The rise of generative AI may not have fundamentally reshaped the economy yet, but AI leaders expect bigger changes are coming.
Brad Lightcap, chief operating officer of OpenAI, during the Bloomberg Technology Summit in San Francisco, California, US, on Thursday, May 9, 2024.
Photographer: David Paul Morris/BloombergHi! It’s Shirin from San Francisco. AI leaders say it will take more time before the world sees the real economic impact of artificial intelligence. Plus: OpenAI is about to unveil product updates. But first…
Three things to know:
• Apple plans to power AI tools with in-house server chips this year
• OpenAI rival Anthropic defended its partnerships with Amazon and Google
• Reddit laid out a new content policy while seeking more AI licensing deals
When OpenAI first released ChatGPT in late 2022, the product set off a wave of excitement that artificial intelligence would supercharge worker productivity and create new jobs — and fear that it might make other roles obsolete.
Nearly 18 months later, the impact is more subtle. While there have been some layoffs driven by AI, we have not seen entire professions replaced or created. It’s also unclear how much more work is getting done because of AI. Earlier this year, Federal Reserve Chair Jerome Powell said generative AI could boost productivity growth, but “probably not in the short run.”
So does that mean we’ve overhyped the potential of AI? On Thursday, I asked this question to OpenAI Chief Operating Officer Brad Lightcap at Bloomberg’s Tech Summit. His answer: It’s still too soon to say.
“We have a year to reshape the economy? That’s a tall order!” Lightcap joked in our chat onstage. If OpenAI halted all future AI development, which it’s certainly not doing, he said it would take significant time for the full impact of its current models to be felt. Lightcap said he’s “fairly confident at this point that there would be this 10- to 20-year diffusion period through the economy of just GPT-4 level technology.”
Lightcap’s sentiment was echoed by others at the event. AI leaders doubled down on the longer-term disruptive potential of the technology even as some tacitly acknowledged its comparatively muted effects in the early days.
“We are probably going to overestimate this technology in the short-term,” Clement Delangue, chief executive officer of AI developer platform Hugging Face, told me. “Maybe we’re going to be disappointed in a year, or two years, but we’ve also probably underestimated the long-term impact of AI in 10 years. The world is going to be very different.”
AI companies are racing to lay the groundwork for that future. In recent months, OpenAI — like much of the AI sector — has been expanding its enterprise business, selling a customized version of ChatGPT to hundreds of corporate customers including Moderna and Estee Lauder. Many of these companies are still in the relatively early days of implementing the tool into their day-to-day workflow.
“The thing we study is how do these systems really make differences in businesses,” Lightcap said.
OpenAI and its peers are also betting on so-called “AI agents” that can perform tasks for us with little to no human supervision. Both Lightcap and Anthropic CEO Dario Amodei agreed that in order to really make AI indispensable to people, it will need to do more on their behalf than spit out a few clever sentences.
“The next systems will have to cross a utility and capability gap of being able to go out in the world and be useful and go do things,” Lightcap said. Likewise, Amodei said the chatbots of today will be extended to become the “personal assistant that takes actions, that interacts with all the stack of existing technology for you.”
Beyond that, Lightcap said the public may still be dealing with a learning curve for getting the most out of AI. “People somewhat have a misguided sense of what these models can and should be used for,” he said. They’re not meant to be “databases” that can “recall facts,” but instead to reason about new information.
To that end, OpenAI is working with publishers to “bring more information into [AI models’] field of view, which we ultimately think is good for the world,” As for whether these efforts are also good for the economy, only time will tell.
Got a question about AI? Email me, Shirin Ghaffary, and I'll try to answer yours in a future edition of this newsletter.
OpenAI is planning to hold a livestreamed event on Monday to show off updates to ChatGPT and GPT-4 amid feverish rumors about the company’s next big product.
“We’ve been hard at work on some new stuff we think people will love,” OpenAI CEO Sam Altman said in a social media post on Friday. “Feels like magic to me.”
The tech world has been abuzz with speculation in recent days. A powerful new chatbot surfaced online that may or may not be from OpenAI — and which received winking acknowledgement from Altman. Users also spotted a web page called search.chatgpt.com that briefly re-routed to chatgpt.com.
As my colleague Rachel Metz reported this week, OpenAI is developing a feature for ChatGPT that can search the web and cite sources in its results, challenging Google and AI startup Perplexity.
In his post Friday, Altman said next week’s event will not include a search engine or GPT-5, a long awaited sequel to GPT-4, the large language model that powers the startup’s AI services. However, he said nothing about GPT-4.5.
Get Tech Daily and more Bloomberg Tech weeklies in your inbox:
