Subscribe to stay ahead of technology trends. Never miss future editions.
Inflection AI CEO Sean White builds AI that brings people together
As we probably know by now, most of the AI race is about IQ: more GPUs, bigger models, and a sprint towards AGI. Our latest guest thinks that’s the wrong goal. As he puts it:
“If I say I’m building AGI, it says nothing about whether your life is better.”
Meet Sean White, CEO of Inflection AI, the $4 billion frontier lab behind the chatbot Pi. He joined us to explain why AI should be designed around emotional and relational intelligence, why we’ll all use several chatbots rather than one, and why “IQ gets you answers. EQ gets you understood.”
You might not know this, but Sean has spent his career on how people and computers work together. His PhD research involved augmented reality glasses that cost millions of dollars. He ran wearables research at Nokia Labs, then served as Chief R&D Officer at Mozilla, where the company made privacy-first choices even when they cost revenue. He took over as CEO of Inflection in 2024, after co-founder Mustafa Suleyman left for Microsoft. He also sits on the board of the Earth Species Project, a nonprofit using AI to decode animal communication.
Inflection is a public benefit corporation, co-founded by LinkedIn co-founder Reid Hoffman. Pi was trained on examples of high-EQ conversations, and Sean’s team is now building what he calls relational intelligence, meaning AI that helps groups of people, not just individuals.
In our episode, we learn from Sean:
Why chasing AGI is a mistake. Sean says the problem is less about danger and more that AGI “focuses us in the wrong direction.” He explains why refocusing on human problems isn’t the same as slowing down, and why Inflection designs AI that won’t “steal more of your time and have you stay down in a dark corner.”
The future of chatbots. Why people already use three to six AI tools a week, and why AI will be even more fragmented than search engines have been since 1994. He also covers why the best AI should work in the background like the thousand processes running on your laptop, rather than leaving you to manage a thousand agents.
Privacy, trust and owning your data. Lessons from Mozilla, Inflection’s role in the Data Transfer Initiative, and his plain-English rule for AI memory: “no surprises.” He also explains why AI may be the best defence against AI-powered phishing and spam.
What AI hardware should look like. Why we only have “finite real estate” on our bodies for one watch, a ring or two and one pair of earbuds. He covers why most AR glasses are years away from being ready, and why companies should consider open-source hardware so devices like cochlear implants don’t become “a giant doorstop” when the company behind them shuts down.
Working with Reid Hoffman. What Sean learned from Reid’s habit of asking “what is the theory of the game,” why Reid responds faster than almost any entrepreneur he knows, and why Sean believes AI is pushing us to ask what it means to be a good human.
Watch or listen now across YouTube, Apple Podcasts, Spotify, and X
Download the transcript 👇
Timestamps
(0:00) Meet Sean White
(2:17) Inflection's AI Mission
(3:09) Should AI Slow Down?
(6:00) Where Inflection Is Today
(7:41) The Role of the Chatbot
(10:25) How We Should Use Chatbots
(12:43) The Chatbot Experience
(17:06) Should We Worry About Privacy?
(21:10) Who Owns Your Data?
(25:47) Building for the Next Few Years
(29:56) Emotional Intelligence in AI
(32:00) Will Inflection Make Hardware?
(40:40) AI in the Next 12 Months
(44:10) Will AI Models Collaborate?
(45:38) Working with Reid Hoffman
(50:20) What Makes a Good Human?
Inside Inflection AI: Sean White on Why Chasing AGI Is the Wrong Goal
Why Chasing AGI Is a Mistake
When we asked whether AI should slow down, Sean’s answer was neither yes nor no.
“The acceleration towards AGI is a mistake, less because of the dangers and more because it focuses us in the wrong direction.”
To him, focusing on human problems instead isn’t a slowdown. It’s “a refocusing in the right direction.” The goal at Inflection is to design against the things people fear most about AI.
“We’re not going to train the AI to somehow steal more of your time and grab more of you and have you stay down in a dark corner.”
He also pushes back on the idea that EQ means a digital boyfriend or girlfriend. For Sean, emotional intelligence is about building the interface that best understands each person.
From IQ to Four Kinds of Intelligence
Inflection started out as one of the earliest frontier labs, in a race to “throw more GPU at this thing.” Bigger models did get smarter. But the team decided that being human-centered meant understanding people, “less about how good we are at thermodynamics and differential equations.”
