this post was submitted on 07 Oct 2024
117 points (100.0% liked)
Technology
37720 readers
145 users here now
A nice place to discuss rumors, happenings, innovations, and challenges in the technology sphere. We also welcome discussions on the intersections of technology and society. If it’s technological news or discussion of technology, it probably belongs here.
Remember the overriding ethos on Beehaw: Be(e) Nice. Each user you encounter here is a person, and should be treated with kindness (even if they’re wrong, or use a Linux distro you don’t like). Personal attacks will not be tolerated.
Subcommunities on Beehaw:
This community's icon was made by Aaron Schneider, under the CC-BY-NC-SA 4.0 license.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
AGI is inevitable unless:
General intelligence is substrate independent and what the brain does cannot be replicated in silica. However, since both are made of matter, and matter obeys the laws of physics, I see no reason to assume this.
We destroy ourselves before we reach AGI.
Other than that, we will keep incrementally improving our technology and it's only a matter of time untill we get there. May take 5 years, 50 or 500 but it seems pretty inevitable to me.
@ContrarianTrail @JRepin well I guess somebody would first need to clearly define what "AGI" is. Currently it's just "whatever the techbro hypers want it to be".
And then there's the matter (ha!) of your assumption that we understand all laws of physics necessary that "matter obeys", or that we can reasonably understand them. That's a pretty strong assumption: individual human minds are pretty limited and communication adds overhead, and we might reach a point where we're stuck.
A chess engine is intelligent in one thing: playing chess. That narrow intelligence doesn’t translate to any other skill, even if it's sometimes superhuman at that one task, like a calculator.
Humans, on the other hand, are generally intelligent. We can perform a variety of cognitive tasks that are unrelated to each other, with our only limitations being the physical ones of our "meat computer."
Artificial General Intelligence (AGI) is the artificial version of human cognitive capabilities, but without the brain's limitations. It should be noted that AGI is not synonymous with AI. AGI is a type of AI, but not all AI is generally intelligent. The next step from AGI would be Artificial Super Intelligence (ASI), which would not only be generally intelligent but also superhumanly so. This is what the "AI doomers" are concerned about.
@ContrarianTrail
> A chess engine is intelligent in one thing: playing chess
No. That's not how the adjective "intelligent" works, outside of marketing drivel of course ("intelligent washing machine" etc).
> Artificial General Intelligence (AGI) is the artificial version of human cognitive capabilities
Can you give a definition of "intelligence" or "human cognitive abilities" that would allow us to somehow unequivocably establish that "X is intelligent" or "X has human cognitive abilities"?
IIRC, within computer science, which is the field most heavily driving AI design and research forward, an 'intelligent agent' is essentially defined as any 'agent' which takes external stimulai from a collection of sensors in some form of environment, processes that stimulai in a dynamic fashion (one of the criteria IIRC is a branching decision tree based on the stimulai), and then applies that processing to a collection of affectors in the environment.
Yes, this definition is an extremely low bar and includes a massive amount of code, software and scripts. It also includes basic natural intelligences such as worms, ants, amoeba, and even viruses. One example of mechanical AI are some of Theo Jansen's StrandBeasts
"Artificial intelligence" is for the marketing department's benefit. At least mainly so. What people envision with the term AI is because of preconceived notions based science fiction not what it actually is.
@JayDee so two things.
First: sure, we can redefine words in any way we want, but then:
talking about "AI" becomes much less interesting if it merely means "walking a decision tree based on data coming from external sensors"
the whole talk about "intelligence" becomes a bait-and-switch, as the conversation started with the term "intelligence" being used in the general sense we tend to apply to people and some higher-order animals.
I am not bait-and-switching here. The switchers were the business-minded grifters which made the term synonymous with LLMs and eventually destroyed its meaning completely.
The definition I gave is from the most popular and widely used CS textbook on AI and has been the meaning used in the field since the early 90s. It's why videogame NPCs are always called AI, because they fit the conventional CS definition, and were one of the major things it was about the most.
As for your '1', AI is a wide-but-very-specialized field and pertains from everything from robots to text autocomplete. If you want the most out of it, you need to get down into the nitty gritty and really research the field.
On a Seperate note, while AI safety, AGI, and the risk of the intelligence explosion are somewhat related to computer science's pursuit of AI systems, they are much more philosophical currently, and adhere to much vaguer definitions of AI, Such as Alan Turing's.
@JayDee I didn't say you are, I clarified in my later post. Sorry, should have been clearer.
I am vehemently agreeing with you here, in fact.
The context is the conversation above in the thread, where it was claimed that "AGI" is "pretty inevitable".
And the point I've been making is:
we don't have a good definition of what "intelligence" is, in the sense presumably used above;
if we decide to use a somewhat simplistic definition, the whole "AI" issue stops being all that exciting.
