Most discussions on AI’s human-augmenting capabilities focus on what the technology can do to make us more productive, efficient, and effective at work. But there is also an opportunity for AI to play an active role in an individual’s everyday personal life.
In fact, I believe that AI tools can even help us become better versions of ourselves.
Just as AI can be trained to amplify our technical and even creative abilities, it can also be trained to help us organize and work through the many emotions and thoughts we have on a daily basis. It can be a non-judgmental listener that gives thoughtful feedback, provides useful information that helps us navigate everyday tasks, and even just keeps us company.
That’s what we are working towards at Inflection AI with the launch of Pi, your personal AI, designed to be a smart, supportive, and always available companion. When Mustafa Suleyman, Karén Simonyan – both noted pioneers in the field of AI – and I co-founded Inflection AI in early 2022, our mission was to build a tool that is designed to understand the user, rather than the other way around. We wanted it to be more than an answerer of questions, but a participant in the dialogue that helps you along your life’s path – whether that be connecting with other human beings, making decisions, figuring out how to navigate a problem at work, or more.
Mustafa and Karen have assembled an incredible team to deliver on this mission. We just released Pi last week, and it can be accessed through multiple channels – through standard text messages and messaging platforms like WhatsApp or Messenger, on your desktop,or through an iOS app. As far as the field has come in just the past year, we think we are just at the beginning of this new paradigm, where interactions between humans and AI are commonplace and natural.
“Everything that you currently do with computers is going to increasingly feel like a conversation, a back and forth, a dialogic,” says Mustafa. “Your AI is going to ask you questions, is going to clarify, is going to try and sharpen your understanding. And through iterative back and forth, you’ll be able to convey your real intent.”
The conversational companion of Pi is just the beginning. In time, we aim to build more AI tools tuned for a variety of purposes, from organizing your schedule to aiding you in developing new skills and performing time-consuming tasks.
We certainly aren’t alone in our quest to bring more AI agents to the world. Even Pi, when asked, predicted that in the coming years, “We’ll see more and more AI companies like me being developed. Eventually, I think AI companions will become ubiquitous in our daily lives.”
Mustafa joined me on the Greymatter podcast to discuss what Inflection AI has built, the importance of building safeguards into Pi from the outset, why we chose to set up the company as a public benefit corporation, and our vision for the future. You can listen to the conversation at the link below or wherever you get your podcasts.
EPISODE TRANSCRIPT
Heather Mack:
Hi everyone. Welcome to Greymatter, the podcast from Greylock, where we share stories from company builders and business leaders. I’m Heather Mack, head of editorial at Greylock.
Today, my guests are two members of the Greylock investing team, Reid Hoffman, who’s been with us since 2009, and Mustafa Suleyman, who joined at the beginning of last year.
Mustafa and Reid are also the co-founders of the consumer AI company, Inflection. The company started last year and has been incubated at Greylock.
So, both of you, thanks so much for being here.
Reid Hoffman:
Always awesome.
Mustafa Suleyman:
Thank you so much for having me. It’s great to be here.
HM:
Now, Inflection has been quietly working on a product that we’ll get to hear a lot about today, but I want to first start with some context setting.
So obviously right now is an action-packed time in AI, to say the least. Both early stage startups and the biggest tech companies on earth are working on many different tools, and mainly designed for enterprise settings. Then we have things like chatGPT, GPT-4, and DALL-E, which have been instrumental in capturing the public’s attention and really driving it home just how advanced AI has become.
Both of you have been deeply involved in AI for some time. Mustafa, you’re widely recognized as a pioneer in the sector. In 2010, you co-founded DeepMind, which was acquired by Google in 2014. You then served as head of applied AI at DeepMind until 2019 when you joined Google as VP of AI products and AI policy until you joined Greylock in January 2022.