Pi was trained on examples of high-EQ conversations. The team then added relational intelligence, which means AI that works with groups rather than just individuals, and then agentic intelligence. Sean says that adds up to “almost four different kinds of intelligence.”
You Won’t Use Just One Chatbot
An Inflection researcher found that people use three, four, five or six different AI tools across a week. Sean keeps a graph of search engine usage from 1994 to 2026, showing a new leader every few years. He thinks AI will be even more varied than that.
He also compares today’s AI to the early Mac, when people put 15 or 20 fonts on one flyer simply because they could.
“Right now, we’re in this stage where we’re trying to shove AI everywhere.”
He isn’t sold on the idea of everyone managing a thousand agents either.
“As someone who has been in various leadership positions over the years, managing a thousand people directly is terrifying.”
His model is the thousand processes already running on your laptop: they’re working all the time, and you never have to think about them.
Privacy: No Surprises
At Mozilla, any data stored in the cloud was encrypted on the user’s device, so the company couldn’t read it. Inflection is part of the Data Transfer Initiative, which lets users take their data with them. Sean’s rule for AI memory is simple.
“You don’t want them to suddenly find that they’re hearing the phrases that they said… added to somebody else’s model, or that their memories have now been shipped off to a marketer somewhere.”
He also sees personal AI as the best defence against AI-powered phishing and spam. Tools for tricking people have become far more sophisticated, and individuals need protection too, not just systems.
The Laughter Metric
The use case that surprised Sean most was people playing multi-user games with Pi around a table. Inflection has since built a prototype from it.
“A reasonable metric is how much laughter there was in the system.”
One version of Pi could join a group text, crack jokes and keep the conversation going.
What AI Hardware Should Look Like
Sean once ran a startup built on Google Glass. The team quickly learned the total addressable market was about 1,500 Glass wearers. His approach to hardware since then is pragmatic:
Finite real estate: “You get a watch, not 15 watches. You get one or two rings, you get one pair of earbuds.”
Glasses aren’t ready: most AR systems “won’t be ready for years” for everyday use.
Brain tech is too soon: adding EEG or focused ultrasound “in the next quarter or two” is premature.
Price matters: only a few people will buy a $2,000 device, while billions already have earbuds and phones.
He also makes a case for open-source hardware. When the company behind a device shuts down, the product can become “a giant doorstop.” In the case of a cochlear implant, that means a device in someone’s head that no longer gets updates.
Lessons From Reid Hoffman
Sean started out on the engineering side. He says Reid showed him how many different things you can build. They talk often, sometimes several times a week. Their last call was about books. Their last work call was about go-to-market.
Reid’s habit of game-playing shaped one of Sean’s favourite questions: what is the theory of the game? “Am I trying to actually build the longest road here, or am I trying to build out the most cities?”
“I think he responds faster than almost any entrepreneur I know.”
What’s Next: Social AI, New Form Factors and Trust
A year ago, Sean points out, there was no Claude Code, no OpenClaw and none of today’s wave of agentic systems. Over the next year, he expects:
More AI built for groups and social settings, not just people working alone.
New kinds of models beyond LLMs, with separate pieces handling EQ and IQ.
More form factors. Some will end up in his “giant prototype drawer,” but some will work.
AI reaching people who don’t own laptops, or who think of AI as only a work tool.
A bigger focus on trust. “Enterprise trust and consumer trust are different creatures.”
He expects collaboration between AI labs to “accordion,” made harder by geopolitics. Even basic values differ: Pi’s principles aren’t the same as Anthropic’s constitution.
What It Takes to Be a Good Human
Sean calls himself a humanist. During the industrial revolution, he says, people feared becoming clockwork, and the arts produced books like Frankenstein. AI is pushing us to ask the same question again: what does it mean to be human?
“IQ gets you answers. EQ gets you understood. Relational intelligence kind of gets you back to the people in your life.”
Where to connect with us
Follow Ollie on LinkedIn: https://www.linkedin.com/in/ollieforsyth
Follow Sean on LinkedIn: https://www.linkedin.com/in/seanwhite
Visit Inflection AI: https://inflection.ai
Previous episodes include
See all previous episodes here 👉
Our partners:
Harmonic: Your go-to startup database. Visit Harmonic to learn more.
Prelint: Keeps every decision reviewed, recorded, and traceable. Head over to Prelint to learn more.
VCX: The public ticker for private tech. Head over to GetVCX.com to learn more.
