[somewhereoutth]

In case anyone is wondering, we have made zero progress on anything even remotely resembling Artificial Intelligence. Zero.

Unfortunately of course, the people who might have some of the skills needed to actually build such a thing (at the bricks and mortar level anyway), are nearly those people whose understanding of what intelligence actually is may be less than ideal. As a hint, it has nothing to do with passing tests or other such mundanity.

A more interesting approach would be to consider language - if cooperating entities can be constructed that (eventually yet spontaneously) created ways to communicate between each other, then maybe some progress has been made.

Further, if we appreciate that any idea, discovery, anything, can be communicated to even the most recently discovered humans in their own language (though we may need to build up the various concepts from basic terms), and that no such feat is possible with the other animals, then we might wonder if another intelligence (artificial or otherwise) might be able to encode concepts that are unreachable in our (any of our) language and thus thoughts - or, alternatively, that our (any of our) language is conceptually complete in some fundamental sense, and so there simply cannot be such 'higher' intelligence (artificial of otherwise).

[wyattpeak]

If you showed somebody from 1921 a page of text produced by GPT-3, told them that it was written by a machine, and then told them that we'd made no progress towards artificial intelligence, they'd laugh in your face.

You can take from that what you will, but I suspect it will always seem as though we've made no progress, because anything we learn to emulate we necessarily understand well enough that it will no longer seem magical. I wouldn't put it past us to start thinking of humans as automata before we declare that machines can think.

[Ginden]

> If you showed somebody from 1921 a page of text produced by GPT-3, told them that it was written by a machine, and then told them that we'd made no progress towards artificial intelligence, they'd laugh in your face.

You can actually do it. 100 year old people usually don't follow news on artificial intelligence, so they will act genuine.

[jjoonathan]

Unless the people running GPT-2 bots all over the internet suddenly gave up when GPT-3 came out, it's been passing Turing tests on audiences much younger than 100 years.

[irrational]

My grandmother is 99 and definitely doesn’t follow news on artificial intelligence. Maybe I should test this on her the next time I’m at her house.

[Cyphase]

This is the AI effect: https://en.wikipedia.org/wiki/AI_effect

[hyperpallium2]

The humans move goalposts because intelligence is political.

[robbedpeter]

It's clear that gpt models are more competent with knowledge work than a significant percentage of humans. This is implicitly threatening, to the extent that it seems people will refuse to even consider the possibility. Dall-e is a better artist than 99% of humans.

We thought we'd have time for the mental tasks as ai encroached on the menial, but it seems to be the reverse.

By every measure Turing himself considered, the Turing test has been passed. It's only the post-gpt-2 peanut gallery that have insisted on moving the goalposts straight into mysticism and magical thinking.

Machines will be better at everything humans can do, and accomplish things we cannot.

We are living in interesting times, different from anything that's come before - we exist in relation to systems that are learning to think like us.

[katabasis]

If we continue to move the goalposts of what defines intelligence into mystical/ineffable territory, we may find that humanity no longer qualifies as “intelligent” either.

[hyperpallium2]

The Turing test employs interview, not artistic creation, to distinguish the human.

The mechanisation of knowledge work has been ongoing (at least) since human accounting with beans - before writing; before number; maybe, even before language.

The humans' real fear will rise when they meet with superior argument.

[TaylorAlexander]

> Dall-e is a better artist than 99% of humans.

Really depends on whether “art” means “making drawings of things” or “making people feel something”. It’s also a very narrow domain. Dall-e can’t sculpt clay (for example), even if you attached a robot arm, without essentially replacing a bunch of the training system logic. Out of the box Dall-e has no provision to manipulate anything to produce art.

[mensetmanusman]

Where can I try Dall-e? If it’s not available to test, how can we know?

[djsbs]

“The machine stops” [1]written in 1909 has AI composing things (I forget if it was poetry or music)

Orwell’s 1984, written in the mid forties, has pop songs written by machine.

In both cases the AI composed works are described in the same way Id describe modern AI composing things - dreadful.

The concept of AI is quite old. Even Medieval Europe you had philosophers making quite penetrating insights on mechanical creativity. But, lacking a computer, there was no point continuing their train of thought

[1] amazing, far seeing, book. Very short, maybe a two hour read.