Reid, you majored in symbolic systems at Stanford, and you’ve been investing in the sector for at least a decade, when you first funded OpenAI. You’re currently on the board of several AI focused companies, including Inflection AI and Tome, and you were on the board of OpenAI for several years.
And of course, let’s not forget the many Greymatter podcasts where you interviewed the AI chatbot, ChatGPT. And your recent book, Impromptu, which you co-wrote with GPT-4.
Now, with that as the backdrop. Give me a quick characterization of this moment. Mustafa, let’s start with you. How would you describe this moment in AI?
MS:
Surreal. I would say it’s most definitely the most exciting time in my living memory in technology. I mean, it is completely stunning. The pace of progress is surprising eventhose of us who have been at the cutting edge of developing systems like these for many, many years.
And it just feels like one of the most productive times in technology, feels like there is a kind of Cambrian explosion of different combinations of software and algorithms. Everybody is experimenting, everybody is trying new things, and I think it’s going to be one of the most productive moments for many, many years. It’s very exciting.
HM:
Reid?
RH:
Well, I think that exactly as Mustafa said, the world has begun to see this [moment], which is that we are at the crescendo of things: The internet was huge, mobile is huge, cloud computing is huge. And artificial intelligence – which in Impromptu I refer to as “amplification intelligence”, as it’s kind of a partner with us in various ways – is going to crescendo all of those things.
And I actually had language about the moment (which is in Impromptu) trying to argue for people to understand this is our “Aha” moment; our amplification of human ability, and [I argue] actually in fact, to seize on that.
One of the things I think is that so, often, so many people – even very smart people, even the creators of the technology – always start with their fears and uncertainties first versus the possibilities and the hopes about what they can get to.
And the way we get to those really great futures, the way… when the transformation of the steam engine, the transformation of the printing press, the transformation of electricity, all of these things is, “Ah, that’s the future we want to steer towards, row towards and that’s the thing that we need to be doing.” And I think that’s the moment that we’re in and I think that literally every professional activity within a small number of years – as a matter of fact, Motamedi and I wrote a piece on this last year – will be amplified by artificial intelligence.
And that’s just the beginning.
HM:
Right. Now, you announced a little over a year ago that you’d co-founded Inflection AI. At the time you described it simply as an AI-first consumer products company. There hasn’t been much else said since, except that Inflection’s attempting to flip roles for humans and computers – instead of us trying to understand how to talk to them, it’s the other way around.
Now, bring me up to speed on where you are on delivering on that goal. What has Inflection built? Mustafa, you want to start us off?
MS:
Sure. I guess one way of thinking about it is that for as long as we have been creating software, we have put the onus on us as humans to try to understand the language of machines. And we’ve had to learn their programming languages, their interfaces, and that’s been a huge constraint. And that is all about to change, because for the first time in history, computers are actually able to communicate with us in the same natural plain English language that we are using to communicate with one another now. And we think that that’s going to completely transform what it means to have a digital experience.
Everything that you currently do with computers is going to increasingly feel like a conversation, a back and forth, a dialogic. Your AI is going to ask you questions, is going to clarify, is going to try and sharpen your understanding. And through iterative back and forth, you’ll be able to convey your real intent. In fact, you’ll be able to share how it is you’re feeling, not just what it is you’re thinking, or what you need to find, but actually your emotional state. And that will create a very, very different experience to the kind of thing that software and technology has ever been able to do before.
“Everything that you currently do with computers is going to increasingly feel like a conversation.”
RH:
I think part of looking at this in terms of what we’re doing with Inflection AI is to say, how do you help people navigate their entire lives? Think of our life as a journey, some of that journey is how we work. Some of that journey is how we connect with our friends, and our family. Some of that journey is what we choose to do for fun on some particular evening. And now that you have these linguistically sophisticated, capable tools for interfacing with us (and obviously English is what we’re speaking now, but as everyone knows they can do any language. That’s part of the thing that makes them amazing). They can be great kind of personal assistants, personal artificial intelligences.
