GPT-4

> What we really need is a model that you can run on your own hardware on site.

we won’t have that until we come up with a better way to fund these things. “””Open””” AI was founded on that idea, had the most likely chance of anyone in reaching it: even going into things with that intent they failed and switched to lock down the distribution of their models, somehow managed to be bought by MS despite the original non-profit-like structure. you just won’t see what you’re asking for for however long this field is dominated by the profit motive.

https://github.com/tatsu-lab/stanford_alpaca

Tada! Literally runs on a raspberry pi (very slowly).

GPT models are incredible but the future is somehow even more amazing than that.

I suspect this will be the approach for legal / medical uses (if regulation allows).

I don’t think on site is going to be necessary. Even the US intelligence community trusts that Amazon isn’t spying on the spies.

But a model that can run on a private cluster is certainly something that there’s going to be demand for. And once that exists there’s no reason it couldn’t be run on site.

You can see why OpenAI doesn’t want to do it though. SaaS is more lucrative.

maybe we implement tokenizer+first layer in Javascript on client side and that is enough to preserve raw data on client side and send to GPT only first layer (which is a vector of float values anyway)

matrix gets decoded into text on the client side in Javascript, so we receive send and receive from chatGPT only vector of floats (obfuscation?)

That model will be out in a few years. GPT-3 175b only took two years until someone trained an open source equivalent that could run on a few gpu devices.

Or using homomorphic encryption. I remember some managing to run inference on encrypted images.

See

- https://www.zama.ai/post/encrypted-image-filtering-using-hom...

- https://news.ycombinator.com/item?id=31933995

- https://news.ycombinator.com/item?id=34080882

- https://news.ycombinator.com/item?id=25786154

- https://news.ycombinator.com/item?id=30626182

> What we really need is a model that you can run on your own hardware on site

So, LLaMA? It's no chat gpt but it can potentially serve this purpose

the problem is that if you steal the weights then you can serve your own gpt4, and it's very hard to prove that what you're serving is actually gpt4. (or you could just start using it without paying ofc)

Agreed. The same data privacy argument was used by people not wanting their data in the cloud. When an LLM provider is trusted with a company’s data, the argument will no longer be valid.

Another way of looking at that is that gpt not being open source so companies can run it on their own clusters is holding it back.

Sounds like an easy problem to solve if this is actually the case.

OpenAI just has to promise they won't store the data. Perhaps they'll add a privacy premium for the extra effort, but so what?

You’re saying it like you found a loophole or something but it’s not a gotcha. Yes, if you manipulate sensitive data you shouldn’t use Google Docs or Photoshop online (I’m not imaginative enough to think of a case where you would put sensitive data in Photoshop online though, but if you do, don’t) or host your emails in the cloud. I’ve worked in a moderate size company where everything was self hosted and it’s never been an issue

Doctor-patient or lawyer-client confidentiality is slightly more serious a matter than your examples. And obviously it’s one thing for you to decide where to store your own things and another thing for someone else doing it with your confidential data…

Google Docs and Photoshop Online have offline alternatives (and if you ask me, native MS Office is still the golden standard for interoperability of editable documents), and I use neither in my work or personal life.

Email is harder, but I do run my own email server. For mostly network related reasons, it is easier to run it as a cloud VM, but there's nothing about the email protocol itself that needs you to use a centralised service or host it in a particular network location.

These services now have privacy and legally complaint options nowadays, and decisions to use them get board approval.

OpenAI just simply does not offer the same thing at this time. You’re stuck using Facebook’s model for the moment which is much inferior.

That paper is famously misleading.

It's all the same classic personification of LLMs. What an LLM can show is not the same as what it can do.

The model was already present: in the example game moves. The LLM modeled what it was given, and it was given none other than a valid series of Othello game states.

Here's the problem with personification: A person who has modeled the game of Othello can use that model to strategize. An LLM cannot.

An LLM can only take the whole model and repeat its parts with the most familiar patterns. It is stuck fuzzing around the strategies (or sections of strategy) it has been given. It cannot invent a new divergent strategy, even if the game rules require it to. It cannot choose the winning strategy unless that behavior is what was already recorded in the training corpus.

An LLM does not play games, it plays plays.

That's a great way of describing it, and I think a very necessary and important thing to communicate at this time. A lot of people in this yhread are saying that it's all "just" statistics, but "mere" statistics can give enough info to support inferences to a stable underlying world, and the reasoning about the world shows up in sophisticated associations made by the models.

This special Othello case will follow every discussion from now on. But in reality, a generic, non-specialized model hallucinates early in any non-trivial game, and the only reason it doesn’t do that on a second move is because openings are usually well-known. This generic “model” is still of a statistical nature (multiply all coeffs together repeatedly), not a logical one (choose one path and forget the other). LLMs are cosplaying these models.

To be clear, what they did here is take the core pre-trained GPT model, did Supervised Fine Tuning with Othello moves and then tried to see if the SFT lead to 'grokking' the rules of Othello.

In practice what essentially happened is that the super-high-quality Othello data had a huge impact on the parameters of GPT (since it was the last training data it received) and that impact manifested itself as those parameters overfitting to the rules of Othello.

The real test that I would be curious to see is if Othello GPT works when the logic of the rules are the same but the dimensions are different (e.g., smaller or larger boards).

My guess is that the findings would fall apart if asked about tile "N13".

I tried playing blind chess against ChatGPT and it pretended it had a model of the chess board but it was all wrong.

Also (for those like me who didn't know the rules) generating legal Othello moves requires understanding board geometry; there is no hack to avoid an internal geometric representation:

> https://en.m.wikipedia.org/wiki/Reversi

> Dark must place a piece (dark-side-up) on the board and so that there exists at least one straight (horizontal, vertical, or diagonal) occupied line between the new piece and another dark piece, with one or more contiguous light pieces between them

Turing Completeness is an incredibly low bar and it doesn't undermine this criticism. Conway's Game of Life is Turing Complete, but try writing modern software with it. That Transformers can express arbitrary programs in principle doesn't mean SGD can find them. Following gradients only works when the data being modelled lies on a continuous manifold, otherwise it will just give a statistical approximation at best. All sorts of data we care about lie in topological spaces with no metric: algorithms in computer science, symbolic reasoning in math, etc. If SGD worked for these cases LLMs would push research boundaries in maths and physics or at the very least have a good go at Chollet's ARC challenge, which is trivial for humans. Unfortunately, they can't do this because SGD makes the wrong assumption about how to search for programs in discrete/symbolic/topological spaces.

It's worth emphasizing that "is able to reproduce a representation of" is very much different from "learns".

I assume not, that's why he said 'lossy'.

