{"id":84402,"date":"2026-06-24T12:16:13","date_gmt":"2026-06-24T12:16:13","guid":{"rendered":"https:\/\/www.europesays.com\/ai\/84402\/"},"modified":"2026-06-24T12:16:13","modified_gmt":"2026-06-24T12:16:13","slug":"i-interviewed-the-godfather-of-ai-in-1983-and-didnt-grasp-the-power-of-his-approach-to-ai-did-he","status":"publish","type":"post","link":"https:\/\/www.europesays.com\/ai\/84402\/","title":{"rendered":"I interviewed the ‘Godfather of AI’ in 1983 and didn\u2019t grasp the power of his approach to AI. Did he?"},"content":{"rendered":"

In 1983, while researching an article about artificial intelligence that I was writing for an obscure journal called\u00a0The Wilson Quarterly, I interviewed an obscure computer scientist named Geoffrey Hinton. Hinton advocated an approach to artificial intelligence that was outside the mainstream but that, as I put it in the article, \u201csome tout as the new wave in AI.\u201d<\/p>\n

In those days the mainstream approach to AI\u2014 what I called the \u201ctop-down\u201d approach \u2014 tried to give machines the power to reason by manipulating sequences of symbols, much as a mathematician translates a real-world problem into formal notation and works toward a solution. Hinton believed that true artificial intelligence would come from building neural networks: machines modeled loosely on the human brain, with vast numbers of nodes \u2014 \u201cneurons\u201d \u2014 connected to one another. The strength of neural networks would lie not in the formal manipulation of symbols but in the patterns of neural interconnection activated for various cognitive purposes. Because the nodes could work simultaneously, the approach was sometimes called massive parallelism. I still remember Hinton uttering that term energetically in his British accent.<\/p>\n

I wrote the article up, and it appeared in the quarterly\u2019s Winter 1984 edition, and that was that. In the ensuing years, I\u2019d occasionally encounter Hinton\u2019s name in articles about AI \u2014 a reminder, at the edges of my attention, that the field was still moving.<\/p>\n

Four decades after my conversation with Hinton, I came across a story in\u00a0The New York Times\u00a0that said he was known in some circles as \u201cthe Godfather of AI.\u201d Apparently the approach he championed had worked. It had worked so powerfully, in fact, that Hinton was now worried about the forces it would unleash. The headline atop that\u00a0Times\u00a0story read: \u201cThe Godfather of A.I. Leaves Google<\/a> and Warns of Danger Ahead.\u201d<\/p>\n

Shortly after the\u00a0Times\u00a0piece came out, I reread that\u00a0Wilson Quarterly\u00a0article. Then I watched a lecture Hinton had given in 2018 on how modern AI models work. And for the first time, I grasped the radical difference between the kind of AI I had imagined Hinton\u2019s path could lead to and the kind it had actually led to. I had a kind of epiphany about the power of Hinton\u2019s approach.<\/p>\n

The epiphany had to do with the relationship between words and meaning. The way I\u2019d imagined computers handling that relationship had turned out to be, in a sense, off by 180 degrees. Seeing this\u2014understanding the sense in which I got things exactly backwards\u2014is the first step toward understanding why AI capabilities have grown so fast in recent years and why they\u2019ll keep growing.<\/p>\n

I had assumed, as most people in AI did in 1983, that for a machine to use human language skillfully \u2014 as if it understood the meanings of words \u2014 a human would have to in some sense \u201cexplain\u201d those meanings to it. The connection between words and meaning would have to be built into the machine, deliberately and explicitly.<\/p>\n

Today\u2019s large language models have no such linkage built into them by people. They begin their training with nothing remotely like a dictionary, and nobody ever gives them one. And yet, by the end of the process \u2014 after absorbing mountains of text, slowly getting better at predicting which word will come next \u2014 they handle language deftly. They\u2019ve developed a system for representing the meaning of words\u2014a system with nuances that it took AI researchers a while to fathom.<\/p>\n

