Pamela Mishkin knows nothing about forestry. So when the OpenAI researcher, who spends her days studying AI policy and safety, needed to find out how AI would affect forestry staff, she turned to the technology she works with on daily basis.
“I put some dense documents about forestry into the model,” she says. “Then I asked it to extract the sections that were most relevant to my query.” It’s research that will have otherwise taken her hours, accomplished much more quickly with next-generation AI.
We’ve grown used to the background AI that offers us recommendations for what to observe, read, and buy. Now, with powerful latest foundation models and accessible natural language interfaces, we’re entering a brand new phase of AI—one which empowers us to create, not only eat. To learn more, we spoke to 6 AI experts about how they use next-generation AI at work, from saving time to considering otherwise to creating speeches just a little more bejeweled.
Jaime Teevan, Microsoft Chief Scientist
“I take all my pre-read documents—you understand, you’re going into a gathering and also you’ve got a bunch of documents you’ve to read—and summarize them as poems. I did it a couple of times as a gimmick, but then I spotted I actually process the data higher—and it makes the strategy of preparing for a gathering just a little bit joyful.”
Erik Brynjolfsson, Director of the Digital Economy Lab on the Stanford Institute for Human-Centered AI
“I had to present a chat on the National Bureau of Economic Research, and I asked GPT-3 to assist me write my remarks. For fun, I had it rewrite my draft within the type of Taylor Swift. It made this totally amazing poem with these terrific metaphors that I had never heard before. Everybody on the conference thought it was just riotously fun and insightful. Combining academic work with just a little little bit of art sparked latest ways of fascinated by things. Ever since then, I’ve been listening just a little bit more to Taylor Swift because I used to be like, ‘Wow, that was some pretty good poetry there.’”
Sumit Chauhan, Microsoft CVP, Office Product Group
“I’m preparing for an off-site, and I actually have to write down this paper about AI. There’s a lot details about it in emails, in documents, in PowerPoints. I said to Microsoft 365 Copilot, ‘Generate me a document with a framing, marketing strategy, monetization, and go-to-market for AI.’ It searched all my relevant documents and emails and generated an overview, so I had a start line. Without it, I probably would’ve spent a complete week preparing. Now I actually have the time to step back and take into consideration how I should structure the conversation, the higher-level strategy. It’s giving me the creative space to give it some thought.”
Eric Horvitz, Microsoft Chief Scientific Officer
“Nature had a bit recently exploring why there was an unexpected jump in levels of methane within the atmosphere in the course of the pandemic. I input the entire paper into the model and asked quite a few questions, including, ‘Imagine that the hypotheses of those authors are incorrect. What else is perhaps happening to elucidate this data?’ And the system got here up with a stupendous set of alternate hypotheses that we would want to examine. That session, and plenty of others I’ve had within the realm of scientific exploration, shows how the system can function a scientific advisory copilot on among the hardest problems we face.”
Pamela Mishkin, researcher at OpenAI
“I actually have a really different communication style than my manager. I’m very Latest York, off-the-cuff and quick, and she or he’s more by the book. I’ll ask ChatGPT to rewrite things in her style. I can ask it to double-check an email I’ve written to ensure that that it’s clear—that it comes off as skilled. I feel it helps tone down my Latest York-ness after I’m communicating with Californians.”
Sam Schillace, Microsoft Deputy CTO
“I used to be in a Teams meeting, and we turned on closed captioning. The model—an internal experiment—took the closed captions and structured them right into a Loop document. So at the top of the meeting, we didn’t just get a transcription, we got: ‘Here’s all of the questions asked, all of the answers that got. Here’s all of the stuff that was referenced and here’s a snippet of any document that referenced that.’ We got this nicely structured log of the meeting to refer back to.”
Just as today we are able to’t imagine computing with out a keyboard, mouse, or the web, in the longer term, we won’t find a way to assume work without AI copilots that help us summarize, reason, and communicate. That is just the start of a complete latest way of working—and what we are able to accomplish with it.