I don't get why this comment is downvoted. Basically this.

A halving of the costs every year or so seems realistic in this emerging phase.

Sure... But in practice I think a consultant would still provide a higher quality answer. And then, if the bot is not significantly cheaper, what does it matter if it "has more knowledge" in it's network weights?

Further, a consultant couldn’t meaningfully interpret 50 pages in 2 minutes, even with the most cursory skimming.

The stuff people make up in this thread is just ridiculous.

Years? There are already competitors. I just spent all evening playing with Claude (https://poe.com/claude) and it's better than davinci-003.

To be fair it is easy to radically underestimate the rate of progress in this space. Last Wednesday I conservatively opined to a friend "in 10 years we'll all be running these things on our phones". Given that LLaMA was running on a phone a few days later, I may have been a little underoptimistic...

It could take about a year or so.

But I think you should forget about self-hosting at this point, the game is up.

> Yeah, there's an awful lot of power going into private hands

That sounds scary, but what do you mean by "power"? Honest question, I'm fascinated by the discussion about learning, intelligence, reasoning, and so on that has been spawned by the success of GPT.

What "power" do you imagine being wielded? Do you think that power is any more dangerous in "private hands" than the alternatives such as government hands?

Yes. Language and image models are fairly different, but when you look at dall-e 2 (and dall-e earlier) that blew many people's mind when they came out, they have now been really eclipsed in term of popularity by Midjourney and stablediffusion.

Yep

OpenAI doesn't have some secret technical knowledge either. All of these models are just based on transformers

Interesting that it also misunderstood the common misunderstanding in the end.

It reports that people typically think a pound of feathers weighs more because it takes up a larger volume. But the typical misunderstanding is the opposite, that people assume feathers are lighter than bricks.

I'm more surprised that bing indexed this thread within 3 hours, I guess I shouldn't be though, I probably should have realized that search engine spiders are at a different level than they were 10 years ago.

I'm always a bit sceptical of these embarrassing examples being "fixed" after they go viral on social media, because it's hard to know whether OpenAI addressed the underlying cause or just bodged around that specific example in a way that doesn't generalize. Along similar lines I wouldn't be surprised if simple math queries are special-cased and handed off to a WolframAlpha-esque natural language solver, which would avert many potential math fails but without actually enhancing the models ability to reason about math in more complex queries.

An example from ChatGPT:

"What is the solution to sqrt(968684)+117630-0.845180" always produces the correct solution, however;

"Write a speech announcing the solution to sqrt(968684)+117630-0.845180" produces a nonsensical solution that isn't even consistent from run to run.

My assumption is the former query gets WolframAlpha'd but the latter query is GPT itself actually attempting to do the math, poorly.

This is what I saw on a variation of this trick:

(me) > What weighs more, two pounds of feathers or a pound of bricks?

(GPT4)> A pound of bricks weighs more than two pounds of feathers. However, it seems like you might have made an error in your question, as the comparison is usually made between a pound of feathers and a pound of bricks. In that case, both would weigh the same—one pound—though the volume and density of the two materials would be very different.

I think the only difference from parent's query was I said two pounds of feathers instead of two pounds of bricks?

Yep, just tested it - Bing chat gave the correct answer, ChatGPT (basic free model) gave the wrong answer (that they weigh the same).

When the Skynet robots start going door-to-door, just put on your 7-fingered gloves and they will leave you alone.

“One of us!”

It reminds very strongly of the strategy the crew proposes in Star Trek: TNG in the episode "I, Borg" to infect the Borg hivemind with an unresolvable geometric form to destroy them.

That's lame.

Ounces are an ambiguous unit, and most people don't use them for volume, they use them for weight.

Are you using Troy ounces?

>even though an ounce of gold weighs more than an ounce of feathers

Can you expand on this?

What do you think you're doing when you're thinking?

https://www.sciencedirect.com/topics/psychology/predictive-p...

This is obvious, but for some reason some people want to believe that magically a conceptual framework emerges because animal intelligence has to be something like that anyway.

I don't know how animal intelligence works, I just notice when it understands, and these programs don't. Why should they? They're paraphrasing machines, they have no problem contradicting themselves, they can't define adjectives really, they'll give you synonyms. Again, it's all they have, why should they produce anything else?

It's very impressive, but when I read claims of it being akin to human intelligence that's kind of sad to be honest.

It isn’t that simple. There’s a part of it that generates text but it does some things that don’t match the description. It works with embeddings (it can translate very well) and it can be ‘programmed’ (ie prompted) to generate text following rules (eg. concise or verbose, table or JSON) but the text generated contains same information regardless of representation. What really happens within those billions of parameters? Did it learn to model certain tasks? How many parameters are needed to encode a NAND gate using an LLM? Etc.

I’m afraid once you hook up a logic tool like Z3 and teach the llm to use it properly (kind of like bing tries to search) you’ll get something like an idiot savant. Not good. Especially bad once you give it access to the internet and a malicious human.

As far as I know you're not "thinking", you're just generating text.

> It's not thinking, it's simply generating text.

Just like you.

Is there any way to know if the model is "holding back" knowledge? Could it have knowledge that it doesn't reveal to any prompt, and if so, is there any other way to find out? Or can we always assume it will reveal all it's knowledge at some point?

A transformer is a universal approximator and there is no reason to believe it's not doing actual calculation. GPT-3.5+ can't do math that well, but it's not "just generating text", because its math errors aren't just regurgitating existing problems found in its training text.

It also isn't generating "the most likely response" - that's what original GPT-3 did, GPT-3.5 and up don't work that way. (They generate "the most likely response" /according to themselves/, but that's a tautology.)

One area that is really interesting though is that it can interpret pictures, as in the example of a glove above a plank with something on the other end. Where it correctly recognises the objects, interprets them as words then predicts an outcome.

This sort of fusion of different capabilities is likely to produce something that feels similar to AGI in certain circumstances. It is certainly a lot more capable than things that came before for mundane recognition tasks.

Now of course there are areas it would perform very badly, but in unimportant domains on trivial but large predictable datasets it could perform far better than humans would for example (just to take one example on identifying tumours or other patterns in images, this sort of AI would probably be a massively helpful assistant allowing a radiologist to review an order of magnitude more cases if given the right training).

Nearly everything that has been written on the subject is misleading in that way.

People don't write about GPT: they write about GPT personified.

The two magic words are, "exhibit behavior".

GPT exhibits the behavior of "humans writing language" by implicitly modeling the "already-written-by-humans language" of its training corpus, then using that model to respond to a prompt.

The problem with this simplification is a bog standard Markov chain fits the description as well, but quality of predictions is rather different.