And part of this amazing team that Mustafa has assembled with expertise from many different major AI efforts – because he’s been doing this for quite some time – part of the thing is they’ve realized how is it that you make these, the Pi, the Inflection AI agent, not just an answerer of questions, but a participant in the dialogue helping you along your life’s path? Helping you connect with other human beings, helping you decide on what you might want to do, asking you questions that help you think about what you want your life to be doing, or how you want to be working. And this team, which has been working very intensely to bring about this kind of different training regime for personal intelligence, Pi, has been doing just a spectacular job.
HM:
So walk me through the experience with Pi. You open it up on your phone, how do you get up to speed? How do you get started?
MS:
So Pi is really your personal AI. It’s with you wherever you are. Sometimes you’ll want to talk to it on WhatsApp, sometimes it’ll be available on Instagram, sometimes you’ll talk to it via desktop. There’s also, of course, an iOS app.
So right from the outset, we’ve tried to establish the dynamic that this is a multi-platform experience. Wherever you need Pi to be, it should be there for you. This isn’t something that you download or an app that you keep on your phone. It’s really a lasting, and in the future, I think, ever-present relationship that you have with an AI that helps you to make sense of the world around you.
Sometimes it’ll be there to listen, provide support, and be there for you when you need to vent. And other times it’ll help you find useful information. It’s obviously super knowledgeable. It’ll present information in a really succinct, conversational and kind of informal way. But other times it’ll be funny and silly and creative and help you unblock those stuck moments where you’re trying to figure out a new idea or make a plan.
On other occasions, you can imagine it being there for you when you are about to make a tough decision. You might be thinking about relocating to a different city or changing jobs, or perhaps you’re preparing to have a tough conversation with your boss. Your AI’s going to be there as your sidekick helping you as a sounding board, helping you to prepare for important moments in your life.
And in many ways, this is the essence of what sort of marks this new era. It’s not just a sort of new platform or a new technology, but it’s an entirely new paradigm. Much like the smartphone or the internet was a new paradigm. In the future, I think everybody is going to have a personal AI and it will make us much smarter, much more productive, and I hope ultimately much more compassionate and much more kind.
HM:
So slightly more advanced than those magic eight-balls…
RH:
Shaking that eight-ball, “is AI going to be very important for the future?” “Yes!”
HM:
Yeah, I’m sure that would be the answer. But it’s also different than what we’re hearing from GPT-4. And you’ve answered a little bit, but there are a lot of tools using large language models. What is the primary differentiator? When you’re saying this is something that’s going to be more emotionally attuned and kinder, how do you build that?
MS:
Yeah, so I think that over the next few years there are going to be lots of different types of AIs and we are very much focused on your personal AI. One that is principally about helping you to be more respectful and more compassionate to yourself as a way, I think, of connecting you to the outside world, being kinder and more compassionate to friends and family and colleagues. You can kind of see it as a thought partner, a place to reflect and make sense of the world around you.
And I think that’s a different tone and a different style and people might turn to different AIs for different purposes. And our take is that most of the focus in AI so far has been around IQ and information, accessing information. But in fact, I think most people are looking for a conversation. They’re looking to form useful relationships and reflect on tricky moments in their life. And that’s where we’re stepping in.
“It’s not just a sort of new platform or a new technology, but it’s an entirely new paradigm.”
RH:
And among other things, part of what we had to do some special development for is things like it’ll remember the previous conversations it had with you. It’s an ongoing dialogue. It’s not just like, “Oh, you’ve asked about this and here’s your instant Wikipedia answer,” which is some of the amazing things that are happening with AI now. But it’s also like, “Okay, I’m talking with you.” And this has been an ongoing context and conversation and it’s kind of developing that. So it feels like your personal AI.
HM:
I have a lot of thoughts here. But I’m really curious, why do you think people want this?
