Interview with Geoffrey Hinton

If there's one thing I've found never came true for me, it's almost any sentence of substantive opinion about "philosophy" which starts with "I think we'll agree"

And I do think this AI/AGI question is a philosophy question.

I don't know if you'll agree with that.

Whilst your analogy has strong elements of "consensus not required" I am less sure that applies right now, to what we think about AI/AGI. I think consensus is pretty .. important, and also, absent.

At what point does a human become intelligent? Is a 12 cell embryon intelligent? Is a newborn intelligent? Is a 1 year old intelligent?

> It's the logically night is day argument - just because we can't point to a clear cut off point

Um...what? There may be more than one of them, but precise definitions exist for the transitions between day and night. I think that is a very poor analogy to intelligence.

There are not just degrees of intelligence but different kinds. It is easier for us to understand and evaluate intelligence that is more similar ours and it becomes increasingly harder the more alien it becomes.

Given that, I don't see how you could reject that assertion that LLMs have some kind of intelligence.

Asking

The conversation (about whether AI is “intelligent”) was already absurd, I’m just pointing it out ;)

The more important conversation is about whether AI is useful, dangerous, and/or worth it. If AI is competent enough at a task to replace a human for 1/10 the cost, it doesn’t really matter if it “has a mortal soul” or “responds to sensory stimuli” or “can modify its weights in real time”, we need to be talking about what that job loss means for society.

That’s my main frustration: that the “is it intelligent” debate devolves into pointless unsettleable philosophical questions and sucks up all the oxygen, and the actual things of consequence go undiscussed.

They're useful. They're not intelligent. He invited the reproach.

This is interesting in its own right, and has propelled the computing industry since it was proposed, but it's not a measurement of intelligence. The reality is that we don't have a good measurement of intelligence, and struggle to define it to begin with.

there is a ton more but you asked for one :)

Of course, because you have different people all predicting a different future, some of them are bound to get it right. That doesn't mean the same person will be right again.

The overwhelming likely thing is that radiologist jobs will change, just like programming jobs will change.

e.g. see my comment on: Did Google and Stack Overflow "replace" programmers?

https://news.ycombinator.com/item?id=43013363

That is, I do not think programmers will be "replaced". The job will just be different; people will come to rely on LLMs for their jobs, like they rely on search engines.

Likewise, you can probably hire fewer doctors now because Google appeared in ~2000, but nobody talked about them being "replaced". There is NOT less demand for doctors.

---

It also reminds me of the prediction around self-driving cars, which is 13+ years ago at this point:

https://news.ycombinator.com/item?id=45149270

I believe Hacker News mostly fell for the hype in ~2012-2016. And even though the predictions turned out to be comically wrong, many people are still attached to them

https://en.wikipedia.org/wiki/List_of_predictions_for_autono...

i.e. I don't think Hinton will be proven "right" with ANY amount of time. The whole framing is just off.

It's not humans xor AI, it's humans + AI. And the world is not static

Some condensed source I found on the topic:

https://www.ing.com/Newsroom/News/The-more-famous-an-expert-...

Calling it "scientism" to care about these things as a way of dismissing the argument out of hand is anti-intellectualism at its worst.

moralestapia's reply contains a series of fallacies and thinking errors.

1. Straw Man Fallacy and Misrepresentation

moralestapia distorted my statement that Hinton is a well-regarded researcher into the idea that I am suggesting any statement he makes about the AI ramp is correct. This is actually the opposite of what I wrote. Despite Hinton's groundbreaking work in AI, I said he is not good at understanding commercial ramps, as talented people often cannot conceive that others cannot follow their vision.

moralestapia created a classic straw man argument: attacking a misrepresented version of the other’s position.

2. Red Herring and Totally Irrelevant Reference

moralestapia brings in Watson's racist viewpoints that got him ostracized.

Watson’s controversial statements about race have no connection to the topic at hand or to the argument being made and serve only to distract.

3a. Anchoring (Behavioral Economics) and Innuendo Effect

This next point comes from my fascination with Kahneman and Tversky—understanding their framework is incredibly important in how we relate. But more than that, I think they basically pull back the curtain and allow us to see what we do to hide the truth.

The first bit of information we’re exposed to disproportionately influences how we interpret subsequent information or people—that’s anchoring. By introducing Watson’s racist statements in response to my comments, he “anchored” the discussion with something emotionally and morally loaded.

The phrase “Do you also back up Watson's statements about race? I’m sure you don’t, as that’s not part of your training.” superficially acknowledges that I am most likely not defending racist statements. However, just raising the hypothetical (“do you also defend X?”) puts the idea into the discussion.

Research in psychology and behavioral economics confirms that once an accusation or association is mentioned, it becomes part of the mental framework—making “I know you are not” less effective at removing the implication than never introducing it in the first place.

In other words, moralestapia does a great job of saying, "look, there seems to be somebody who could possibly support racist thought," by specifically saying, "I know you don't have this viewpoint."

3b. Guilt by Association and Ad Hominem (Implied)

This is slightly more subtle: if we unpack it, moralestapia is suggesting that if I did support Hinton's commercial ramp (which I don't), then somehow I must also accept all of another person’s viewpoints. As referenced above, moralestapia introduces the notion that I may (although he says he doubts it) be a racist.

4. Hasty Generalization

moralestapia's blanket statement “Accomplished researchers can say dumb things too, it happens all the time” is true in a trivial sense, but it’s being used to generalize and dismiss Hinton’s credibility out of hand, regardless of the context or specifics. This is a generalization that bypasses engagement with the actual substance of Hinton’s expertise or my original point.

I’m a bit less picky here because you could argue that this was exactly my point: although Hinton is clearly brilliant, he can't always appreciate that others may not be able to follow or implement his vision.

But specifically, moralestapia doesn't seem to understand or appreciate Hinton. I’ll try to explain how to listen to a brilliant mind and not dismiss him with statements like "can say dumb things too." I think moralestapia is cutting himself off from an enormous amount of learning by making such dismissals or by accusing others of “Scientism.” In this case, I would also submit that moralestapia does not understand the historic context of how this was derived by Hayek. As somebody that leans heavily toward Austrian School of Economics and Hayek, it's funny of being accused of a term that I would say that I agree with as per Hayek's definition.

So, let’s discuss Hinton.

It’s already been cited in this thread that Hinton has lost credibility because he predicted in 2016 that radiologists would be replaced.

But I’d hope that most people can recognize that Hinton is very capable of making a technical judgment, just not necessarily a marketing or commercial one. That’s the root of my comment: listen to him, but understand his background.

From a technical perspective, using a tensor-based approach and all the tools we have today should, in theory, allow us to replace radiologists with AI. There’s nothing technically preventing this if a business case could be made for the investment. From this standpoint, Hinton was—and is—100% correct.

So, why was he so off?

While it’s not a rigorous argument, I’ll just mention that 1 out of every 5 dollars of our GDP is spent on medicine—basically double any other country—and that number is growing faster than inflation or GDP growth.

I’ll propose, without full argument, that a quick review shows the field is flooded with rent-seeking, as defined by Krueger & Tullock.

As an investor in the medical field, you quickly realize that breaking into medicine is a minefield of regulation, bureaucracy, and entrenched interests.

Navigating the labyrinth of government approvals and encountering medical practitioners—who often seem intent on protecting their own positions—can open your eyes to what rent-seeking really looks like. Why do we pay U.S. doctors 200–300% more than Scandinavian doctors? This is not a rigorous argument, but it should highlight that something looks wrong from 50,000 ft.

Penetration into this field is not a technical problem; it’s a political one. Hinton is a talented researcher who finished his career at a massive company called Google, shielded from market realities. That doesn’t mean we should dismiss his ability to perceive the capabilities of the technology. However, to get a valid time frame, we need to process his output.

Finally, do we say that "well I'll just listen to Stuart Russell." While I think you should listen to him, you need to recognize his impact on the technical community is much less. More than this, the specifics of what has redefined AI, which I would like to call a tensor based approach, is NOT were he made his mark in AI. On the flip side of this, Hinton is all about applying Tensors, and is the reason why he is so widely cited and recognized.

The debt to the workers almost never goes unpaid. The workers quickly call the debt to get food and shelter in return.

More often the workers fail to repay their debts to the rich. This is how you get entities like Berkshire Hathaway or Apple sitting on mountains of money. That money is the symbol of the loans that were extended to the workers, with the workers not being able to offer equivalent value in return.

Even among the rich, holding money is unusual, though. They usually like to call the debt for something of real value (e.g. land) as well.