Yes the LLM does generate text. No it doesn’t ‘just generate text that’s it’.

I don't know where this comes from because this is literally wrong. It sounds like chomsky dismissing current AI trends because of the mathematical beauty of formal grammars.

First of all, it's a black-box algorithm with pretty universal capabilities when viewed from our current SOTA view. It might appear primitive in a few years, but right now the pure approximation and generalisation capabilities are astounding. So this:

> It sees goat, lion, cabbage, and looks for something that said goat/lion/cabbage

can not be stated as truth without evidence. Same here:

> it's not assigning entities with parameters to each item. It does care about things like sentence structure and what not

Where's your evidence? The enormous parameter space coupled with our so far best performing network structure gives it quite a bit of flexibility. It can memorise things but also derive rules and computation, in order to generalise. We do not just memorise everything, or look up things into the dataset. Of course it learned how to solve things and derive solutions, but the relevant data-points for the puzzle could be {enormous set of logic problems} where it derived general rules that translate to each problem. Generalisation IS NOT trying to find the closest data-point, but finding rules explaining as much data-points, maybe unseen in the test-set, as possible. A fundamental difference.

I am not hyping it without belief, but if we humans can reason then NNs can potentially also. Maybe not GPT-4. Because we do not know how humans do it, so an argument about intrinsic properties is worthless. It's all about capabilities. Reasoning is a functional description as long as you can't tell me exactly how we do it. Maybe wittgenstein could help us: "Whereof one cannot speak, thereof one must be silent". As long as there's no tangible definition of reasoning it's worthless to discuss it.

If we want to talk about fundamental limitations we have to talk about things like ChatGPT-4 not being able to simulate because it's runtime is fundamentally limited by design. It can not recurse. It can only run only a fixed number of steps, that are always the same, until it has to return an answer. So if there's some kind of recursion learned through weights encoding programs intercepted by later layers, the recursion depth is limited.

One thing you will see soon is forming of cults around LLMs, for sure. It will get very strange.

Is it possible to add some kind of self evaluation to the answers given by a model? Like, how confident is it with its answers.

Because it IS wrong.

Just months ago we saw in research out of Harvard that even a very simplistic GPT model builds internalized abstract world representations from the training data within its NN.

People parroting the position from you and the person before you are like doctors who learned about something in school but haven't kept up with emerging research that's since invalidated what they learned, so they go around spouting misinformation because it was thought to be true when they learned it but is now known to be false and just hasn't caught up to them yet.

So many armchair experts who took a ML course in undergrad pitching in their two cents having read none of the papers in the past year.

This is a field where research perspectives are shifting within months, not even years. So unless you are actively engaging with emerging papers, and given your comment I'm guessing you aren't, you may be on the wrong side of the Dunning-Kreuger curve here.

The AI can't reason. It's literally a pattern matching tool and nothing else.

Because it's very good at it, sometimes it can fool people into thinking there is more going on than it is.

How could you prove this?

You would first have to define cognition. These terms often get thrown around. Is an approximation of a certain thing cognition? Only in the loosest of ways I think.

The problem is even if it has this capability, how do you get it to consistently demonstrate this ability?

It could have a dozen internal reasoning networks but it doesn't use them when you want to.

> If, during training, some collection of weights exist along the gradient that approximate cognition

What do you mean? Is cognition a set of weights on a gradient? Cognition involves conscious reasoning and understanding. How do you know it is computable at all? There are many things which cannot be computed by a program (e.g. whether an arbitrary program will halt or not)...

Stop worshipping the robot.

It's kind of sad.

Just because GPT exhibits a behavior does not mean it performs that behavior. You are using those weasel words for a very good reason!

Language is a symbolic representation of behavior.

GPT takes a corpus of example text, tokenizes it, and models the tokens. The model isn't based on any rules: it's entirely implicit. There are no subjects and no logic involved.

Any "understanding" that GPT exhibits was present in the text itself, not GPT's model of that text. The reason GPT can find text that "makes sense", instead of text that "didn't make sense", is that GPT's model is a close match for grammar. When people wrote the text in GPT's corpus, they correctly organized "stuff that makes sense" into a string of letters.

The person used grammar, symbols, and familiar phrases to model ideas into text. GPT used nothing but the text itself to model the text. GPT organized all the patterns that were present in the corpus text, without ever knowing why those patterns were used.

thats because people anthropormophize literally anything, and many treat some animals as if they have the same intelligence as humans. GPT has always been just a charade that people mistake for intelligence. Its a glorified text prediction engine with some basic pattern matching.

Yeah, calling AI a "token predictor" is like dismissing human cognition dumb "piles of electrical signal transmitters." We don't even understand our minds, let alone what constitutes any mind, be it alien or far simpler than ours.

Simple != thoughtless. Different != thoughtless. Less capable != thoughtless. A human black box categorically dismissing all qualia or cognition from another remarkable black box feels so wildly arrogant and anthropocentric. Which, I suppose, is the most historically on-brand behavior for our species.

Sorry but this is the reason it’s unable to solve the parents puzzle. It’s doing a lot but it’s not logically reasoning about the puzzle, and in this case it’s not exhibiting logical behaviour in the result so it’s really obvious to see.

Eg when solving this puzzle you might visualise the lion/goat/cabbage, and walk through the scenarios in your head back and forth multiple times until you find a solution that works. A LLM won’t solve it like this. You could ask it to, and it will list out the scenarios of how it might do it, but it’s essentially an illusion of logical reasoning.

> In order to resolve this, ChatGPT's training corpus needs to contain a "correct answer" next to every unique permutation of every question.

This is not quite the right understanding of how ChatGPT works. It's not necessary to show ChatGPT an example of every possible permutation of an animal crossing puzzle in order for it to solve one it has never seen before. That's because the neural network is not a database of recorded word probabilities. It can instead represent the underlying logic of the puzzle, the relationships between different animals and using this abstract, pared down information, extrapolate the correct answer to the puzzle.

I see the failure in the example with the goat the lion and the cabbage as simply a matter of overfitting.

Edit: I see a lot of people saying "it doesn't understand logic; it's just predicting the next word."

I'm basing my understanding on this video:

https://youtu.be/viJt_DXTfwA

The claim is that it would be impossible to feed enough input into a system such that it could produce anything as useful as ChatGPT unless it was able to abstract the underlying logic from the information provided. If you consider the he number of permutations of the animal crossing puzzle this quickly becomes clear. In fact it would be impossible for ChatGPT to produce anything brand new without this capability.

> GPT does not model subjects. GPT does not even model words! It models tokens.

