Illustration: Sarah Grillo/Axios
A new artificial intelligence thingamajig called ChatGPT set the internet abuzz this week.
Why it matters: Essentially an artificial intelligence (AI) interface that texts you like a know-it-all human, ChatGPT could portend major disruptions ahead for Big Tech — particularly for the business of search.
How it works: Simply type a prompt into its interface, not unlike how you would use Google — and instead of returning links, ChatGPT writes you back in paragraphs.
- When asked “how do you crush inflation,” ChatGPT gave an answer that Jerome Powell would probably feel OK about:
- “There are a few different ways to try to crush inflation, but most of them involve using monetary policy to manage the money supply and demand in the economy,” went the opening of the bot’s answer.
State of play: Experts are hailing ChatGPT, developed by a company called OpenAI, as a major breakthrough in the decades-long push to create a bot that texts with humans as though it were a person, too.
- ChatGPT works well enough that people are starting to grasp just how powerful a chatbot could be, as tech writer Alex Kantrowitz told me on the podcast What Next: TBD on Friday.
- “For years we’ve been hearing from tech companies like Meta, like Google or even Amazon, that voice and chat talking conversationally with computers is the future — and the actual products never really measured up,” he said. “Now we’re starting to see actually there is powerful tech here.”
- ChatGPT is much farther along than, say, the awkward text interaction you might have with a bank or airline.
- In its first five days, more than 1 million users signed up to try the ChatGPT, according to OpenAI president Greg Brockman.
🍪 The intrigue: When asked how to make chocolate chip cookies, ChatGPT gave me a short, clear recipe that looked pretty standard. Compare that to Googling, which returns links to long, overwritten blog posts that require endless scrolling before you get a recipe.
- Those recipe websites are part of a whole ecosystem of businesses built around optimizing for Google searches — including finance sites like Investopedia, media outlets and other services.
- This is part of why Google has spent so much money on chat AI. The tech giant pulled in $149 billion from its search business last year; more than half of total revenue.
Yes, but: ChatGPT has no idea whether anything it says is true.
Between the lines: Writers and teachers are already fretting about the AI’s ability to spin out essays.
- The New York Times’ Paul Krugman is asking if ChatGPT is coming for skilled jobs — he’s one of several writers who used the interface to crank out a column in the past week.
- Axios’ Ina Fried calls it “scary good,” but also worries about the fact that ChatGPT doesn’t say where it’s getting its info and can be “confidently wrong.”
Zoom out: Notably, this new AI was not released by a Big Tech company, but by a risk-taking startup. OpenAI made a splash earlier this year when it released Dall – E2, an AI you can use to create art images. Artists freaked out.
- That a small, private player is the one making waves with AI makes sense; these companies can release something experimental with less risk of angering users or shareholders or advertisers if mistakes arise.
- Open AI does have some powerful friends in the biz, though. It announced a partnership with Microsoft in 2019.
Reality check: The future isn’t here yet. Users already dug up flaws — like getting the bot to tell you how to shoplift.
- And, there are big red flags to watch for: AI is prone to reflecting the biases that humans have and to manipulation (like when a Microsoft chatbot started cheering Hitler.) ChatGPT seems to have done a somewhat better job of avoiding this so far.
What’s next: Expect to see more bots from bigger players. “It’s game time for Google,” Kantrowitz said. “I don’t think it can sit on the sidelines for too long.”