Monday, March 11, 2024 Austin Convention Center
I wrote these notes during a keynote at South By Southwest, using an Apple Vision Pro headset and wireless Magic Keyboard. Words in direct quotes are exactly what the speakers said. The rest is paraphrase, as close to what was said as my fingers could type it. One the way home, I started correcting typos and tried to clarify content., but I ran out of time during my layover at DFW and decided to post this as a work in progress..
The full conversation between Peter and Josh Constine, former editor at large of TechCrunch, is now available at YouTube. Highly recommended!
Peter and Josh Constine (SignalFire VC, TechCrunch) take the stage.
AI makes us more human, Peter says. AI is a powerful tool that enables us to go deeper about things we’re interested in. “Kids are the purest form of what it means to be human.” They ask questions all the time, right? Right.
Peter: My kids ask a lot of great questions. Someone said ChatGPT helped them understand Shakespeare. The person just asks questions about the play, using audio with ChatGPT. It’s like a tireless professor. AI opens up more of what we’re interested about.
Josh: What is worth recalling? Do you feel we are shifting from being answerers and creators to the askers and curators?
Peter: Yes.
Josh: Will AI change our perception of ourselves?
Peter: To quiet the mind and think about some of these questions. My wife and I took a 10-day meditation retreat. It may give us our time back. Language helps me process my thinking. The act of writing is clarifying. I’ve used ChatGPT to talk about an idea, to gain insights. Just start asking questions. You will get deep into the weeds and gain insights.
Josh: At TechCrunch I only knew what I wanted to say when I got to the last sentence.
Peter: ChatGPT is awake all the time. I can explore ideas any time.
Josh: We are going from being the engine to being the compass. Why did you take this job. “You could have just called in rich.” Why take such a tough job?
Peter: I think its just really important work. I was inspired by the mission of the company. I’ve never seen technology as powerful any time in my lifetime. I keep a gratitude journal every day. It’s going to be of such benefit to humanity, and I just want to be a part of it.
Josh: How do you know where to focus?
Peter: The space is getting larger. Picking the right path—there’s a lot of pressure. We look what humans have learned so far and build that. What is the right mental model to think about writing or coding. The goal is to remove friction from things people want to do.
Josh: What does the world need from ChatGPT?
Peter: A flexible tool that removes barriers from goals people have. A doctor who uses it to summarize new research to stay current. An artist takes picture of painting and gets critique in real-time. Someone accomplished task in 20 minutes that used to take 4 hours. The co-evolution is putting product out there and see how people use it.
Josh: We don’t do less work with email. Do we get our time back?
Peter: It gives individuals that choice. AI accelerates what your vision is. Someone who rents GPUs out to startups. I can ask ChatGPT how to set up a remote server. Each individual will be able to do more.
…
Peter: My daughter and I are wired in a similar way, but she is going to have new tools. She is going to be able to do a lot more in her lifetime.
Josh: Values?
Peter: AI takes on values of person using it. There are so many different cultures in the world. You reset your brain on the airplane and notice the differences in a different culture. Different cultures have different values. Meticulous attention to detail in Japan, for example.
Josh: OpenAI is an American company. Best AI talent is at OpenAI and other companies, not Ivy League schools. Does it take on American values by default.
Peter: We work with people from all around the world. We build a better product and better future by reflecting values of the world.
Josh: Can we push American ideals?
Peter: That’s not how I do it. It should take the shape of each individual user.
Josh: Board changes, a distraction? Or better aligned with values.
Peter: I’m glad to see new board members. Glad to have them on board. There’s a lot of work to do.
Josh: Can I offer my values? Best teachers are idiosyncratic. Determination find answer, not copout. Beauty that inspires. Honest about its own limitations. You might want to consult Perplexity or Google. Fairness. Stewardship—society and earth as a whole. What am I missing?
Peter: Seeking the truth, the actual facts. Also, just being helpful. Technology in service with people.