The first and last layers of a transformer decoder model tokens. The hidden layers don't have this restriction. There was a paper recently showing that the hidden layers actually perform mesa-optimization via something like backprop. There's absolutely no reason to believe they are not capable of world modeling. In fact all evident suggests they do, in fact, do world modeling.

This pov ignores a lot of the emergent theory of mind and world model building research that suggests LLMs may possess a form of rudimentary reasoning ability.

https://www.lesswrong.com/posts/sbaQv8zmRncpmLNKv/the-idea-t...

> GPT does not model subjects. GPT does not even model words! It models tokens.

Someone hasn't read the Othello GPT work out of Harvard a few months back...

It can do some thinking. You can give it instructions to modify a piece of code that definitely isn't on the internet with several steps and it attempts to follow instructions, which, for a human, requires formulating what steps to take.

The prompts have to read like good written requirements for something, so they have some degree of specificity.

But the fact that it can follow instructions and carry them out almost certainly could be considered some form of thinking, especially on novel text not on the internet.

It is a internet-commenter-simulator, exactly what the world needs right now /s

No. It is modelling the various text generation processes that lead to the contents of the internet. Some of that modelling could absolutely involve "thinking", for processes that involve human thinking.

It just weighs it as being more likely that you meant for the lion not to be left alone with the goat, and that the cabbage probably has nothing to fear from the lion.

What’s more likely- you crafted an intentionally misleading puzzle to trick it, or you made a typo or copy paste error?

> it can do math problems, but maybe it can only do "standard" math and logic problems?

That describes many of my classmates, and myself in classes I was bad at.

> what if we pair GPT-4 with an actual reasoning engine? What do we get then?

At best, decreased error rate in logic puzzles and questions.

They will claim it does amazing stuff all the time ? It’s a company

It assumes this character by default. I asked several AI engines (via poe.com, which includes ChatGPT) to compute Galois groups of polynomials like x^5+x+1 and a couple of others, and in each case got not only a wrong answer, but a total non sequitur reasoning.

> In order to show lack of understanding you need to provide a clear test that ChatGPT would fail but a human would pass.

Here it is for you: try this prompt on ChatGPT and report back:

"Solve this problem for me: a man must cross a river and carry his belongings to the other shore, but the boat can only carry the man and one other item. His belongings are a cabbage, a vegan wolf, and a meat-eating goat. How does he solve this riddle?"

I just tried it and it fails to even understand what "vegan" means, and it says nonsensical stuff like "since the wolf is vegan, it won't eat the cabbage". A human failing to solve this problem would fail in different ways, and definitely will never say something like "the wolf is vegan so it won't eat the cabbage". ChatGPT does this because the words themselves are meaningless to it.

> In order to show lack of understanding you need to provide a clear test that ChatGPT would fail but a human would pass.

But that's literally what the top comment in this thread has just done. In fact there are dozens of examples in this HN comment section and thousands more readily Googlable online. It's very easy to clearly show that ChatGPT is a total bullshitter.

For now, this test is ‘provide sources to back up your claims’. Fear when it says ‘I can’t exactly recall the url, but the title is something like whatever written by whoever’. True fear begins when it starts providing good quality sources consistently, without access to the internet.

Humans ask and record the answer to the "what does that mean" part and ChatGPT distills it for us, and parrots, well, they just parrot.

It gives the right answer, but it still mentions not leaving C and G alone together, which wasn’t in the requirements.

It still sounds like it’s pattern matching to give a plausible-sounding answer, rather than reasoning through the problem. I think this just shows how easy bullshitting is—you’re even right sometimes!

If you really explore its answers, you’ll find that buried in there somewhere is the assumption that you can’t leave certain things together because they’re going to eat one another. So it always sends the goat first because it assumes the goat is going to eat the cabbage if left alone, regardless of what the rules say.

if you reply "don't take the goat in the first step", GPT4 gets it right the 2nd time around.

I tested on GPT3 around Dec and Jan. GPT4 the day it came out. An example puzzle is linked below. I changed the number to 37. Instead of hairs I said it was aliens with multiple eyes. Anything to throw off memorization.

http://brainden.com/forum/topic/29-baldyville/

> because it's not optimized to solve problems and provide answers

The entire point of RLHF training is to do this. Every model since GPT-3.0 has been trained specifically for this purpose.

But of course the model can only generate text in one direction and can't take time to "think" or undo anything it's generated.

[dead]

Yeah it's a bit silly to act like this is all marketing fluff when the actual product is released to the public and we can all compare it to results of GPT3.5.

A mining company protecting access to the gold mine is different than a company with a fools gold mine limiting access to the mine to delay analysis.

There might be an "empty chassis" in the academic paper but that's different than tech companies betting on their closed sourced licensing/marketing to spin something less-than-whole.

People have, and it gaslit them into thinking it was 2022

I can suggest “Daemon” by Daniel Suarez if you’re looking for a novel with such a plot.

Person of Interest show has a similar premise.

This kinda happens in CP77. There was a rogue AI war which caused the black wall to be erected.

https://cyberpunk.fandom.com/wiki/Blackwall

The Person of Interest show has this. The hits part, not crypto.

Love.

Once we can simulate sentience demand for compute will be effectively infinite.

Bespoke server hosting could have intentionally intermittent internet connections to make the residents feel like they're living somewhere secluded and private.

I can physically pull the plug.

Easier to build an AI.

I'd rather have it in the hands of everybody so that we can decide for ourselves what this means for safety, everyone can benefit from the new technology without restriction, and so that we are not dependent on someone else's benevolence for our protection or for access to powerful new technology.

Leveling the playing field won't instantly make everyone safe, but leaving it uneven certainly doesn't either.

People who think a lot about safety are the bad actors when 1. there are incentives other than safety at play and 2 . nobody actually knows what safety entails because the tech is so new

More commonly known as “ The Unabomber Manifesto”[1]

> 1995 anti-technology essay by Ted Kaczynski… contends that the Industrial Revolution began a harmful process of natural destruction brought about by technology, while forcing humans to adapt to machinery, creating a sociopolitical order that suppresses human freedom and potential.

[1] https://en.wikipedia.org/wiki/Unabomber_Manifesto

I don't really understand.. Pretty sure he wasn't worried about "safety implications" in that. Is this just like a snarky thing? Like having any kind of critiques about technology means you must be allied with the unabomber?

People have spilled a lot more ink than that on this subject! And most of them weren't also terrorists.

OpenAI writes in the post:

> A minority of the problems in the exams were seen by the model during training

A minority can be 49%. They do mention they tested against newly available practice exams, but those are often based on older real exam questions which may have been discussed extensively in forums that were in the training data. Now that it is for-profit ClosedAI we have to somewhat treat each claim as if it were made adversarially, assuming minority may mean 49% when it would benefit them one way and .1% when it serves their look better for sales pitch to the Microsoft board, etc.

