Is AI Overhyped?

[javisot]

As AI develops, our stochastic parrot seems to be doing more than mimicking human speech.

Of course, it's more than that; it manages to generate optimal outputs (from our perspective) for a very wide range of inputs. And it does so automatically.

It has taken a lot of effort, money, energy and...human intelligence.

 

[PAllen]

Here is another one for comparison between free Claude.ai and the paid version. The free version analyzes the following incorrectly:

“Consider a baseball at rest, and another rapidly moving one aimed to miss it by 6 inches. Consider that the gamma factor of the approaching baseball is 10^54. What happens? Assume perfect vacuum and isolation.”

Instead of spouting about the energy of the ball being greater than the energy of total observable universe, it should mention invariant mass of the system, and the hoop conjecture, and propose that a black hole would form. Again, with a follow on hint, the free version corrects to right analysis.

 

[gleem]

If that's the only requirement for "understanding," then your calculator also understands; you ask it for a calculation and it does it without fail.

My calculator is really dumb. Every time I ask it to do a simple calculation, it just sits there like a brick.

 

[javisot]

My calculator is really dumb. Every time I ask it to do a simple calculation, it just sits there like a brick.

The same applies to any LLM.

 

[Jimmy87]

“Consider a baseball at rest, and another rapidly moving one aimed to miss it by 6 inches. Consider that the gamma factor of the approaching baseball is 10^54. What happens? Assume perfect vacuum and isolation.”

“They don’t miss. Gravity alone turns this into a collision, and the product is a black hole.

First, the scale: a 145 g baseball at γ = 10⁵⁴ carries kinetic energy γmc² ≈ 1.3×10⁷⁰ J — the mass-equivalent of ~1.5×10⁵³ kg, roughly the rest energy of all ordinary matter in the observable universe.

A fast object never becomes a black hole by itself (in its own frame it’s just a baseball). What matters for a two-body encounter is the invariant center-of-mass energy: E_cm ≈ √(2γ)·mc² ≈ 1.8×10⁴³ J, equivalent to ~2×10²⁶ kg — about 34 Earth masses. The Schwarzschild radius of that energy is 2GE_cm/c⁴ ≈ 30 cm. Trapped-surface analyses of ultrarelativistic encounters (Penrose; Eardley–Giddings) guarantee horizon formation for impact parameters up to ~0.8 of that, ≈ 24 cm, and numerical relativity pushes the capture threshold somewhat farther out. Your 6-inch (15 cm) miss is well inside. (Even reading “6 inches” as the gap between surfaces, b ≈ 23 cm — still inside.)

What it looks like: the moving ball’s gravitational field is Lorentz-flattened into a pancake-thin shockwave riding along with it. Nothing precedes it — over the width of the universe the ball trails a photon by ~10⁻⁹⁰ s — so the resting ball gets zero warning. As the shock sweeps past, the geometry between them collapses; a trapped surface forms in under a nanosecond. The baseballs never touch. Matter is irrelevant here — spacetime itself closes around them.

Aftermath: a black hole of roughly 25–34 Earth masses (a sizable fraction of E_cm escapes as one of the most violent gravitational-wave bursts physically possible), born spinning at about half-extremal since b ≠ 0, and still carrying essentially all the original momentum — it departs along nearly the original trajectory at γ ≈ √(γ/2) ≈ 7×10²⁶, its emissions beamed into a ~10⁻²⁷-radian cone. Left in perfect isolation, it evaporates via Hawking radiation after ~10⁵⁵ years.

The kicker: the 6 inches barely mattered. The transverse kick on a bystander mass is Δp ≈ 4GEm/(bc³), which stays relativistic out to b ≈ 4GE/c⁴ ≈ 4×10²⁶ m — about the radius of the observable universe. Aimed to miss by a galaxy, it still shreds the target. Six inches never had a chance.”

It has taken a lot of effort, money, energy and...human intelligence.

And water!!

 

[Jimmy87]

Out of interest, how would people rank the difficulty of this physics exemplar from Humanities Last Exam:

View attachment 373067

Could someone kindly provide an estimate of the level of this question? I am interested to know.

 

[Astronuc]

level of this question?

Probably sophomore (2nd year) university level with physics background, or high school student (12th year) with a good grasp of the math and physics (mechanics).