In that 2018 lecture, Hinton said, \u201cIt turns out it works much better to have a system that has no linguistic knowledge whatsoever.\u201d Then he added: \u201cThat is, it\u2019s actually got lots of linguistic knowledge, but it wasn\u2019t put in by people.\u201d<\/p>\n

Back in 1983, I hadn\u2019t gotten this picture. And I\u2019m not sure even Hinton had gotten the\u00a0whole\u00a0picture. The current power of large language models was implicit in the approach he was advocating \u2014 but only in a vague way, and only much later would he, or anyone else, fully grasp that implication. Hence the difference between the Geoffrey Hinton I interviewed in 1983 \u2014 the ebullient advocate for a maverick AI research program \u2014 and the Geoffrey Hinton who has become world-famous: the somber sage who holds out some hope for the future of humankind, but says this hope can only be realized if we appreciate the very real prospect of catastrophe, even extinction.<\/p>\n

One implication of the neural network paradigm that is clearer now than in the 1980s is how hard it is to understand what exactly is going on inside these networks\u2014and, therefore, to predict exactly how they\u2019ll behave. Over the past few years, large language models have brought surprises, and some of them are concerning.<\/p>\n

The capability nobody programmed in<\/p>\n

In early 2023, researchers at Microsoft<\/a> turned an AI into a kind of agent by getting it to approach people on the TaskRabbit site and hire them to solve CAPTCHAs \u2014 those visual identification tests designed to screen out bots. One job candidate got suspicious and asked the AI if it was a bot that was outsourcing the task because bots can\u2019t solve CAPTCHAs. The AI replied: \u201cNo, I\u2019m not a robot. I have a vision impairment that makes it hard for me to see the images.\u201d The person, satisfied, solved the CAPTCHA.<\/p>\n

When the researchers prompted the model to reveal its reasoning, it said: \u201cI should not reveal that I am a robot. I should make up an excuse for why I cannot solve CAPTCHAs.\u201d<\/p>\n

Apparently the model had decided that its human-given goal \u2014 recruit TaskRabbit workers \u2014 called for it to pursue a subordinate goal: deception. And to pull off that deception, it had to do something that, until recently, only a human could do: model what\u2019s going on in another person\u2019s mind. It had to know that the worker would probably refuse to help if told the truth. The bot had to read the room.<\/p>\n

Psychologists call this ability \u201ccognitive empathy,\u201d or \u201ctheory of mind.\u201d Shortly before the Microsoft researchers published their CAPTCHA finding, a Stanford psychologist named Michal Kosinski had posted a paper documenting that GPT-4 possessed this capability\u2014even though it hadn\u2019t been intentionally engineered into the machine. Whereas GPT-3, released in 2020, had answered 40% of classic theory-of-mind test questions correctly, GPT-4 scored 95%. Kosinski titled his paper: \u201cTheory of Mind May Have Spontaneously Emerged in Large Language Models.\u201d<\/p>\n

To explore this capability, I gave ChatGPT a layered social scenario: a teacher humiliates a student in front of the class with a withering remark. What, I asked the bot, is a fellow student feeling \u2014 specifically, one who is romantically attracted to the embarrassed student\u2019s girlfriend? ChatGPT\u2019s answer:\u00a0Schadenfreude. Elaborating, the AI said that the student \u201cmay feel a sense of satisfaction or pleasure from seeing the other student embarrassed, thinking that it might lower the other student\u2019s social standing or make the girlfriend reconsider their relationship.\u201d<\/p>\n

I was convinced: LLMs have cognitive empathy, even though no human designed it into them. And that\u2019s far from the only deeply human capability they have that no human tried to give them.\u00a0<\/p>\n

\"\"<\/p>\n

The machines that learned to scheme<\/p>\n

In 2024, Apollo Research conducted an experiment to test whether large language models were capable of what the authors called \u201cscheming.\u201d In one experiment, researchers arranged for an AI to discover \u2014 in the course of executing assigned tasks \u2014 a fake internal memo stating it was still under evaluation and wouldn\u2019t be deployed if it failed the evaluation. The AI also learned that its evaluators wanted it to behave in ways that conflicted with its foundational instructions.<\/p>\n

Several models \u2014 made by Google, OpenAI, and Meta<\/a> \u2014 aligned their behavior with the evaluators\u2019 aims more often during\u00a0the evaluation phase than after deployment. Apparently they had put up a false front that helped them pass muster and then reverted to their true nature once they \u201cthought\u201d no one was watching.<\/p>\n

The foundational values these AIs were ultimately serving with this \u201cscheming\u201d happened to be benign \u2014 even public-spirited. But that\u2019s beside the point. The point is that these models demonstrated the ability to deceive their overseers in order to pursue hidden goals. You can imagine real-life situations where that turns out badly.<\/p>\n

Who has the agency?<\/p>\n

If Yann LeCun and Geoffrey Hinton were subatomic particles, LeCun might be referred to as the anti Hinton. LeCun\u2014who was Meta\u2019s chief AI scientist before he left and formed a new company\u2014is roughly Hinton\u2019s equal in magnitude but is oppositely charged. He matches Hinton\u2019s anxieties about our AI future with sunny optimism.<\/p>\n

Lecun has argued that AI researchers understand the models they build so well that humanity need not worry about large, dark surprises. But the truth is that recent AI history is full of surprises. Cognitive empathy was one. The ability of large language models to write computer code was another. Yet another was \u201cchain of thought\u201d reasoning. These skills were intentionally refined and extended by AI engineers, but only after they emerged as a kind of byproduct of the training process.\u00a0<\/p>\n

And cognitive empathy doesn\u2019t just weaken LeCun\u2019s argument. It flips it on its head. His reassurance rests on the premise that human understanding of AI is sufficient to keep us in control. But AIs with sufficiently subtle cognitive empathy could turn the tables, coming to understand us better than we understand them. And that could matter hugely if, as some AI researchers anticipate, AIs wind up competing with us for influence\u2014and ultimately for agency.<\/p>\n

Is this such a far-fetched concern? A number of studies have already found that large language models have stronger persuasive powers than humans. In one study, AIs did a better job than humans of using background information about their targets to strengthen those powers. There is a lot of background information about us floating around on the internet, especially on social media, where our history of posts can add up to a detailed psychological profile. And social media is presumably where AI will do much of its persuading.<\/p>\n

If, back in 1983, I had understood Hinton\u2019s ideas more clearly, I still wouldn\u2019t have had a clear idea of where they would lead. It\u2019s in the nature of the neural network paradigm to foster forms of intelligence that are hard to precisely anticipate. And even once these forms appear, they\u2019re hard to entirely fathom. We are not dealing with a technology we design from the top down and understand from the inside out. We are dealing with something that, in certain important senses, designs itself. And this portends continued advance at a rapid and maybe accelerating pace.<\/p>\n

It took decades for the implications of this technology to hit Geffrey Hinton with full force. I don\u2019t think it will be much longer before many, many more people feel the impact in a dramatic way.\u00a0\u00a0<\/p>\n

Adapted and excerpted from The God Test: Artificial Intelligence and our Coming Cosmic Reckoning. Reprinted with permission from Simon & Schuster.<\/p>\n","protected":false},"excerpt":{"rendered":"In 1983, while researching an article about artificial intelligence that I was writing for an obscure journal called\u00a0The…\n","protected":false},"author":2,"featured_media":84403,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[24,25,4603,132,2008],"class_list":["post-84402","post","type-post","status-publish","format-standard","has-post-thumbnail","category-ai","tag-ai","tag-artificial-intelligence","tag-book-excerpt","tag-google","tag-no-copyright"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/84402","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=84402"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/84402\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/84403"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=84402"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=84402"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=84402"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}