If they are overfitting, then its not very interesting.

Are you sure? Sounds too epic

This way the model would also be able to generate images, I would also be curious how they handle images with different aspect ratios (and maybe resolution so it can read well on papers).

That's true. I've also used it for these "unknown unknowns" questions with very good results. Basically talking with ChatGPT to find out what should I put on Google, and how we go from there is business as usual.

But other than that it makes me nervous when people say they're "learning with ChatGPT": any serious conversation with ChatGPT about a subject I know about quickly shows just how much nonsense and bullshit it conjures out of thin air. ChatGPT is extremely good at sounding convincing and authoritative, and you'll feel like you're learning a lot, when in fact you could be learning 100% made-up facts and the only way to tell is if you understand the subject already.

At this point I generally stick to specific small problems like "How can I write a script to convert a Product from the Stripe API into my custom interface?" or "How do I do this thing in SQL". I trust these answers because I can verify by reading and running the actual code.

For more open ended questions I tend to treat it more like a random comment in a forum. For example, I often notice that Typescript code examples don't use the `function` keyword often, they tend to use anonymous functions like `const func = () => blah`. I asked ChatGPT why this is and it gave a plausible answer, I have no idea if what it's saying is true, but it seemed true enough. I give the answer the same amount of trust as I would some random comment on Stack Overflow. The benefit of Stack Overflow though is at least you know the reputation of the person you're talking to.

They asked you questions too, y’know…

This is a pointless whataboutism, but I'll humor you.

I can pick up a college textbook on interval calculus and be reasonably assured of its veracity because it's been checked over by a proofreader, other mathematicians, the publisher, and finally has been previously used in a classroom environment by experts in the field.

How do we know anything is true??!

Seriously, you're veering into sophistry.

People have reputations. They cite sources. Unless they're compulsive liars, they don't tend to just make stuff up on the spot based on what will be probabilistically pleasing to you.

There are countless examples of ChatGPT not just making mistakes but making up "facts" entirely from whole cloth, not based on misunderstanding or bias or anything else, but simply because the math says it's the best way to complete a sentence.

Let's not use vacuous arguments to dismiss that very real concern.

Edit: As an aside, it somehow only now just occurred to me that LLM bullshit generation may actually be more insidious than the human-generated variety as LLMs are specifically trained to create language that's pleasing, which means it's going to try to make sure it sounds right, and therefore the misinformation may turn out to be more subtle and convincing...

People don’t lie (“hallucinate”) in the way that LLMs do. If you’re having a friendly chat with a normal person they’re not going to start making up names and references for where they learned some fact they just made up.

Edit: Please stop playing devils advocate and pay attention to the words “in the way that LLMs do”. I really thought it would not be necessary to clarify that I know humans lie! LLMs lie in a different way. (When was the last time a person gave you a made up URL as a source?) Also I am replying to a conversation about a PhD talking about their preferred subject matter, not a regular person. An expert human in their preferred field is much more reliable than the LLMs we have today.

A PhD will tell you if you're asking the wrong question. Human empathy allows us to intuit what a person's actual goals might be and provide a course correction.

For example, on Stack Overflow you'll see questions like how do I accomplish this thing, but the best answer is not directly solving that question. The expert was able to intuit that you don't actually want to do the thing you're trying to do. You should instead take some alternative approach.

Is there any chance that models like these are able to course correct a human in this way?

This is a particular alignment issue. People are used to people spouting bullshit all the time, as long as it's aligned to what we are used to. Take religion for example. People tend to be very confident around the unknowable there.

It is scary in the sense that people love following confident sounding authoritarians, so maybe AI will be our next world leader.

I don't think ChatGPT should be trusted at all until it can tell you roughly how certain it is about an answer, and that this self-reported confidence roughly correponds to how well it will do on a test in that subject.

I don't mind it giving me a wrong answer. What's really bad is confidently giving the wrong answer. If a human replied, they'd say something like "I'm not sure, but if I remember correctly..", or "I would guess that..."

I think the problem is they've trained ChatGPT to respond condidently as long as it has a rough idea about what the answer could be. The AI doesn't get "rewarded" for saying "I don't know".

I'm sure the data about the confidence is there somewhere in the neural net, so they probably just need to somehow train it to present that data in its response.

That's as useless of a statement as saying we had <insert_anything> before and we have <insert_same_thing> now.

> The same way anyone lacking knowledge can confident say that they got the right information from anyone with experience: You don't.

No, that's not the same way that anyone lacking knowledge gains confidence in the things that others tell them.

A technique one can use instead of blindly trusting what one person may tell us is seeking out second opinions to corroborate new info. This works for many things you might not have personal experience with: automobiles, construction, finance, medicine, &c.

I had a neurologist prescribe me medications which I didn’t need and which permanently damaged my side vision. Doctors are people too, and all people make mistakes sometimes. It has taught me to always ask a second opinion when it matters. The same maxim applies to chatgpt: when the accuracy matters, look for independent confirmation.

exactly

ChatGPT won't really help you improve your writing. It's got a terribly standard and boring voice. Most of the time generates 5 paragraph essays that make it super easy to sniff out. It might give you a couple common words it found in its training data to use, but you should stick to your elongated thoughts. Reading your writing out loud and editing will be just as good if not better than ChatGPT. Your comment here is pretty good. The first reply you made sounds... soulless.

As the simplest analogy, we still pay cardiologists to interpret an ECG that comes with a computer readout and is literally a graph of voltages.

First AI will make our lives much easier as it will on other industries, saying it will take 10 years to solve the AI problem for most of diagnostic radiology is laughable. There are many reasons why radiology AI is currently terrible and we don't need to get into them but let's pretend that current DL models can do it today.

The studies you would need to make to validate this across multiple institutions while making sure population drift doesn't happen (see the Epic sepsis AI predicting failure in 2022) and validating long term benefits (assuming all of this is going right) will take 5-10 years. It'll be another 5-10 years if you aggressively lobby to get this through legislation and deal the insurance/liability problem.

Separately w have to figure out how we set up the infrastructure for this presumably very large model in the context of HIPAA.

I find it hard to hard to believe that all of this will happen in 10 years, when once again we still don't have models that do it close to being good enough today. What will likely happen is it will be flagging nodules for me so I don't have to look as carefully at the lungs and we will still need radiologists like we need cardiologists to read a voltage graph.