[wyattpeak]

I'm not saying the idea would be new to them, the idea of thinking machines had been around for a lot longer than that. I'm saying that the idea that modern text-generation is "zero progress on anything even remotely resembling Artificial Intelligence" would be absurd to them.

Jules Verne wrote about a trip to the moon. It doesn't follow that he would regard the NASA missions as old-hat.

[jimmaswell]

Some AI generated music certainly passes as normal music.

[mensetmanusman]

Alternatively, they could have said, ‘oh great! Computers continued to improve and you were finally able to implement our algorithms on enough data!’

[wyattpeak]

To whatever extent that people in the 1920s can be said to have had algorithms for machine learning, they certainly didn't bear any relation to modern algorithms.

Even the idea of requiring enough data to build a good system is fairly new. As late as the 1980s, expert systems were the dominant approach to artificial intelligence, and they didn't require information corpi at all but instead involved experts programming in all of the rules they could think of for a system.

[mensetmanusman]

Do we know who published the first conceptual framework for the algorithms behind AlphaGo etc? It seems like they would get a Nobel prize at some point…

[Rury]

I'd posit intelligence isn't what people make it out to be, and that we already have AI. People just aren't impressed by it when they learn the magic behind it, and hence disagree on that we have it.

I mean, people seem to hold human intelligence as something extraordinary, despite having no idea what precisely makes us intelligent. Isn't that kind of pulling the cart before the horse? For all we know, humans might just be biomechanical robots operating on the "stimuli" inputted to us, behaving in completely predictable ways, no different than how computers operate on the "data" inputted to them.

[nine_k]

Wolves definitely don't have language.

Still, they possess an undeniable degree of intelligence. They also have cultures, that is, forms of knowledge passed between generations by teaching, not genetically, and differing between packs.

I suspect that a robot as intelligent as a dog, but with an easier interface, would be a great help to humans.

OTOH, what currently is called "AI" is mostly deep learning, a very important part of cognition and perception. Without modern results in computer perception and low-level cognition and control, a "more general" AI would be blind, deaf, and paralyzed in the real world.

I suspect that the older approaches based on more supervised ways to construct cognitive functions have not born all the fruit they could, and may eventually help create an AI with better higher-level reasoning. They are just not in vogue now, so the best researchers and fattest grants are in deep learning and around. Also, the hardware may not be there yet.

(A similar thing happened to neural networks. The first, one-layer, neural network was the perceptron created in 1958 [1] The approach, while valid and constantly developed, did not see a real uptake until early 2010s, when incomparably better hardware finally became available.)

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

[akamoonknight]

One thing sort of related to language that I see (as an entire outside observer to the field) as being required is some sort of shared communication 'channel'. Biological life works with atoms to form proteins that seem to do much of the communication that eventually guides higher level functions. Computers and computational processes can work on bytes and bits and package those into messages or results but on their own I'm not sure what it means for one process to consume another processes bytes/bits/messages, whereas proteins have physical results that lead to responses. Not that biological life should necessarily be the goal, but its definitely been good at guiding us in a lot of different ways. It seems like some sort of shared medium (that can be dynamically combined/recombined as needed) is required to communicate between disparate processes is required to dynamically change/improve systems and I just don't have any idea what that really looks like.

[mgraczyk]

More to the point, we've also made no progress on supersizing existing unintelligent machines. In fact, machines have become dramatically smaller over the last several years.

If you look at the people who have the skills to make such machines larger, those who built bigger and better vacuum tubes and larger cathode displays with more oomph, they all appear to have disappeared, replaced by the misguided miniaturizers.

Your last point is already addressed in the paper, argument #3.

[highspeedbus]

Before we can commonly use a new noun, we need to fully fit its meaning in our limited working memory. So I believe there is a natural upper bound in human intelligence, for things that are beyond our brain power to get the full picture.

That must be why we haven't solved P = NP yet. This would take a person with twice the L1 cache to accomplish.

[jimmaswell]

> zero progress on anything even remotely resembling Artificial Intelligence

I know it's a bit hyperbolic but Skynet comes to mind every day I use Copilot. It's just amazing the kind of things it can suggest/adapt to. We're definitely on some path of progress.

[hiddencost]

Checkv out the reinforcement learning on hanabi task. A cool approach to cooperation.