When billion-dollar AIs break down over puzzles a child can do, it's time to rethink the hype

[cbabe]

https://www.theguardian.com/commentisfree/2025/jun/10/billion-dollar-ai-puzzle-break-down

When billion-dollar AIs break down over puzzles a child can do, it’s time to rethink the hype

Gary Marcus

The tech world is reeling from a paper that shows the powers of a new generation of AI have been wildly oversold

Tue 10 Jun 2025 06.18 EDT

…

Apple did this by showing that leading models such as ChatGPT, Claude and Deepseek may “look smart – but when complexity rises, they collapse”. In short, these models are very good at a kind of pattern recognition, but often fail when they encounter novelty that forces them beyond the limits of their training, despite being, as the paper notes, “explicitly designed for reasoning tasks”.

…

The Tower of Hanoi is a classic game with three pegs and multiple discs, in which you need to move all the discs on the left peg to the right peg, never stacking a larger disc on top of a smaller one. With practice, though, a bright (and patient) seven-year-old can do it.

What Apple found was that leading generative models could barely do seven discs, getting less than 80% accuracy, and pretty much can’t get scenarios with eight discs correct at all. It is truly embarrassing that LLMs cannot reliably solve Hanoi.

…

What the Apple paper shows, most fundamentally, regardless of how you define AGI, is that these LLMs that have generated so much hype are no substitute for good, well-specified conventional algorithms. (They also can’t play chess as well as conventional algorithms, can’t fold proteins like special-purpose neurosymbolic hybrids, can’t run databases as well as conventional databases, etc.)

… more …

[highplainsdem]

1. K&R! I'd bookmarked the op-ed earlier this morning while catching up with news, had planned to post

it, but am always glad to see other DUers posting about AI.

This type of AI is so badly flawed. And now both the US and UK governments want to force agencies to use it as much as possible. Idiocy.

[SheltieLover]

2. AI should be illegal!

[brush]

4. I fear for the next-gen college students as they're being allowed to use in now for their studies.

Will they be able to do critical thinking to solve the nation's problems?

[highplainsdem]

5. They'll be dumbed down and easily manipulated by politicians, advertisers, and especially the tech lords

controlling the AI they've become dependent on.

[brush]

6. Yep, it's scary. They'll be in positions of responsiblity in 20 years or so...

and tethered to AI and it's overseers.

[highplainsdem]

7. "Idiocracy" combined with "1984" - what a wonderful AI-assisted future.

[brush]

8. Ah, George Orwell, certainly ahead of his time, but the aging and senile-adjacent trump...

is certainly not who the magats visualize as Big Brother.

Oh fuck, wait. Can it possibly be the nasty, full of himself and always lecturing someone — JD Vance?

Can't be, they'll fuck it up as they've done with the economy, trump, Musk and Vance. Not to mention the border/deportation debacle.

[stopdiggin]

3. "can't fold proteins like special-purpose neurosymbolic hybrids"

where do we think the S-P neurosymbolic hybrid came from .. ?

[dchill]

9. Why do I feel like the developers of AI are just not that bright?

[Bernardo de La Paz]

11. When people don't understand a subject, they find it easy to call developers "not that bright".

Your comment is complete nonsense. I could be more specific in other settings but not here.

[Bernardo de La Paz]

10. Current consumer AI is Large Language Models, LLMs. They do not have reasoning modules. . . . nt

[eppur_se_muova]

12. Correction: Past time. Hype pushes strock prices higher, which is what it's all about.

It should be noted that the term "artificial intelligence" has actually been applied to several different varieties of algorithms, which only muddies the water. But that obfuscation benefits those who are just trying to sell a product for whatever the market will bear -- confused customers are more readily separated from their money, afraid that they'll be at a disadvantage without "AI", even if they don't really know what they're buying.

The huge power consumption of AI, especially applied to topics that can't really justify it, is enough reason to detest the hype and overuse.