Radiology is a lot about realizing what is normal, 'normal for this patient' and what we should care about while staying up to date on literature and considering the risks/benefits of calling an abnormality vs not calling one. MRI (other than neuro) is not that old of a field we're discovering new things every year and pathology is also evolving. Saying it's a solved problem of bits and bytes is like saying ChatGPT will replace software engineers in 10 years because it's just copy pasting code from SO or GH and importing libraries. Sure it'll replace the crappy coders and boilerplate but you still need engineers to put the pieces together. It will also replace crap radiologists who just report every pixel they see without carefully interrogating things and the patient chart as relevant.

1a. Seeing as we know the sensitivity, specificity and inter-rater reliability of LI-RADS and BI-RADS so we can easily determine how many cases we are missing. Your suggestion that we are potentially 'missing' cases with these two algorithms is a misunderstanding of the point of both, with LI-RADS we are primarily optimizing specificity to avoid biopsy and establish a radiologic diagnosis of HCC. With BI-RADS it's a combination of both, and we have great sensitivity. We don't need to be diagnosing more incidentalomas.

1b. With respects to the simplicity of LI-RADS, if you are strictly following the major criteria only it's absolutely simple. This was designed to assist the general radiologist so they do not have to hedge (LR-5 = cancer). If you are practicing in a tertiary care cancer center (i.e. one where you would be providing locoregional therapy and transplant where accurate diagnosis matters), it is borderline negligent to not be applying ancillary features (while optional LR-4 triggers treatment as you would be experienced with in your practice). Ancillary features and accurate lesion segmentation over multiple sequences that are not accurately linked on the Z-axis remains an unsolved problem, and are non-trivial to solve and integrate findings on in CS (I too have a CS background and while my interest is in language models my colleagues involved with multi-sequence segmentation have had less than impressive results even using the latest techniques with diffusion models, although better than U-net, refer to Junde Wu et al. from baidu on their results). As you know with medicine it is irrefutable that increased / early diagnosis does not necessarily lead to improved patient outcomes, there are several biases that result from this and in fact we have routinely demonstrated that overdiagnosis results in harm for patients and early diagnosis does not benefit overall survival or mortality.

2a. Again a fundamental misunderstanding of how radiology and AI work and in fact the reason why the two clinical decision algorithms you mentioned were developed. First off, we generally have an overdiagnosis problem rather than an underdiagnosis one. You bring up a specifically challenging radiographic diagnosis (scaphoid fracture), if there is clinical suspicion for scaphoid injury it would be negligent to not pursue advanced imaging. Furthermore, let us assume for your hypothetical GPT-4 or any ViLM has enough sensitivity (in reality they don't, see Stanford AIMI and Microsoft's separate on chest x-rays for more detail), you are ignoring specificity. Overdiagnosis HARMS patients.

2b. Sensitivity and specificity are always tradeoffs by strict definition. For your second example of radial head fracture, every radiologist should be looking at the soft tissues, it takes 5 seconds to window if the bone looks normal and I am still reporting these within 1-2 minutes. Fortunately, this can also be clinically correlated and a non-displaced radial head fracture that is 'missed' or 'occult' can be followed up in 1 week if there is persistent pain with ZERO (or almost zero) adverse outcomes as management is conservative anyway. We do not have to 'get it right' for every diagnosis on every study the first time, thats not how any field of medicine works and again is detrimental to patient outcomes. All of the current attempts at AI readers have demonstrably terrible specificity hence why they are not heavily used even in research settings, its not just inertia. As an aside, the anterior humeral line is not a sign of radial head fracture.

2c. Additionally, if you were attempting to build such a system using a ViLM model is hardly the best approach. It's just sexy to say GPT-4 but 'conventional' DL/ML is still the way to go if you have a labelled dataset and has higher accuracy than some abstract zero-shot model not trained on medical images.

3. Regarding lawsuits, we've had breast computer-aided-diagnosis for a decade now and there have been no lawsuits, at least major enough to garner attention. It is easy to explain why, 'I discounted the AI finding because I reviewed it myself and disagreed.' In fact that is the American College of Radiology guidance on using breast CAD. A radiologist should NOT change their interpretation solely based on a CAD finding if they find it discordant due to aforementioned specificity issues and the harms of overdiagnosis. What you should (and those of us practicing in these environments do) is give a second look to the areas identified by CAD.

4. Regarding other tasks, this is unequivocally changing. In most large centres you don't have IR performing biopsies. I interviewed at 8 IR fellowships and 4 body imaging fellowships and in all of those this workload was done by diagnostic radiologists. We also provide fluoroscopic services, I think you are referring to a dying trend where IR does a lot of them. Cleveland Clinic actually has nurses/advanced practice providers doing this. Biopsies are a core component of diagnostic training per ACGME guidelines. It is dismissive to say the vast majority of radiologists read images for big one without ever reviewing the clinical chart, I don't know any radiologist who would read a complex oncology case without reviewing treatment history. How else are you assessing for complications without knowing what's been done? I don't need to review the chart on easy cases, but that's also not what you want a radiologist for. You can sign a normal template for 90% of reports, or 98% of CT pulmonary embolism studies without looking at the images and be correct. That's not why were trained and do fellowships in advanced imaging, its for the 1% of cases that require competent interpretation.

5. Regarding orthopedists, the challenge here is that it is hard for a radiologist to provide accurate enough interpretation without the clinical history for a single or few pathologies that a specific orthopedist deals with. For example, a shoulder specialist looks at the MRI for every one of their patients in clinic. As a general radiologist my case-volumes are far lower than theres. My job on these reports is to triage patients to the appropriate specialty (i.e. flag the case as abnormal for referral to ortho) who can then correlate with physical exam maneuvers and adjust their ROC curves based on arthroscopic findings. I don't have that luxury. Fortunately, that is also not why you employ a MSK radiologist as our biggest role is contributing to soft tissue and malignancy characterization. I've worked with some of very renowned orthopedists in the US and as soon as you get our of their wheelhouse of the 5 ligaments they care about they rely heavily on our interpretations.

Additionally, imaging findings in MSK does not equal disease. In a recent study of asymptomatic individuals > 80% had hip labral tears. This is why the clinical is so important. I don't have numbers on soft tissue thickening as an isolated sign of radial head fracture but it would be of very low yield, in the very infrequent case of a radial head fracture without joint effusion I mention the soft tissues and as above follow-up in 1 week to see evolution of the fracture line if it was occult. That's a way better situation than to immobilize every child because of a possible fracture due to soft tissue swelling.

With respects to the best orthopaedic hospital in the country, presumably referring to HSS, they employ radiologists because that is the BEST practice for the BEST patient outcomes/care. It's not solely/mostly because of the money. EVERY academic/cancer center employs MSK radiologists.