MS:
So I think over the last 20 years, software has been principally useful to us. It has helped us in moments where we are trying to get something practical done. It’s been a utilitarian function- you might want to find a particular piece of information, you might want to book a ride in a car, you might want to buy your groceries, and UI has been about efficiency and getting you exactly the answer you want really, really quickly.
In this new age of technology, we can actually create much richer experiences that address the kind of full range of emotional needs and desires and wants that we actually have in the world. Providing just a little bit of affirmation, a bit of confidence boosting, providing a place to get feedback and have somebody or an AI pay attention to you and ask you questions, I think is a great service.
Many of us in the world don’t always have easy access to that kind of support. And actually I think that tends to be a huge differentiator when it comes to our health and wellbeing and our success in our lives. Just having access to social support in your life makes you feel more confident and more resilient, more able to go about being successful at work or in the home.
And I think that could be a really significant contribution that if we are successful, we might be able to make over the next few years.
RH:
I mean, I think the quality of your life is the people that you go through life with. It’s your friends. When you are fortunate enough to have healthy interactions with your family, your family. It’s your colleagues, part of why you enjoy particular work and particular groups and companies and why you stay there or move to another one, all of that. And so you are, to some degree, the combination of these various quote, unquote, “Social interactions.”
Part of what Pi is designed to do is to help you make those richer and more present, it’s another tool that’s there to amplify you, there to say, “Oh, I need to have this kind of tricky conversation with my friend Sarah or Bob, and how do I do that?” Well, you have a tool that you can talk to with it, one that we’ll remember the conversations you’ve had. And we’re all much better off when we can have that. I mean, if we could all have more people to talk to about all of these various things and they can be trained to be helping us to become our better selves and navigate these things, that’s great.
HM:
So it’s not supposed to replace your personal relationships, but it’s a dry run for a lot of things in certain ways.
RH:
Because it’s not like, “Talk to me, not to people.” It’s like, “Let me help you in how you can have really rewarding interactions with other people. And of course, you can have a rewarding interaction with me too, but it’s not a replacement, it’s a help.”
HM:
You’ve been deeply thoughtful about agents’ personal assistance and how it can impact someone’s personal life. What does the existence of Pi mean for other businesses and developers, if we’re thinking about the broader enterprise AI ecosystem?
MS:
Well, the way that we think about our organization is that we’re an AI studio. Pi is the first AI that we create, but over the next few years, I can imagine us developing a number of different AIs that have different personalities, different types of expertise, and certainly different capabilities. Some of those you could imagine being enterprise or commercial AIs that may be such that we partner very deeply with certain organizations and help them build out an AI that really reflects their brand, that can talk about their products or can help them with their consumer challenges.
And so as part of our launch, we will also be releasing our API that will give other businesses the chance to play with our foundation models. One of the things that is, I think, quite unique about our approach is that we develop our own foundation models. And over the last year or so, we’ve actually been able to build some of the best large language models in the world. And we’re very excited to share those with other developers and other enterprises that might be able to use the conversational style that we’ve been developing.
RH:
And part of, again, to kind of speak to what the team that Mustafa has assembled is by having had been at the very earliest days of this as co-founder of DeepMind, he has met great people and worked with great people at DeepMind and [Google] Brain and across in other organizations. And so assembling a group of the folks who were amongst the leading scale developers, scale researchers, inventors to bring the design thinking and culture and what would be the next generation and assembling that team and doing it. And that’s part of the reason why the models that Inflection are building are an additional contribution to the models that are great models out there. The OpenAI models are great models and are also available through APIs. And so this creates a richer landscape for developers and for people who want to build not just agents, but any kind of artificial intelligence capabilities.
MS:
Yeah. I mean, it’s kind of an incredible time because the really large models are going to be very successful, there’s no doubt about that. But there are also models that are getting much, much smaller and much more efficient every year. Now we can actually produce the same capabilities that used to take GPT-4, 175 billion parameters at 15 or 20 times smaller, fewer parameters. And those kinds of models are going to be increasingly open source. Lots of smaller developers will be able to take advantage of them.
