[gjm11]

I'm not a fan of the glib "everyone knows AI systems don't really think, they are just stochastic parrots, all they do is regurgitate ideas they've stolen" schtick, but this article is the reverse of that only worse.

Today's AI systems are pretty impressive but they are absolutely not, not even slightly, the equivalent of Einstein + Hawking + Tao. The reason they get used a lot for tasks along the lines of "rewrite this so it sounds smarter" is that that's what they're best at.

If we did as the author seems to want and tried to use these systems to solve the kinds of problems we need Einsteins, Hawkings and Taos for, then we would be in for one miserable disappointment after another. Maybe some day -- maybe some day very soon -- they'll be able to do that, but not now.

An article proclaiming that today's AI systems are at the level of Einstein mostly suggests to me that the author's own intellectual level isn't much higher than that of the AI systems he falsely equates with them. That seems unlikely, but I don't have a better explanation for how someone could write something so very far from the truth.

[jw1224]

> If we […] tried to use these systems to solve the kinds of problems we need Einsteins, Hawkings and Taos for, then we would be in for one miserable disappointment after another

We can literally watch Terence Tao himself vibe coding formal proofs using Claude and o4. He doesn’t seem too disappointed.

https://youtu.be/zZr54G7ec7A?si=GpRZK5W1LDvWyBBw

[wizzwizz4]

He's the only person I know of who can actually get good results out of these systems (though I know several people who claim they can). What he's doing is fundamentally not the same thing as what most "vibe coders" are doing: take the autocomplete away, and he's still a talented mathematician.

[gjm11]

Sure, but what he's doing is very much not using Claude or o4 to do things we need Terence Tao for.

I'm not saying today's AI systems aren't useful for anything. I'm not saying they aren't impressive. I'm just saying they're nowhere close to the "Einstein, Hawking and Tao in your house" hyperbole in the OP. I would be very, very surprised if Terence Tao disagreed with me about that.

[mufthun]

You can literally watch Terrence Tao stream himself formalizing existing proofs that he already formalized before.

[personjerry]

> The reason they get used a lot for tasks along the lines of "rewrite this so it sounds smarter" is that that's what they're best at.

I disagree. The reason is that that's what aligns best with what most people are looking for help on.

There is a disconnect between reality and the AI product consumer envisioned here. There is no magical enlightened user who's going to unleash their inner potential.

How much physics or math does the average person know? How much do you think they even WANT to know? The answer is surprisingly little.

On a day-to-day basis the layman writes emails and other mundane tasks, and wants to do them faster and easier.

Having a squad of geniuses in my pocket doesn't pay my bills.

[csallen]

This is the right answer.

Usage of products is determined by what people are driven to do. People are driven by their desires and their problems. And most of these are fairly simple and mundane… eating, paying the bills, feeling healthy, connecting with others socially, etc.

"Expending copious amounts of mental energy on difficult work to create scientific breakthroughs that may-or-may-not allow engineers to build things that contribute to the betterment of the human race" is not how most people want to spend their time, even if there are tools available to help them do that.

[aprilthird2021]

Sorry, you can disagree, but LLMs are generative, meant to generate pleasing-to-humans text, specifically, and that is what they do best...

They are not going to come up with a theory of relativity

[aprilthird2021]

I agree 100%. Additionally, this article ignores the existence of Google. Even the high level questions the person asked Einstein before he devolved to asking for email editing help were things you could have just googled.

The greatness of great minds was how they thought about problems and how they changed how we thought about things. An AI cannot do that. It's designed to tell you what people combined have already agreed upon. It's not designed to break the frontier of our knowledge

[justonceokay]

There’s a little nagging thought in my head when I hear that some people are helped immensely by AI and others are not. It’s that there is a threshold for intelligence that the AI either impresses you or it does not. I’m sure this threshold will continue to rise

[jxjnskkzxxhx]

In my experience, people who understand llms better are more impressed. Not impressed like "wow so smart" but impressed like "wow can't believe that just training to predict the next token actually works so amazingly well"

[Xeoncross]

It feels like the level of skill needed to remain above GenAI's ability to code, write, produce songs, or drawings keeps rising. All of us have strengths, things we do better than AI still (even basic companionship abilities), but I wonder how long that will be true.

[aprilthird2021]

If you have imagination, you will always be above this line.

[cgriswald]

I see that same threshold, but rather than intelligence, it is imagination, and the people below the threshold are unable to find ways to make AI useful to themselves. I think this threshold lowering is in agreement with your threshold rising: People will become more savvy.

[jxjnskkzxxhx]

> Today's AI systems are pretty impressive but they are absolutely not, not even slightly, the equivalent of Einstein + Hawking + Tao

Oh is that what the point of the article was? That is so stupid that it didnt even cross my mind.

[gjm11]

I mean, that's what the article explicitly says. (Perhaps it's all a metaphor for something else, or something, and some subtler point went over my head, in which case I owe the author an apology.)