Josh: Would you let it defend you in court?
Peter: I hope I won’t need it. I would absolutely like to have AI part of my legal team. Imagine I’m falsely accused. I have my counsel. You have an assistant listening to testimony, looking for all the case law, suggesting to the attorney: you might want to ask this question. A super-power assistant. I would love to have that on my side of the courtroom. One woman used ChatGPT for a medical issue. She went to a doctor and confirmed it, helping her child. These are already starting to happen.
Josh: I feel like it’s on the cusp of writing jokes. I have used Dr Gupta tool, which gave good answer on medical question. Health companies use AI for coding on bills. Big jumps forward. How do we become more comfortable with it?
Peter: It’s going to take time and repeated exposure, feeling it first hand. You will see and trust it in different ways. When I first got my Tesla it had Auto Pilot that was cruise control. Over time it got a little better, now change lanes. Now I trust it enough for some stretches letting it do its thing. I’m not fully there yet. The path of getting there is repeated exposure. We’re not keeping AI in the lab, so people can try it out and gain that literacy. It’s awesome my kinds get an advance peak. Being able to play with AI is important.
Josh: Doom scrolling. Health impacts. What have you learned?
Peter: At every job I’ve taken something with me. Instagram, etc. I don’t know what doomscrolling would look like with ChatGPT. It has a different shape, as a product.
Josh: It takes optimists to make great products. Someone needs to be in touch with bad aspects of human behavior. Do you have someone for that?
Peter: Yes. We think about safety as part of building the product. Before we build a feature we consult with safety team. Multiple perspectives on each product.
Josh: Would you first let your kids use ChatGPT, Google, social media.
Peter: The are 9 and under, no unsupervised access to anything. Things that help them explore their creativity. Daughter writing stories. It takes a handwritten story and types it for her, make a story book. Use Dall-E to create images.
Josh: How do we deal with not knowing if something was made by humans or AI?
Peter: In short term, it absolutely does matter. In election year it’s important. Beyond that, I don’t know if people will care? We will find out. Let’s say you get a text from a friend, well formed. How much do you care if they used autocorrect? Or you’re a VC and you get a pitch from someone who used Grammerly to make it look better. You see a billboard—do you care if it was Photoshopped? Take bread: artisan bread versus Wonderbread. The bread made by hand is what they want.
Josh: Someone needs a Certified Human label. For art and politics, need rules for disclosure. [Applause] My wife sends me a poem, I want to know if she wrote it. Will laws catch up with these things?
Peter: We need to start with societal norms. Products should adapt to that. A wedding photographer told me he sees a trend is grooms reading vows off ChatGPT. Socially enforced norms will lead to disclosure.
Josh: You should disclose it when you’re using AI for personal interactions.
Peter: Iterative release is best way for new AI abilities. Better than doing it only in lab. Boston Consulting Group compared groups using ChatGPT, others don’t. AI was beneficial. Productivity went up. Bottom half of performers improved more than top performers. AI is a great equalizer. This is about future proofing roles. But let’s think about how every person can accelerate their work with AI. The more they can use tools, get literacy. Finding ways to getting critique on something you wrote. Instead of future proofing, how might every person use AI to get that extra boost.
Josh: I hope companies empower employees to try the tools. Better than the CIO trying to figure it out. Now I can illustrate with Dall-E. How would you feel if your talent was used to train AI?
Peter: If we can make art creation flywheel faster, that will be good. Artist can accelerate how they can improve the painting. Every artist has been inspired by artists before?
Josh: Do they deserve compensation?
[Audience] “Yes, they do!”
Josh: If we keep strengthening AI while keeping it chained, eventually it breaks free. Can we make AI feel kinship with us? Are we more alike than not?
Peter: ChatGPT is told to be a helpful assistant. And also becoming a partner.
Vision Pro battery is at 6 percent. Time to shift to paper.