What is AI? (Everyone thinks they know but no one can agree. And that’s a problem.)
Liam / July 12, 2024
Internet nastiness, name-calling, and other not-so-petty, world-altering disagreements
AI is sexy, AI is cool. AI is entrenching inequality, upending the job market, and wrecking education. AI is a theme-park ride, AI is a magic trick. AI is our final invention, AI is a moral obligation. AI is the buzzword of the decade, AI is marketing jargon from 1955. AI is humanlike, AI is alien. AI is super-smart and as dumb as dirt. The AI boom will boost the economy, the AI bubble is about to burst. AI will increase abundance and empower humanity to maximally flourish in the universe. AI will kill us all.
What the hell is everybody talking about?
Artificial intelligence is the hottest technology of our time. But what is it? It sounds like a stupid question, but it’s one that’s never been more urgent. Here’s the short answer: AI is a catchall term for a set of technologies that make computers do things that are thought to require intelligence when done by people. Think of recognizing faces, understanding speech, driving cars, writing sentences, answering questions, creating pictures. But even that definition contains multitudes.
And that right there is the problem. What does it mean for machines to understand speech or write a sentence? What kinds of tasks could we ask such machines to do? And how much should we trust the machines to do them?
As this technology moves from prototype to product faster and faster, these have become questions for all of us. But (spoilers!) I don’t have the answers. I can’t even tell you what AI is. The people making it don’t know what AI is either. Not really. “These are the kinds of questions that are important enough that everyone feels like they can have an opinion,” says Chris Olah, chief scientist at the San Francisco–based AI lab Anthropic. “I also think you can argue about this as much as you want and there’s no evidence that’s going to contradict you right now.”
But if you’re willing to buckle up and come for a ride, I can tell you why nobody really knows, why everybody seems to disagree, and why you’re right to care about it.
Let’s start with an offhand joke.
Back in 2022, partway through the first episode of Mystery AI Hype Theater 3000, a party-pooping podcast in which the irascible cohosts Alex Hanna and Emily Bender have a lot of fun sticking “the sharpest needles’’ into some of Silicon Valley’s most inflated sacred cows, they make a ridiculous suggestion. They’re hate-reading aloud from a 12,500-word Medium post by a Google VP of engineering, Blaise Agüera y Arcas, titled “Can machines learn how to behave?” Agüera y Arcas makes a case that AI can understand concepts in a way that’s somehow analogous to the way humans understand concepts—concepts such as moral values. In short, perhaps machines can be taught to behave.
Hanna and Bender are having none of it. They decide to replace the term “AI’’ with “mathy math”—you know, just lots and lots of math.
The irreverent phrase is meant to collapse what they see as bombast and anthropomorphism in the sentences being quoted. Pretty soon Hanna, a sociologist and director of research at the Distributed AI Research Institute, and Bender, a computational linguist at the University of Washington (and internet-famous critic of tech industry hype), open a gulf between what Agüera y Arcas wants to say and how they choose to hear it.
“How should AIs, their creators, and their users be held morally accountable?” asks Agüera y Arcas.
How should mathy math be held morally accountable? asks Bender.
“There’s a category error here,” she says. Hanna and Bender don’t just reject what Agüera y Arcas says; they claim it makes no sense. “Can we please stop it with the ‘an AI’ or ‘the AIs’ as if they are, like, individuals in the world?” Bender says.
It might sound as if they’re talking about different things, but they’re not. Both sides are talking about large language models, the technology behind the current AI boom. It’s just that the way we talk about AI is more polarized than ever. In May, OpenAI CEO Sam Altman teased the latest update to GPT-4, his company’s flagship model, by tweeting, “Feels like magic to me.”
There’s a lot of road between math and magic.
AI has acolytes, with a faith-like belief in the technology’s current power and inevitable future improvement. Artificial general intelligence is in sight, they say; superintelligence is coming behind it. And it has heretics, who pooh-pooh such claims as mystical mumbo-jumbo.
The buzzy popular narrative is shaped by a pantheon of big-name players, from Big Tech marketers in chief like Sundar Pichai and Satya Nadella to edgelords of industry like Elon Musk and Altman to celebrity computer scientists like Geoffrey Hinton. Sometimes these boosters and doomers are one and the same, telling us that the technology is so good it’s bad.