And that’s part of what’s driving this kind of Cambrian explosion of innovation and creativity and development, which is to say that everyone everywhere is going to start playing with these large language models, integrating them into their own use cases. And I think there’s just going to be this amazing explosion of creativity.
HM:
Reid, you’ve shared a lot of insights over the years based on your time founding and scaling LinkedIn. One of the main takeaways, especially as you outlined in a discussion you had on the importance of friendship in both your personal and professional life, was identifying that it had to happen in that exact moment in time in order to really work. And why is now the right time for Pi to come out into the wild?
RH:
Well, maybe not into the wild, into civilization.
HM:
It’s true. Fair enough.
RH:
But I think part of it is obviously we’re at this huge inflection point, hence, hence the name which Mustafa and team came up with, Inflection.
HM:
It’s perfect.
RH:
Yes. Where AI is that amplification from the internet, from mobile, from cloud, that is that amplification of giving us superpowers. Well, part of it is we also want that amplification to play across the scope of our lives. Part of having a personal AI is to say, “Hey, how do I navigate many elements of my life?”
And actually having an AI that doesn’t just answer questions but asks you questions. Because by the way, part of how I discover myself in various ways is when my friends ask me questions like, “Ooh, I hadn’t thought about that question yet.” And that’s a way that I grow and I evolve.
And the reason why it’s now is because, as I put in one of our Fireside Chatbots conversations, that this is going to be the year that in some way, some of the person of the year (even though it’s tools), but the person of the year thing will be AI. Because we will be giving all of these kinds of magical moments and these superpowers to people. And some of the superpowers should be: how do we navigate our life in its broad context, of course, including work, but not just only work?
HM:
So people have been working on AI (people like Mustafa) for decades, even before that of course. But in terms of feeling it on a tangible level, it kind of feels like it just got here and it’s moving faster than we can keep up with, like that famous phrase, “Gradually then suddenly.” So that’s prompting a lot of understandable concern. How is Inflection addressing the numerous safety/privacy concerns and what are the guardrails you’re building into the product?
MS:
Yeah, that’s a great question and something that is very, very important to us at Inflection. In fact, safety is probably the most important part of the way that we build out Pi
The way that we do that is that we show both good and bad examples of the kinds of behaviors that we want Pi to imitate. Sometimes it’s a human doing that imitation, sometimes it’s other versions of the AI doing the imitation. And then Pi is able to observe side by side comparisons of two or three or four different examples of a good response to a particular question. And once it does this, tens of thousands, hundreds of thousands of times in some cases, millions of times, it learns to build a model of the kind of behavior policy that we really wanted to adhere to. That’s one of the key things that we do.
The other thing is that we pay very close attention to the ways that users interact with Pi. It’s super important that people can trust Pi over time. And we think that trust is really a function of setting clear boundaries. So Pi is very careful in proactively trying to lay out types of interactions that it’s happy with and other types of interactions that are off limits.
So for example, some people have created AIs to foster romantic relationships, that’s going to be off limits for Pi, and it’s not something that we are going to be pursuing. And so whenever Pi sees that it looks like a conversation might be heading that direction, it’s pretty firm about pushing back on that.
HM:
Friend zoning you right away?
MS:
Yeah, you get friend zoned pretty quickly, and it’s very polite and kind and respectful in the way that it friend zones you, but it’s also pretty firm. And that kind of boundary setting and clear expectations is a very fundamental part of our safety program.
RH:
And part of safety isn’t just the avoidance of the negative, something that might lead to self-harm or other harm, but it’s also the amplification of the positive. And so the idea is to say, how do you… asking you questions that help you discover things that you like about yourself, things that you may want to bring out more presently with other folks that you’re talking to you with, your friends and the people in your life or your colleagues.