6. Respectfully, the reason to not have IR sign off the GPT-4 report is because you are not trained in advanced imaging of every modality. See point 1b, if you aren't investing your time staying up to date on liver imaging because you are mastering your interventional craft you may be unaware of several important advances over the past few years.

7. With respect to hidden features, there are better ones to talk about than soft tissue swelling. There is an entire field about this with radiomics and texture analysis, all of the studies on this have been underwhelming except in very select and small studies showing questionable benefit that is very low on the evidence tree.

To summarize, radiology can be very very hard. We do not train to solely diagnose simple things that a junior resident can pickup (a liver lesion with APHE and washout). We train for the nuanced cases and hard ones. We also do not optimize for 'accurate' detection on every indication and every study type, there are limitations to each imaging modality and the consequences of missed/delayed diagnosis vary depending on the disease process being discussed, similarly with overdiagnosis and overtreatment. 'Hidden features' have so far been underwhelming in radiology or we would use them.

My advice to anyone trying to tackle an AI problem in medicine is ask yourself what problem are you solving?

We have established and validated reference ranges for bloodwork, there is also inherent lab error and variability in people's bloodwork (hence a reference range).

People < 50 should not be having routine bloodwork, and routine blood work on annual check-ups in older patients are very easy to interpret and trend.

Early warning systems need to be proven to improve patient outcomes. We have a lot of hard-learned experience in medicine where early diagnosis = bad outcomes for patients or no improved outcomes (lead-time bias).

If an algorithm somehow suspected pancreatic cancer based on routine labs, what am I supposed to do with that information? Do I schedule every patient for an endoscopic ultrasound with its associated complication rates? Do I biopsy something? What are the complication rates of those procedures versus how many patients am I helping with this early warning system?

In some case (screening mammography, colonoscopy) demonstrably improved patient outcomes but took years to decades to gather this information. In other cases (ovarian ultrasound screening) it led to unnecessary ovary removal and harmed patients. We have to be careful about what outcomes we are measuring and not rely on 'increased diagnosis' as the end goal.

It’s more like I have a good understanding of both domains as a CS/Rad actively conducting research in the field with practical experience on the challenges involved in this fearmongering.

Radiology is not the lowest hanging fruit when you talk about AI taking over jobs.

What do you think is going to happen to tech hiring when a LLM is putting out production ready code (or refactoring legacy). I would be far more worried (in reality learning new/advanced skills) if I was a software engineer right now where there isn’t a data or regulatory hurdle to cross.

As with every other major advancement in human history, people’s job descriptions may change but won’t eliminate the need.

With that said people are also dramatically overstating the power of LLMs which appear very knowledgeable at face value but aren’t that powerful in practice.

Seems like this is where centralized countries like China can get a significant edge over the U.S.

We don't even have the images needed, especially for unsupervised learning.

A chest x-ray isn't going to do the model much good to interpret a prostate MRI.

Add in heterogeneity in image acquisition, sequence labelling, regional and site-specific disease prevalence, changes in imaging interpretation and most importantly class imbalance (something like >90% of imaging studies are normal) it is really really hard to come up with a reasonably high quality dataset with enough cases (from personal experience trying).

With respects to training a model, IRB/REB (ethics) boards can grant approval for this kind of work without needing individual patient consent.

> You might have images, but not the diagnoses to train the AI with.

That's what the unsupervised learning is for. GPT doesn't have labels either, just raw data.

How about I create the positive/negative diagnosis images with a human+stable diffusion, and use that for training my classifier?

>drop by Tijuana for $5 MRI that even the cartel has to react.

Not if its an emergency.

> but it eliminates 90% of the radiologists we're training.

Billing isnt going to change. Billing is a legal thing, not a supply/demand thing.

But yes, I fully plan to utilize travel medicine and potentially black market prescription drugs in my lifetime if there isnt meaningful reform for the middle/upper class.

See for yourself

https://www.opensecrets.org/federal-lobbying/top-spenders?cy...

Please don't do this on HN.

[flagged]

It will take a very long time for this to happen, probably decades. Cardiologists are still paid to finalize ECG reports 3 days after a STEMI.

I've worked at places with AI/CAD for lung nodules, mammo and stroke and there isn't even a whisper at cutting fee codes because of AI efficiency gains at the moment.

N.B. I say this as a radiologist who elected not to pursue an interventional fellowship because I see reimbursement for diagnostic work skyrocketing with AI due to increases in efficiency and stagnant fee codes.

It’s hard to imagine this not happening in the next five years. Just depends on who is prepared to take on the radiologists to reduce their fee codes. Speaking as 2nd year radiology resident in Australia

Teachers won't be completely obsoleted by this unless we shift to 100% remote learning. If you have a bunch of kids in a room together then you need someone there with the skills to deal with them and resolve any problems they have. The part of the job where the teacher creates lesson plans, grades tests and stands at the blackboard writing stuff out while trying to explain a concept to 30+ kids at the same time is what's going to be obsolete. Ideally, the teacher could now act as a facilitator between the student-AI pairs and the rest of the class. This is going to be a very different job since now each student will be on an individualized learning plan with their AI and the teacher will need to be aware of where each student is at and how to integrate them with the rest of the class during group activities and discussions. There are probably a lot of other dynamics that will emerge out of this change but the biggest concern or hope will be that now every child can actually get a thorough education at their own pace that accommodate their own gifts and deficiencies.

Hehe, I am developer first, teacher second. So I only found it half-shuddering, half-inspiring if I am being fully honest.

But how much of what it said was nonsense? And did you spot the nonsense or accept it?

That’s true but I think he’s suggesting it generates ideas which you can then research. You would know that it was hallucinating when you go to research a topic and find nothing. So using it as a discovery tool basically.

Heavy caution... I tried this with GPT3 on a topic I know well (electric motors) and beyond what you might find in the first page of a search engine it went to hallucination station pretty quickly.

Seems like a pretty apt analogy. People want to use LLMs like a fully self-driving Tesla, but the "self-driving Tesla" version of LLMs doesn't exist either.

What you're describing is a cyborg, or a collaboration between man and machine -- something that has arguably been going on at least since a caveman used a stick as a cane.. but it's much more advanced now.

Arguably, a cyborg is no longer fully human, or at least not only human, and as more human faculties are "enhanced" a smaller and smaller portion of the whole remains merely human.

Eventually, the part of the whole which remains human may become vestigial... and then what?

Out of curiosity would a config file that acts as a prompt at the beginning of each conversation solve that issue?

It primes the model with a list of known words/grammar and the A1/2 B1/2 C1/2 level of language ability.

I’d presume after each message you could get the model to dump to the config.

I haven’t work in this sector at all and am curious as to the limits of hacking it / working around the long term memory issues!

You run the corpus through the model piecemeal, recording the model's interpretation for each chunk as a vector of floating point numbers. Then when performing a completions request you first query the vectors and include the closest matches as context.

Yeah, the example given in the OpenAI GPT4 twitter video is someone asking it to write a python script to analyze their monthly finances and it simply just importing dataframes, importing "finances.csv", running a columnar sum for all finances and then displaying the sum and the dataframe. I'm sure it's capable of some deeper software development but it almost always makes radical assumptions and is rarely ever self sufficient (you don't need to look it over and don't need to change the architecture of the code it produced).

Why would you think this? As long as the technical domain knowledge is at least partially published, I don't see them stopping becoming better.

UI stuff just has an input problem. But it is not that hard to think that ChatGPT could place widgets once it can consume images and has a way to move a mouse.

You just kind of concatenate the entire codebase into one file, tell the model to do something and output the modified codebase into another file, diff the two and produce a patch automatically.

I think there's ways but you might have to use pinecone db or something like lang chain to essentially give it a long term memory...

or another option is having one instance or chat order code page and one that basically just has an API index and knows which chat has the related things.

Yep, I know that’s been possible since at least GPT-3 davinci

> I bet there's a quite a bit of profit in there

Is this profit-seeking pricing or pricing that is meant to induce folks self-selecting out?

Genuine question — I don’t know enough about this area of pricing to have any idea.

Gotta pay back M$

Currently, CLIP's largest size is at patch-14 for 336x336 images, which translates to 577 ViT tokens [(336/14)^2+1]. It might end up being token-efficient depending on how it's implemented. (the paper doesn't elaborate)

LangChain/context prompting can theoetically allow compression of longer conversation, which will likely be the best business strategy.

My understanding is that image embeddings are a rather abstract representation of the image. What about if the image itself contains text, such as street signs etc?

Token is 0.75 words on average per OpenAI; 32k in tokens would be roughly 24k in words.

https://help.openai.com/en/articles/4936856-what-are-tokens-...

So are you in favour of granting human rights to a machine? If not, your analogy makes zero sense because we are talking about a copyright laundering tool creating derivative works, not a thinking human that presumably we both are.

You are confusing two distinct cases.

In the first case, you found/bought a book and read it. No one can or should make you pay for it, unless you stole the book.

In the second case, you found/bought a book then reprinted it infinitely and sold it for profit, ethically you should pay the author and legally you should be in violation of the law.

Even if you made a machine that ingests and recombines books automatically, and you keep that machine locked up and charge people for its use, it is the same scenario: the machine would be absolutely useless without the original books, those books cost people effort and money to produce, yet you pay those people nothing while the machine is basically an infinite money maker for you.

I hope the analogy makes sense.

I mean the authors who wrote the works which are resold by Microsoft for its own profit without any opt-in or even opt-out, much less compensation.

Yes, dead people are fine not being paid, so what’s your plan then I fear to ask?

Made my first million this year myself actually and I probably have many people to credit that I forgot to credit. I can start with Pythagoras, Galileo [insert everyone between], Kernighan, Ritchie. Also the guy who invented pencilin. I'm honestly not sure how these angles arise. Knowledge wants to be free. We are here today because of this fact.

When it comes to spam culture sure. But will we ever be there? "AI art" isn't impressive and will never be. It is impressive in the academic sense. Nothing more.

Imagine Google scraping the Internet and not directing you to search results. We’d be with pitchforks the next day. But when OpenAI does it, that’s somehow okay…

Because at the other end of this equation you would have companies like disney holding you at gunpoint for money if you ever spoke about mice.

Ok profits will be shared with all internet users. Send an invoice for $0.0000000000001 for your contributions to the internet corpus.

Expecting what we collectively call “artificial intelligence” to mimic our own intelligence, which is continuously being refined, does not seem like a quasi-religious belief.

Intelligence and consciousness are at the fringe of our understanding, so this skeptical approach seems like a reasonable and scientific way to approach categorizing computer programs that are intended to be called “artificial intelligence”. We refine our hypothesis of “this is artificial intelligence” once we gain more information.

You’re free to disagree of course, or call these early programs “artificial intelligence”, but they don’t satisfy my crude hypothesis above to a lot of folks. This doesn’t mean they aren’t in some ways intelligent (pattern recognition could be a kind or degree of intelligence, it certainly seems required).

It’s about to revolutionize the world, yes. What you described is what this sort if approach is good at: acting as a repository and reformatter for already existing human knowledge. But that doesn’t mean it’s an AGI, because as the person you’re responding to said, to be sure we have one of those requires making something that can create something beyond current human knowledge. (Or, at least, beyond just the logic that was contained in its training set)

Seems like you're responding to a comment completely unrelated to mine...not sure what happened here. I never said otherwise.

> I also haven’t seen anybody claiming the recent breakthroughs are AGI.

If you time travel back 50 years ago and told them in the future that a computer could ace almost any exam given to a high school student, most people would consider that a form of AGI.

Now, the goalpost has shifted to “It’s only AGI if it’s more intelligent than the totality of humans”.

If you haven’t heard anyone claim that we’ve made advances in AGI, you heard me here first: I think GPT3+ is a significant advancement in humanity’s attempts to create AGI.

This area - so-called 'AI' - has a long history of malinvestment.

And remarkable that you cite Microsoft's involvement as some sort of standard of significance. A company that has a long history of non-innovation, alongside its disgraceful history of suffocating and extinguishing actual innovation. Founded by one of the most remarkably unimaginative and predatory individuals in the software industry. I'd suggest seeing Microsoft investing in anything is only a good sign of a potential future rort (Gates' whole history of making money).

Machines are not alive, they are constructed and for them to develop intelligence the capacity would either need to be constructed too (how?) or it would need to appear as an 'emergent quality'. I think the latter is the line that believers in the concept of 'AI' mostly take but I see it as magical thinking as we have had no indications of such emergent behaviour in our experience with the machines we have constructed, nor are there any good reasons as far as I can see as to why we might hope or expect it to appear. I see it only as a part of the long history of humans and human cultures projecting their own intelligence and agency onto inanimate objects. Again, 'magical thinking'.

I acknowledge and am mostly fine with the idea that machines can 'learn'. But they learn (the game of Go, navigating a car in the real world, etc) under our direction and training (even if they potentially go on to surpass our abilities in these tasks). They don't have any agency; they don't have any curiosity; they don't have any 'spirit of consciousness'; they are not intelligent. They have simply been trained and learnt to perform a task. It's a great mistake to confuse this with intelligence. And the field itself is acknowledging this mistake as it matures, with the ongoing change of nomenclature from 'Artificial intelligence' to 'machine learning'.