As AI hype has ballooned, a vocal anti-hype lobby has risen in opposition, ready to smack down its ambitious, often wild claims. Pulling in this direction are a raft of researchers, including Hanna and Bender, and also outspoken industry critics like influential computer scientist and former Googler Timnit Gebru and NYU cognitive scientist Gary Marcus. All have a chorus of followers bickering in their replies.
In short, AI has come to mean all things to all people, splitting the field into fandoms. It can feel as if different camps are talking past one another, not always in good faith.
Maybe you find all this silly or tiresome. But given the power and complexity of these technologies—which are already used to determine how much we pay for insurance, how we look up information, how we do our jobs, etc. etc. etc.—it’s about time we at least agreed on what it is we’re even talking about.
Yet in all the conversations I’ve had with people at the cutting edge of this technology, no one has given a straight answer about exactly what it is they’re building. (A quick side note: This piece focuses on the AI debate in the US and Europe, largely because many of the best-funded, most cutting-edge AI labs are there. But of course there’s important research happening elsewhere, too, in countries with their own varying perspectives on AI, particularly China.) Partly, it’s the pace of development. But the science is also wide open. Today’s large language models can do amazing things. The field just can’t find common ground on what’s really going on under the hood.
These models are trained to complete sentences. They appear to be able to do a lot more—from solving high school math problems to writing computer code to passing law exams to composing poems. When a person does these things, we take it as a sign of intelligence. What about when a computer does it? Is the appearance of intelligence enough?
These questions go to the heart of what we mean by “artificial intelligence,” a term people have actually been arguing about for decades. But the discourse around AI has become more acrimonious with the rise of large language models that can mimic the way we talk and write with thrilling/chilling (delete as applicable) realism.
We have built machines with humanlike behavior but haven’t shrugged off the habit of imagining a humanlike mind behind them. This leads to over-egged evaluations of what AI can do; it hardens gut reactions into dogmatic positions, and it plays into the wider culture wars between techno-optimists and techno-skeptics.
Add to this stew of uncertainty a truckload of cultural baggage, from the science fiction that I’d bet many in the industry were raised on, to far more malign ideologies that influence the way we think about the future. Given this heady mix, arguments about AI are no longer simply academic (and perhaps never were). AI inflames people’s passions and makes grownups call each other names.
“It’s not in an intellectually healthy place right now,” Marcus says of the debate. For years Marcus has pointed out the flaws and limitations of deep learning, the tech that launched AI into the mainstream, powering everything from LLMs to image recognition to self-driving cars. His 2001 book The Algebraic Mind argued that neural networks, the foundation on which deep learning is built, are incapable of reasoning by themselves. (We’ll skip over it for now, but I’ll come back to it later and we’ll see just how much a word like “reasoning” matters in a sentence like this.)
Marcus says that he has tried to engage Hinton—who last year went public with existential fears about the technology he helped invent—in a proper debate about how good large language models really are. “He just won’t do it,” says Marcus. “He calls me a twit.” (Having talked to Hinton about Marcus in the past, I can confirm that. “ChatGPT clearly understands neural networks better than he does,” Hinton told me last year.) Marcus also drew ire when he wrote an essay titled “Deep learning is hitting a wall.” Altman responded to it with a tweet: “Give me the confidence of a mediocre deep learning skeptic.”
At the same time, banging his drum has made Marcus a one-man brand and earned him an invitation to sit next to Altman and give testimony last year before the US Senate’s AI oversight committee.
And that’s why all these fights matter more than your average internet nastiness. Sure, there are big egos and vast sums of money at stake. But more than that, these disputes matter when industry leaders and opinionated scientists are summoned by heads of state and lawmakers to explain what this technology is and what it can do (and how scared we should be). They matter when this technology is being built into software we use every day, from search engines to word-processing apps to assistants on your phone. AI is not going away. But if we don’t know what we’re being sold, who’s the dupe?
“It is hard to think of another technology in history about which such a debate could be had—a debate about whether it is everywhere, or nowhere at all,” Stephen Cave and Kanta Dihal write in Imagining AI, a 2023 collection of essays about how different cultural beliefs shape people’s views of artificial intelligence. “That it can be held about AI is a testament to its mythic quality.”
Above all else, AI is an idea—an ideal—shaped by worldviews and sci-fi tropes as much as by math and computer science. Figuring out what we are talking about when we talk about AI will clarify many things. We won’t agree on them, but common ground on what AI is would be a great place to start talking about what AI should be.