So, safety is not just a, “Oh, let’s just avoid the negative.” It’s also looking at where the river of the conversation of the shared experience with Pi go? It’s trying to be in a place that leads you to a much kind of healthier, fulfilled, meaningful, happier place.
“Part of safety isn’t just the avoidance of the negative, something that might lead to self-harm or other harm, but it’s also the amplification of the positive.”
HM:
Now, in that same vein, Reid, I want to double click on this, you were involved in building social media platforms. Now, there was a lot of optimism for the power and potential of these platforms, and many of the founders and investors in these companies sort of underappreciated the potential downsides that could emerge. And now we’re kind of stuck with those issues and it’s hard to change. And how is AI different? And maybe talk about the lessons you’ve learned from social platforms that could help us not make the same mistakes with AI.
RH:
Well, Mustafa has mentioned a few of these themes already, but broadly it’s make sure you’re asking some of the right questions and creating the alignments in the right way. So for example, if we were pursuing an engagement model with Pi, and it’s like, “spend as much time with me as possible,” that could be a misalignment to what would be helpful to people’s lives.
And when you look back at LinkedIn, part of our early design of it was time saving, not time-wasting. It’s like, how do you try to make each minute as useful to the person? And if a person only uses LinkedIn for five minutes, that’s great. If a person only uses Pi for five minutes, that’s great. If it’s an amplifying, if it’s a quality experience, something that helps the person with, in the LinkedIn case, their work and their lives, in the Pi case, the broad context, which can include work, but includes what are we doing on Friday night? Or how am I having this challenging conversation with my friend, Sarah or my friend, Bob as a way of doing it? These things are that kind of thing. And you ask the questions to get those alignments, and you design your business model to have the alignment.
And so I think that’s the things we’ve learned from the social media side: when you look at all the kinds of rampant challenges in question, LinkedIn gets much fewer of those than the others, partially because of the intentional purpose beforehand and being clear about what you’re about. And similarly with Inflection and Pi, we’ve been clear about what we’re about and which ways we add value to people’s lives and which ways we try to steer away from because we don’t think that adds value.
MS:
Yeah. I think the other thing that’s super important and ties into the business model, as Reid said, it’s really important to us that we figure out a monetization mechanic that doesn’t incentivize us to keep you engaged or maybe distracted on the platform for as long as possible. And likewise, we sort of don’t want to drive the kind of addictive, dopamine hits of outrage or anger or despair or viral misinformation. These kinds of things really have incentivized people to spend longer on the platform, and sometimes that’s because they’re getting paid for it. And we are very, very conscious that that is a moment that needs to change with AI and that new business models need to emerge to do that.
One important element of that is the focus that we’ve put on the legal structure of our company. So Reid and I were very clear when we co-founded the company, and a lot of our discussions over many years actually have been around how do you fundamentally incentivize an organization to do the right thing? What does that actually mean in practice?
And a big part of it is saying, “Well, the goal of this company is to try to balance the financial needs of the stockholders with not just the needs of our users, but also the kind of people and the society that is impacted beyond just our immediate users, the wider stakeholders. And I think that the interests of society more generally need to be integrated in part of the decision-making and incentivization mechanic of a company, and a public benefit corporation provides a fiduciary obligation on the directors of the company to actively balance these considerations when making all of its decisions.
And that’s really quite different to a regular company. It doesn’t solve all the problems, and it really is a kind of experimental structure. So it might throw up challenges of its own as we get more used to how this works in practice. But for us, I think it sets out a north star and an intention, a kind of direction for us that hopefully if we’re successful in the long term, we’ll actually make a real difference.
“I think that the interests of society more generally need to be integrated in part of the decision-making and incentivization mechanic of a company.”