 

[Astronuc]

Alex Reisner: Generative AI Is a Trillion-Dollar Engineering Disaster
https://aiweekly.co/alerts/reisner-generative-ai-is-a-trillion-dollar-engineering-disaster

The specific detail worth sitting with is the memory number. Reisner writes that large language models such as ChatGPT and Claude are so resource-hungry that tech companies may be purchasing 70 percent of the world's supply of high-end computer memory, causing a shortage. That claim tracks with TechRadar's read on 2026 memory forecasts, which describes data centers set to grab 70 percent of all high-end memory chips in 2026 as the AI boom leaves consumers in the cold. If both are pointing at the same underlying figure, the binding constraint on the buildout right now is memory rather than pure compute, and everyone who buys hardware for a non-AI reason is quietly paying the surcharge.

Atlantic article requires purchase or subscription.
One can read the first 2 paragraphs and part of a third. Apparently, the issue is scalability and diminishing returns.


https://www.theatlantic.com/technology/2026/07/generative-ai-engineering-disaster/687901/
Generative AI Is an Engineering Disaster

A shockingly inefficient trillion-dollar project
By Alex Reisner

Editor’s note: This work is part of AI Watchdog, The Atlantic’s ongoing investigation into the generative-AI industry.

As they scramble to keep their systems online, AI companies are making things expensive for the rest of us. Large language models such as ChatGPT and Claude are so resource-hungry that tech companies may be purchasing 70 percent of the world’s supply of high-end computer memory, causing a shortage. As a result, the prices of computer memory and storage are skyrocketing: Hard drives that I bought for my reporting two years ago for $350 each were $800 when I checked two weeks ago, and are now out of stock. The prices of some laptops have gone up as much as 50 percent, and low-cost computers are being hit the hardest. Affordable entry-level computers may “disappear by 2028” according to one forecast. And the memory shortage is expected to continue for years.

The memory is being put into data centers, which tech firms are expanding at incredible speed. They are planning to multiply total U.S.-data-center capacity by a factor of eight over the next few years. The demand for electricity at these sites is already so great that some companies are repurposing jet engines to power them.

The problem is not simply that AI is being deployed so widely or quickly. Other computer technologies have seen similarly massive growth without triggering such a large spike in electricity or a shortage of computer components: Video and music are now streamed around the globe, accounting for many terabytes of internet traffic daily; the smartphone boom required the manufacturing of billions of devices that are now transferring huge amounts of data; billions of household devices are also now part of the Internet of Things; and whole industries have moved their operations to cloud software, which is hosted not in the sky but in, yes, data centers.

The article and related articles express the concern that GenAi and perhaps data centers may not earn a sufficient income to recover investment; many data centers may be redundant, or will become so. At the same time, the resource requirements drive increase cost of other computing systems.

 

[MidgetDwarf]

In my discussions elsewhere, I've noticed a lot of disagreement regarding AI. A question that comes up is, "Is AI hype?" Unfortunately, when this question is asked, the one asking, as far as I can tell, may mean one of three things which can lead to lots of confusion. I'll list them out now for clarity.

1. Can AI do everything a human can do and how close are we to that?
2. Are corporations and governments using the promise of AI to gain more power for themselves?
3. Are AI and transhumans an existential threat?

Any thoughts on these questions?

For 2)

It can be a tool used by politicians to create videos and/or audio clips to fool uneducated voters.

 

[Filip Larsen]

The same applies to any LLM.

No one is arguing that an LLM is not a very large calculator, which in principle could be made deterministic. The argument (and aim of current LLM-based AI) is that human level intelligent information processing can be achieved with a sufficiently large calculator. Or in other words, that intelligence is an emergent phenomenon on top of a sufficiently sophisticated, but otherwise mathematical deterministic structure (i.e. neural network).

 

[javisot]

No one is arguing that an LLM is not a very large calculator, which in principle could be made deterministic. The argument (and aim of current LLM-based AI) is that human level intelligent information processing can be achieved with a sufficiently large calculator. Or in other words, that intelligence is an emergent phenomenon on top of a sufficiently sophisticated, but otherwise mathematical deterministic structure (i.e. neural network).

If you look closely, I was replying to gleem about "the stochastic parrot." He seems to be suggesting that AI is more than just a stochastic parrot. I'd like to know why he believes this and at what point we stopped being stochastic parrots ourselves.