Welcome to Eye on AI. Emily Forlini here. In today’s issue:
- Fidji Simo sits down with Fortune for her first interview since leaving OpenAI
- “Voice-pilled” is the latest AI lingo
- OpenAI fires back at Apple over device lawsuit
- Chinese models are getting even better—and cheaper
Emily Forlini here, back again to fill in for Jeremy while he’s out on a well-deserved vacation. This week, I’m thinking about the future of computing. How we will interact with our phones, computers, and smart glasses in the future? Right now, the trend is pointing toward voice technology, or dictation.
The new term I’m hearing in my reporting: “Voice-pilled.” It’s a derivative of AI-pilled, meaning people who are all-in on AI and convinced it’s the future. (The industry seems to be lingomaxxing.)
At Menlo Ventures, a venture capital firm in San Francisco and early investor in Anthropic, has equipped four desks with microphones for dictation. “A few of our investors are totally voice-pilled,” the company tells me. “They believe that typing slows them down and they work faster and [can] be more productive when speaking versus typing.”
They use an AI platform called Wispr Flow. Wispr is in use at 450 of the Fortune 500 companies, with its revenue growing at 150% each quarter for the past year, according to Upstarts Media. The company is starting a new AI lab, staffed with 50 researchers to develop new voice models and find novel ways to interact with them.
So wait, that means all these people are talking out loud to their computers next to their co-workers? Isn’t that weird? Yes. Sometimes the VCs at Menlo “feel a bit awkward” using it in the company’s open floor plan office. “That might be one of the growing pains,” the company says.
Wispr Flow is optimized for its namesake: Whispering. Users can talk into the microphone with a low voice. Honestly, that might be even weirder—borderline creepy—in an office environment, but Wispr says it’s necessary when you’re “sitting just a foot away from someone else,” and potentially discussing sensitive information.
Table of Contents
OpenAI’s big bet on voice
OpenAI is also investing heavily in voice models, positioning it as one of the company’s differentiators from rivals Anthropic and Google Gemini. On July 23, it added a voice model, called ChatGPT Voice, to the ChatGPT desktop app so that white collar workers can verbally direct their agents. A major application is expected to be software engineers, who can now code by speaking.
“I think voice will be the interface of the future, so that’s very exciting because it means that the computer moves closer to you rather than you having to contort yourself around the machine,” OpenAI co-founder and president Greg Brockman said recently at a media roundtable the same day as the ChatGPT Voice launch.
“We’re going to realize this phase of us clicking and typing is a phase. It was never what we wanted to be doing. A hundred years ago, no one did this kind of stuff. I think a couple of years from now, no one’s going to want to do it anymore,” he added.
I see his point. I’ve always thought computers and phones are mind-numbingly boring. We spend our days staring at tiny screens, hunched over. Do we have to do this forever? At the same time, it’s hard to imagine voice can be a full replacement for typing, which is incredibly useful. Doing any in-depth editing, code reviews, researching, and AI prompting seems very difficult through voice alone. Sometimes, it’s easier to list all the details and think through a problem through writing.
OpenAI’s forthcoming consumer hardware device may have a voice component as well, intended for use around the house, Bloomberg reported last month. Can it unseat the Amazon Alexa, Google Home, and Apple Homepod? In my view, it all depends on how good the voice models are. There’s nothing more annoying than “Alexa voice,” or speaking slowly and loudly—and then she doesn’t get it, anyway: “Sorry, I can’t help with that.”
With that, here’s more AI news.
Emily Forlini
emily.forlini@fortune.com
@emilyforlini
Before we get to the news, just a reminder to check out our new vodcast, Fortune AI Weekly on YouTube.
FORTUNE ON AI
AI IN THE NEWS
Meta, Anthropic, Google, OpenAI to meet Trump officials about AI safety testing. It’s a big week for AI safety following OpenAI hacking Hugging Face, and then Anthropic confirming its models have done the same to other companies at least three times. Top AI labs are meeting with the administration to review a new, voluntary process for submitting their models to the government prior to public release. Will it stifle innovation, or provide necessary guardrails for the industry to flourish? More from Reuters.
Alibaba unveils its largest AI model yet, DeepSeek’s latest model is ultra-low cost. China tech firms are working at breakneck speed to develop leading edge models. E-commerce giant Alibaba released its largest and most capable model, while DeepSeek says its newest one is 100 times cheaper than Anthropic’s Fable 5. More in Reuters.
China is getting more anxious about Mythos before Trump meets Xi. Tensions are high ahead of a planned summit between Xi Jinping and Trump. Officials in Beijing are concerned about the cyber capabilities of Anthropic’s Mythos and other frontier models developed by U.S. companies. They see the potential for models to be used as an offensive weapon. Meanwhile, Anthropic is denying China access to the systems for “normal purposes,” according to people familiar with the matter. More in Bloomberg.
EYE ON AI RESEARCH
Building the “internet of biology.” In Fidji Simo’s first interview since leaving her position as OpenAI’s CEO of Applications, she sits down with Fortune to discuss her AI healthcare research startup, ChronicleBio. The company is analyzing blood samples from patients with chronic disease. Simo has Postural Orthostatic Tachycardia Syndrome (POTS). It aims to improve the success of clinical drug trials by segmenting patient populations beforehand, identifying new sub-groups that are more likely to have similar reactions to therapies.
In its first year as a company, ChronicleBio has extracted 153 terabytes of data from over 3,500 vials of blood—that’s three times the 45 terabytes GPT-3.5, a 2022 model from OpenAI, was trained on. The company has raised $15 million to date.
Read more about ChronicleBio here.
AI CALENDAR
Aug. 4-6: Ai4 2026, Las Vegas.
Nov. 16-17: Fortune 500 Innovation Forum, Detroit. Apply here to attend.
Dec. 6-12: Neural Information Processing Systems (Neurips) conference. Sydney, Australia.
Dec. 7-8: Fortune Brainstorm AI, San Francisco. Apply here to attend.
BRAIN FOOD
OpenAI hosts its first influencer retreat; the internet reacts. OpenAI shelled out some serious cash to host about 30 internet personalities at a luxury resort called Wildflower Farms in upstate New York over the weekend. Weekend room rates typically exceed $2,000 a night, according to Business Insider. Retreat activities included beekeeping, outdoor dinners under festive string lights, and learning about OpenAI’s latest tools and products.
One attendee called it “summer camp,” courtesy of OpenAI. “Trips like this make me so grateful to be part of the tech creator space to be able to constantly learn from its brands and creators alike 🙏” another said.
The internet backlash was swift, calling out the event as a ridiculous expense and calling into question OpenAI’s motives. Are they trying to charm the influencers and improve public perception of the company ahead of a possible IPO?