RH:
How it works is when you’re not a public benefit corporation, the default legal structure is you must make decisions on the basis of the interest of shareholders. And there’s some room for flexibility in that because it’s kind of the question of, well, short-term shareholders, long-term shareholders, et cetera, but it’s basically shareholders. What being a public benefit corporation allows you to do to sustain a clear mission as we have, which is a kind of beneficial healthy AI. And to then make decisions based on that, the impact of your technology for your customers or for communities.
HM:
I’m thinking about the possible other Pis out there that you would build, and if some of them were purpose-built for certain businesses or organizations. And thinking about how you would control for it doing what is in the best interest of those companies or those businesses?
MS:
Well, I mean, principally, our focus at the moment is on the consumer. And our goal is really to try to create Pi to be able to help you to uncover your intention. So at the moment, your attention is sort of a commodity that you basically trade on other platforms in return for access to information, access to connection, access to news. And the more time you spend on those platforms, the more they can be operated as free.
We’d really like to have Pi be an AI that is truly on your side, is aligned with your interest commercially and is there to help you filter and sort through credible information and do so with your interests in mind first and foremost. That’s about the legal structure, it’s about the business model, it’s also about the design mechanics that we’re trying to do the opposite of driving ultimate engagement. And those are the values that we’ll take with us when we meet and work with other partners. That’s just what we stand behind and it won’t be appropriate for everybody. Some partners may want a different set of priorities and that’ll make a lot of sense for them. And there will be many, many different providers and suppliers of AI technologies and so on. And so we are just trying to focus on the best that we can do given the sort of mission that we’ve set out for ourselves.
HM:
Got it. I have this fantasy of airlines having this amazing customer service using a tool like this, but I’ll be patient.
MS:
Exactly. I think in the future, your Pi or your personal intelligence is going to talk to the airline and it will know ahead of time that there is a delay. It’ll know that there is a better route that you can take. It’ll know that it needs to negotiate for a discount so you get some credit for your next flight. And it will help you book the hotel that you need given the delay that you’re about to encounter. And that interaction will take place between your Pi and the airline AI in plain natural language, you’ll be able to review and audit the interaction that took place on your behalf.
And that’s kind of how I see Pis, or AIs in general, is that they’re going to be your digital representatives. They’ll advocate on your behalf, they’ll negotiate on your behalf, they’ll organize things for you. It’ll almost be like having something between a chief of staff and a sort of aide or a coach or a mentor, all in one. And that’s pretty cool because from a time-saving perspective, I think it will help everybody claim back more of their own attention to be able to direct their time and their precious thoughts and their precious moments in life towards being in the real world, making new connections. And less of our time doing administration and logistics and fighting for refunds and all the other annoying things that we have to do today.
RH:
Well, the amplification is not just of the efficiency, but also of the joy and meaningfulness of the path you’re on. And by the way, once you begin to realize that future, it’s like the airline case, what’ll happen is your personal AI will reach out to and say, “Hey, by the way, I’m the personal AI for Mustafa who’s on this plane.” And by having that channel early, you can begin to enable new functions that are possible. For example, the airline agent or other weather information might come back to Mustafa’s personal AI and say, “Oh, by the way, there’s 20% chance this flight’s going to get delayed currently.”
And the things we’re saying, does that matter to know? And it’s like, oh, if there’s 20% chance, let’s rebook to something else now and let’s just handle it. Or, “No, we can wait and let’s stay in contact on what we’re doing.” All that it can handle in the flow of things to make everything just much better. Not just the, “Oh, I’m heading to the airport and it’s been rebooked now, and now I’ve got an extra couple of hours to do something.” Or, “I’m going to have to be at the airport for six hours and what am I going to do?” Those will still happen, but even that will be pre-navigated because we’ll have these personal AIs.
MS:
Yep.
HM:
Yeah, I can imagine another number of scenarios where a tool like this would be so helpful.
And so we’re talking, it’s going to amplify abilities, not replace them. Best case scenario, makes us all more productive, creative, knowledgeable, supported, listened to, advocated for, in Pi’s case.
But to be real, when we look at the people who are building AI and are thus best positioned to benefit from successful businesses, there’s a lot of familiar faces. Put more bluntly, many people who did well from the last tech waves are the same ones who are really leaning into AI right now and might do very well from this wave. Would you say that observation’s accurate?
RH:
Yeah. So look, I think it’s a great question. The most important thing is to make sure that the benefits are spreading around to many people, not just the people who have an opportunity. I feel privileged here at Greylock to invest in and to also be not just part of Web 1.0 and Web 2.0. But also if you kind of say AI, Web 3.0 or whatever as a way of doing it, and I feel lucky and fortunate. But you have folks like Mustafa who has come up from a working class background and is now an entrepreneur leading the edge in AI. And we hope to have underrepresented voices as part of this. And as part of the reason we work with women and entrepreneurs from communities of color and other folks in order to make this happen.
Look, it’s super important to be inclusive in all of this. And in terms of the benefits of the technology to make sure that’s very broad. The important thing is to make sure that the benefit is spreading around. What’s important is to bring everybody into it.
MS:
When I look back over the last 20 years of the explosion of consumer hardware, which comes off the back of 50 years of Moore’s law, reducing the cost of chips, it’s kind of a remarkable story. I mean, today, no matter whether you are a billionaire or you earn 20,000 bucks a year, pretty much everybody within that bracket gets access to the same quality hardware, both on mobile and in your laptop and on desktop.
Now, that’s a remarkably meritocratic story, and my bet is that over the next decade we’re going to see the same trajectory with respect to access to intelligence and access to kindness, compassion, and emotional support. That is going to be one of the most transformational moments in history because now no longer do you need to have resources to get access to the best tutor in the world. No longer do you need to have the comfort and fortune of having a really supportive, kind family or partner when you were growing up to encourage you and so on. That is actually now going to be available to everybody. And I think it’s going to rapidly become a very, very cheap and accessible resource.
It’s kind of incredible to think what that actually does to our civilization over the next 10 to 20 years. How that turbocharges productivity, how that brings so many new voices of creativity, entrepreneurialism and so on into the productive economy. And I think that as much as there are good reasons for us to be concerned and cautious and super attentive to the redistribution questions and to amplification, as Reid says, there’s also loads of reasons to be very optimistic and excited about what’s coming in this wave.
HM:
Okay, I’m ready for a non-judgmental listener who’s going to learn about me and help me get through life with my conversations with people. How can I access Pi? How can I get started?
MS:
Oh, great question. So Pi is available at heypi.com. It is also available on the iOS store as an app. And you can also text it. It’s available on Instagram, on Messenger and via WhatsApp.
HM:
And can I expect it to mirror my personality?
MS:
So one of our foundational commitments is that in everything that we do, we want Pi to be honest and truthful. It’s still an early version, and so it doesn’t always live up to that. It often misremembers things and gets things factually wrong. An AI is a new class of thing, it’s an “it”, it’s a piece of software that we designed.
HM:
Will my Pi act like me, have my sense of humor or…?
MS:
I mean, initially it isn’t going to adapt to your style too much. It will pick up on a few expressions that you like to use and some of your sort of favorite words. But in terms of your overall tone, we’re initially trying to design it to be fairly consistent in its own style. And that’s part of our approach of trying to create a boundaried Ai. It should be clear that Pi is distinct and separate from you and has its own entity and its own sort of style. And sometimes it reflects some of the things that you are interested in, but it doesn’t always agree with you and it has its own kind of flavor every now and then.
HM:
Great. Well, this is very exciting. I’m really excited to experiment with Pi and I’m sure many other people are as well. We’ll see what happens in the next few months.
Reid Hoffman, Mustafa Suleyman, Greylock investors and co-founders of Inflection AI. Thank you so much for joining me today.
RH:
Always a pleasure.
MS:
Thanks so much. It was a lot of fun.