@JayDee AI as the wide, specialized field you mention makes no claims about building anything with *actual* human-like intelligence, I feel. People who understand how the math and code work in these systems know better than to do that.
And yes, "AGI" debate is a philosophical one. The problem is it is not recognized as such, because of the AI hype. People seem to think that AGI is "inevitable" and "just around the corner", because salespeople from companies that benefit from that hype say so.
Alright, I see what you're saying now. We're on the same page.
As an additional thing regarding AGI, I think it should be noted that 'human-level' and 'human-like' are importantly distinct when talking about this topic.
In reality, if an AGI is ever created, it will most likely not be human-like at all. Humans think the way we do out of an evolutionary conditioning for survival, a history an AGI will not be coming from. One example given by Robert Miles is a staple making machine becoming an ASI, where it essentially would exist solely to make as many staples as it could with its hyperintelligence.
We mean to say that this AGI is a 'human-level' intelligence in that it can learn to utilize abstractions and tools, be able to function in a large variety of environments without intervention or training, and be able to learn in a realtime fashion.
Obviously, these criteria for any AI shows just how far away we are from achieving anything right now.these concepts are very vague and the arguments for each one's impossibility or inevitability are equally vague and philosophical. It's still mostly just stuffy academics arguing with each other.
One statement I agree with, though, comes from the AI safety collective: We don't know what we're doing, and we should really sort that out. If any of this is actually possible and we accidentally make an AGI/ASI before having any failsafes or contingencies, it could be very bad.
@JayDee
> We don’t know what we’re doing, and we should really sort that out.
True. But the bigger problem is not the mythical and hypothetical "AGI/ASI" stuff that maybe will happen one day, but very real harms already being caused by misuse and misapplication of algorithmic and "AI"-based systems.
So that's what I think we should be focusing on instead.
True, I would say that there's multiple issues dealing with AI that are more pressing:
These aren't all of them. One thing I've noticed, however, is that these aren't really AI-specific issues - these are all issues caused by automation and lack of regulation. This lack of proactive regulation is also very likely a failing of our currently neoliberal government systems.
I think that is why so many AI hype-mongers draw attention towards A(G)I safety, because they don't want attention drawn to the actual danger which is automation safety in general.
You're right that we need a clear definition of intelligence if we are to make any predictions about achieving AGI. The researchers behind this article appear to mean "human-level cognition" which doesn't seem to be a particularly objective or useful yardstick. To begin with, which human are we talking about? If they're talking about an idealised maximally intelligent human, then I don't think we should be surprised that we aren't about to achieve that. The goal is not to recreate human cognition as if that's some kind of holy grail. The goal is to make intelligent systems which can give results which are at least as good as what would be produced by a skilled and well-trained human working on the same problem.
Can I ask you how you would define intelligence? And in particular, how would you - if you would at all - differentiate intelligence from being clever, or from being well educated?
@lightstream I wouldn't, because I am not the one making claims about "AGI" being just around the corner.
That's the thing, OpenAI and others benefiting from the hype make extraordinary claims – along the lines of "human-level AGI is just around the corner" – so they are the ones that need to define their terms.
You are asking all the right questions here ("which human are we talking about"), the point is that these questions should be answered by those who make such extraordinary claims.
I certainly am not surprised that OpenAI, Google and so on are overstating the capabilities of the products they are developing and currently selling. Obviously it's important for the public at large to be aware that you can't trust a company to accurately describe products it's trying to sell you, regardless of what the product is.
I am more interested in what academics have to say though. I expect them to be more objective and have more altruistic motivations than your typical marketeer. The reason I asked how you would define intelligence was really just because I find it an interesting area of thought which fascinates me and has done long before this new wave of LLMs hit the scene. It's also one which does not have clear answers, and different people will have different insights and perspectives. There are different concepts which are often blurred together: intelligence, being clever, being well educated, and consciousness. I personally consider all of these to be separate concepts, and while they may have some overlap, they nevertheless are all very different things. I have met many people who have very little formal education but are nonetheless very intelligent. And in terms of AI and LLMs, I believe that an LLM does encapsulate some degree of genuine intelligence - they appear to somehow encode a model of the universe in their billions of parameters and they are able to meaningfully respond to natural language questions on almost any subject - however an LLM is unquestionably not a conscious being.
@ContrarianTrail @JRepin and finally, there's a question of whether we actually decide to pursue it.
Nuclear power was supposed to be the "inevitable" power source for all of humanity mere 50 years ago. But at some point we decided not to pursue that goal.
Cryptocurrencies were supposed to be "inevitable" replacement for the banking system.
And we *have* cryptocurrencies and nuclear power. These exist. As opposed to whatever nebulous concept hides beneath "AGI".
Since they still exist, only time will tell if the promise of nuclear power and/or cryptocurrencies come to be.
AGI and even IMHO AI do not exist. Whatever product is being marketed as AI isn't what I would consider AI. "AI" can have its uses but I really do not think they will be what people expect because it fundamentally lacks what I would consider crucial aspects of human intelligence.
AI makes for a very good grammar checker. It is good at producing filler content for SEO. And it is good at producing "stuff" that looks like it could be right. Probably will have some uses in creative work since it doesn't have to be "correct" so as a tool to aid an artist, that's seems pretty cool - I'm sure that is already happening. It will have its uses and a lot of companies will find out the hard way, it is not that they think. That's my prediction.
incremental improvements on a dead end, still gets you to the dead end.
Then you need to give me an explanation for why it's a dead end
because, having coded them myself, I am under no illusions as to their capabilities. They are not magic. "just" some matrix multiplications that generate a probability distribution for the next token, which is then randomly sampled.
You seem to be talking about LLMs now and I'm not. LLMs being a dead end is perfectly compatible with what I just said. We'll just try a different approach next then. Even the fact of realising they're a dead end is yet another step towards AGI.
yeah, so that means that it's not incremental improvement on what we have that we need. That will get us nowhere. We need a (as yet unknown) completely different approach. Which is the opposite of incremental improvement.
I didn't say we need to improve on what we have. We just need to keep making better technology which we will keep doing unless we destroy ourselves first.
Did you read the article, or the actual research paper? They present a mathematical proof that any hypothetical method of training an AI that produces an algorithm that performs better than random chance could also be used to solve a known intractible problem, which is impossible with all known current methods. This means that any algorithm we can produce that works by training an AI would run in exponential time or worse.
The paper authors point out that this also has severe implications for current AI, too--since the current AI-by-learning method that underpins all LLMs is fundamentally NP-hard and can't run in polynomial time, "the sample-and-time requirements grow non-polynomially (e.g. exponentially or worse) in n." They present a thought experiment of an AI that handles a 15-minute conversation, assuming 60 words are spoken per minute (keep in mind the average is roughly 160). The resources this AI would require to process this would be 60*15 = 900. The authors then conclude:
"Now the AI needs to learn to respond appropriately to conversations of this size (and not just to short prompts). Since resource requirements for AI-by-Learning grow exponentially or worse, let us take a simple exponential function O(2n ) as our proxy of the order of magnitude of resources needed as a function of n. 2^900 ∼ 10^270 is already unimaginably larger than the number of atoms in the universe (∼10^81 ). Imagine us sampling this super-astronomical space of possible situations using so-called ‘Big Data’. Even if we grant that billions of trillions (10 21 ) of relevant data samples could be generated (or scraped) and stored, then this is still but a miniscule proportion of the order of magnitude of samples needed to solve the learning problem for even moderate size n."
That's why LLMs are a dead end.
But I wasn't talking about LLMs
You literally were LMAO
Literally a direct quote. In what world is this not talking about LLMs?
There's not a single mention of LLM's in my entire post. The argument I'm making there isn't even mine. I heard it from Sam Harris way before LLMs were even a thing.
Yeah, suuuuure you weren't.
Note that the proof also generalizes to any form of creating an AI by training it on a dataset, not just LLMs. But sure, we'll absolutely develop an entirely new approach to cognitive science in a few years, we're definitely not boiling the planet and funneling enough money to end world poverty several times over into a scientific dead end!
There's not a single mention of LLMs in my post. Not one.
Another possibility is that humans just aren't smart enough to figure out AGI. While I'm sure that we will continue incrementally improving technology in some form, it's not at all self-evident that these improvements will eventually add up to AGI.
I get what you're saying but to me, that still just sounds like a timescale issue. I can't think of a scenario where we've improved something so much that there's just absolutely nothing we could improve on further. With AI we only need to reach the point of making it have human-level cognitive capabilities and from there on it can improve itself.
There are a couple of reasons that might not work:
To be clear, most of the arguments I'm making aren't really about AGI specifically but about humanities capability to develop arbitrary in principle feasible technologies in general.
Progress itself isn't inevitable. Just because it's possible doesn't mean that we'll get there, because the history of human development shows that societies can and do stall, reverse, etc.
And even if all human societies tends towards progress, it could still hit dead ends and stop there. Conceptually, it's like climbing a mountain through the algorithm of "if there is a higher elevation near you, go towards that, and avoid stepping downward in elevation." Eventually that algorithm brings you to a local peak. But the local peak might not be the highest point on the mountain, and while it is theoretically possible to have gotten to the other true peak from the beginning, the person who is insistent on never stepping downward is now stuck. Or, it's possible to get to the true peak but it requires climbing downward for a time and climbing up past elevations we've already been to, on paths we hadn't been on. One can imagine a society that refuses to step downward, breaking the inevitability of progress.
This paper identifies a specific dead end and advocates against hoping for general AI through computational training. It is, in effect, arguing that even though we can still see plenty of places that are higher elevation than where we are standing, we're headed towards a dead end, and should climb back down. I suspect that not a lot of the actual climbers will heed that advice.