The Great A.I. Awakening | Hacker News Reader

The Great A.I. Awakening

nytimes.com

63 pointskanamekun12 comments

From what I understand (someone correct me if I am wrong), the syntactical approach in AI has its roots in classical analytic philosophy, as in Russell and Whitehead's Principia Mathematica. The idea isto replace ordinary language with a perfect language composed of clear, simple concepts that would describe all truth and reality in a completely precise manner.

Russell's student Wittgenstein originally supported this approach, but later demolished it in his book Philosophical Investigations. He showed that any useful language would be imperfect.

It seems to me that the fact the machine learning approach is succeeding where syntactical AI failed helps demonstrate Wittgenstein was right.

I am of the opinion that in the end strong AI will be a mishmash of many different components with a unified ego which will be simple enough to write once all the other components are in place. And many of the patterns that it will have in its subconscious will mirror those that we have in our which are acquired through various sorts of learning. In the end it won't really understand itself and why it does the things that it does, just like we don't.

"intelligent AI"... so IAI?

sorry, couldn't resist ;-)

changed to strong ai.

it's almost like the problem with high end cars. First there were Supercars, then there was an inflation of that name being used, then the term "Hypercars" was invented.

Isn't "AGI" (artificial general intelligence) typically what is used in this context? Or do you mean something else? I like that term a bit more because at least it becomes somewhat clear what's the difference to the machine learning algorithms we already have.

https://www.ocf.berkeley.edu/~arihuang/academic/research/str...

"Strong AI's goal is to develop artificial intelligence to the point where the machine's intellectual capability is functionally equal to a human's."

I'd be careful saying that one is succeeding and another failing, at least to the point of drawing very large conclusions about the nature of intelligence. Under our current level of technology and our current mathematical models, neural networks and other squishier techniques are outperforming rule-based agents in certain fields. Basic neural networks were proposed long ago, though, and Minsky and Papert made quite a convincing argument in the 70s that the success of syntactical AI over neural networks at many important tasks was a clear sign that rule-based logic was correct! We're still so far from general intelligence that it's hard to draw conclusions.

That's a good point. It makes me think that strong AI would have to have both neural network and rule-based components. However, I think that the meta-function that decides when to use one or the other would have to be a neural network. I think that's the way it works in human beings.

I think that is so because the realities that intelligence is supposed to deal with are such that they can't be fully described by any single or multiple systems of precise concepts. That was Wittgenstein's point, and also many other philosophers such as Husserl, Heidegger, Merleau-Ponty, Ricoeur, Whitehead, Putnam, Strawson, Rorty, and all the Pragmatists.

I agree that a lack of precision with respect to correspondence with the real world is inevitable. I don't think that at all mandates a neural network (and artificial neural networks are only superficially similar to human neurons, and their performance often improves by making them even less analogous.) But the idea that the world can't be fully and accurately described by a simple system of precise concepts, as much as it would pain Leibniz, doesn't necessitate that an intelligent agent couldn't use strictly rule-based logic.

I still think we have the bar set too high for human intelligence

Christ what a crock of shit. Talk about fake news.

Every single day something is being called AI. Every time the output is relatively shit and doesn't improve in any noticeable way even over long periods of time.

Remember when Siri was AI? Hahaha. Siri is a borderline outsourced retard who barely understands English.

Please stop fantasising everything is AI. It's not. It's marketing.

We've already asked you to please comment civilly and substantively; we ban accounts that continue like this.

https://news.ycombinator.com/newsguidelines.html

Reply on news.ycombinator.com