"There have been no major breakthroughs in the academic discipline of artificial intelligence for a couple of decades." umm... ok...

You can tell even as a non-expert that the author has no clue about the academic landscape of AI.

Source: GPT-4

It's like no one realizes AI is just another tool. It is only as powerful as the wielder. Guns can cause mash shootings, but they can also bring freedom. Who's side will you help?

> It's like no one realizes AI is just another tool. It is only as powerful as the wielder.

For now! But even if it never wields itself, what it is is a swiss army knife with a bunch of uses, including "writer", "painter", "dystopia manufacturer", and "weapon of mass destruction". Writer and artist associations are busy debating how best to use it ethically, while the military industrial complex pants lustfully in the corner.

What a shallow and ignorant article.

* * *

As a liberal-left lifestyle tabloid, they had to bash something out on this hot topic.

Yeah, well, tell that to marketing departments.

(Until they get replaced by AI, of course).

Today’s AI threatens the livelihoods of talented rhetoricians, so expect many more articles like this one.

This is more of an issue with marketing requirements than AI being stupid. ChatGPT might be more aptly called as compressed knowledge storage than AI but then who will pour the billions in it?

That being said, knowledge storage was a major issue to create a true AGI and it seems deep learning and related ML methods have actually solved this problem.

Next task is implementing the actual reasoning abilities and ability to take unguided actions.

This article left me with a profound sadness.

Not only was it paint-by-numbers, it managed to downplay its own alarm, undercutting the point of reading the article in the first place: for if AI is so 'dreary', then why are we dancing like water drops on a hot skillet?

The overall point of this piece seemed to be an attempt by a 'harrumph' (a term a friend of mine coined for out-of-touch English department creatures) to steer the conversation about AI back towards something the author actually knows something about.

It reminded me of WWI generals, sending wave after wave of calvary towards death, unable to ken that horses now had nothing to do with warfare.

Harrumphs gotta harrumph. In doing so, they will miss out on the now-vanishing window to actually say something meaningful and novel about this terrifying new world into which we have awoken.

It's a shame, because much vital conversation must occur. Immediately. Instanter. Now.

The material analysis (that AI represents wealth concentration) is surely correct, but also well beside the point. We need to move the conversation along to UBI, social housing, and counselling for the hundreds of millions of white-collar workers who no longer have marketable skills.

The author is concerned that LLMs confidently hallucinate false information, and also shares his confident hallucinations about how LLMs work.

This was flimsy all the way through, but loses all credibility when this part rolls around:

> It risks poisoning the well of collective thought, and of our ability to think at all. ... To place all of our trust in the dreams of badly programmed machines would be to abandon such critical thinking altogether.

The author seemingly couldn't resist leaning all the way into hyperbole and made the ridiculous claim that human beings will no longer be able to think once AI really takes off. The unnecessary name calling in saying that LLMs are "badly programmed" is the cherry on top.

A better criticism of the recent AI hype is this article by Noam Chomsky, Ian Roberts and Jeffrey Watumull. [0] which critiques the explainability and reasoning of LLMs with its opaque regurgitation of pre-trained text that it has been trained on.

Then compare it with the original arguments made by a recent OpenAI employee [1] and you will see that even with the "Important Updates" in his blog attempting to revive his weaker rebuttal, the stronger evergreen claims in [0] still stand.

[0] https://archive.is/AgWkn

[1] https://scottaaronson.blog/?p=7094