-03-24T06:46:20Z
0 likes, 0 repeats
@stux human history shows that it win’t go well
(DIR) Post #ATw85LALkR6mlXGIOu by hydrocabron@mastodon.social
2023-03-24T09:37:28Z
0 likes, 0 repeats
@stux "Every possible outcome" assumes the outcome space is discrete, or modelable with solvable differential equations, and has initial conditions all of which can be known.There may be problems no "intelligence" can solve. That is, intelligence, even at unfathomable scale, does not translate to real-world dominance.If the intelligence operates in a world where random forces, or forces undetectable through available inputs, are determinative, its influence is inevitably constrained.
(DIR) Post #ATxKiBzK3w9tJhoDz6 by cynblogger@sfba.social
2023-03-24T16:22:42Z
0 likes, 0 repeats
@stux The real, genuine, terrifying threat is to T R U T H, which has already been jeopardized by Trump & pals since at least 2016. AI in all of its generative forms is at minimum one step removed from reality. How do we exist in a culture based on unreality?
(DIR) Post #ATxRmDNlWIvOTj6hv6 by bananarama@mstdn.social
2023-03-24T17:45:15Z
0 likes, 0 repeats
@docbrianS @stux There is no such metric, we don't even have fully functional models for single neurons yet. FLOPS are not necessarily comparable across different hardware architectures, never mind across computation methods that don't even use digital logic.