We Now Know How AI ‘Thinks’—and It’s Barely Thinking at All

Wall Street Journal ^ | 04/26/2025 | Story by Christopher Mims

[Lazamataz]

The big names in artificial intelligence—leaders at OpenAI, Anthropic, Google and others—still confidently predict that AI attaining human-level smarts is right around the corner. But the naysayers are growing in number and volume. AI, they say, just doesn’t think like us.

The work of these researchers suggests there’s something fundamentally limiting about the underlying architecture of today’s AI models. Today’s AIs are able to simulate intelligence by, in essence, learning an enormous number of rules of thumb, which they selectively apply to all the information they encounter.

This contrasts with the many ways that humans and even animals are able to reason about the world, and predict the future. We biological beings build “world models” of how things work, which include cause and effect.

Many AI engineers claim that their models, too, have built such world models inside their vast webs of artificial neurons, as evidenced by their ability to write fluent prose that indicates apparent reasoning. Recent advances in so-called “reasoning models” have further convinced some observers that ChatGPT and others have already reached human-level ability, known in the industry as AGI, for artificial general intelligence.

For much of their existence, ChatGPT and its rivals were mysterious black boxes.

There was no visibility into how they produced the results they did, because they were trained rather than programmed, and the vast number of parameters that comprised their artificial “brains” encoded information and logic in ways that were inscrutable to their creators. But researchers are developing new tools that allow them to look inside these models. The results leave many questioning the conclusion that they are anywhere close to AGI.

---

TOPICS:
KEYWORDS: ai; artificial; intelligence; llm

[Lazamataz]

Interesting development(s). AGI seems fairly far away after all, which corresponds to my experience with LLM codegen. It can handle the smallest tasks, but anything weighty, and it falters. Hallucinations of objects or functions that do not exist, failure to follow my instructions... you name it, the thing just doesn't operate as well as advertised.

I expect a long period of time where senior- and mid-level developers and computer analysts will still be in high demand.

[Lazamataz]

The ᎪᎡᎢᏆᎱᏆᏟᏆᎪᏞ ᏆᏁᎢᎬᏞᏞᏆᏀᎬᏁᏟᎬ ᏢᏆᏁᏀ ᏞᏆᏚᎢ

[Jonty30]

Having said that, millions of jobs are about checklists, which is what AI excels at. It doesn’t need any real intelligence to keep checking off a or b all day long and replace millions of workers in the process.

[bankwalker]

AI doesn’t think, but it is a very useful tool.

[BenLurkin]

If the “aliens” annoying Earth with their inconsiderate light shows really are robot intelligences, maybe they can take over AI for us and usher in a new age for mankind.

[Harmless Teddy Bear]

The LLM were trained on fiction. And I do mean that literally. Stolen fiction at that.

Which explains the strange stuff it comes up with.

[Texas Eagle]

AI attaining human-level smarts is right around the corner.

Omygosh! I hoped they'd set their sights higher than that.

[sjmjax]

Give it time. Change may come quicker than expected.

Keep in mind early attempts to build cars, airplanes and non sail powered ships wasn’t very impressive by modern standards.

Pretty much the same for all technology. Breakthroughs often occur in strange ways and at odd times.

[Lazamataz]

AI doesn’t think, but it is a very useful tool.

I agree with this...though for codegen, not as useful as I would have hoped.

[Lazamataz]

Having said that, millions of jobs are about checklists, which is what AI excels at. It doesn’t need any real intelligence to keep checking off a or b all day long and replace millions of workers in the process.

LLM is also quite skilled at understanding and responding to natural language. People in the Customer Service field are in serious jeopardy.

[Lazamataz]

Give it time. Change may come quicker than expected. Keep in mind early attempts to build cars, airplanes and non sail powered ships wasn’t very impressive by modern standards. Pretty much the same for all technology. Breakthroughs often occur in strange ways and at odd times.

My response?

Instant gratification takes too long.

[McGruff]

I know of some people with “artificial general intelligence”. All are liberals.

[Lazamataz]

"AI attaining human-level smarts is right around the corner."

Omygosh! I hoped they'd set their sights higher than that.

I know, right?

Look at the bright side: At least they are not trying to generate liberal-level intelligence. 😂😂😂

[sphinx]

AI as a tool will be operationally important, but the question also arises in speculative fiction, especially sci fi.

I instinctively start with the fact that a robot or android, even with very sophisticated programming, would have/could have no real understanding of pain.

Or humor. Or any other organic, emotional or spiritual reaction.

The AI could have a set of programed responses to mimic human behavior, but there is no real mind-body-soul connection. As we have discussed before.

We do not really understand this about ourselves ... and then there’s are the sociopaths.

[Ronaldus Magnus III]

Look at the bright side: At least they are not trying to generate liberal-level intelligence

That's what the government schools main focus is all about.

Two plus two can equal five, as long as your pronouns are all arranged, and you are convinced that the Globull Hoax is real.

[jeffersondem]

I don’t think there is much to AI, at least for now.

If anyone knows how to do it, have AI write a new, good country music song and post it on this thread.

I am talking about a sure enough good country music song, not more of the stuff Nashville has been churning out for the past decade.

[TexasGator]

GEMINI

Calculate the value of $7 per day for 10 years assuming a 4% interest rate per year

Okay, let’s figure out the future value of saving $7 per day for 10 years with a 4% annual interest rate. Since the contributions are made daily, we should consider the compounding effect more frequently than annually.
Here’s how we can approach this:
1. Calculate the total number of contributions:
* Number of years = 10
* Days per year = 365
* Total number of contributions = 10 \times 365 = 3650
2. Calculate the daily interest rate:
* Annual interest rate = 4% = 0.04
* Daily interest rate (approximately) = \frac{0.04}{365} \approx 0.000109589
3. Use the future value of an ordinary annuity formula:
Since the $7 is saved at the end of each day, this is an ordinary annuity. The formula for the future value of an ordinary annuity is:
FV = P \times \frac{(1 + r)^n - 1}{r}
Where:
* FV = Future Value
* P = Periodic Payment ($7)
* r = Interest rate per period (daily rate \approx 0.000109589)
* n = Total number of periods (3650)
4. Plug in the values and calculate:
FV = 7 \times \frac{(1 + 0.000109589)^{3650} - 1}{0.000109589}
FV = 7 \times \frac{(1.000109589)^{3650} - 1}{0.000109589}
FV \approx 7 \times \frac{1.491756 - 1}{0.000109589}
FV \approx 7 \times \frac{0.491756}{0.000109589}
FV \approx 7 \times 4487.28
FV \approx 31410.96
Therefore, saving $7 per day for 10 years at a 4% annual interest rate would result in approximately $31,410.96.

[Chickensoup]

Not really.

They are awful.

I insist they forward my call to a human.

[bankwalker]

Very good ... A useful tool indeed ... How long did that take?

[imabadboy99]

garbage in = garbage out.

AI is BS