Section Insights
Reflections on Leadership in AI
What insights does Emmett Shear have about the current state of AI leadership?
Emmett Shear emphasizes that while CEOs can be high performers, they are often unpredictable. He notes that attention is often misplaced on immediate events rather than the underlying forces shaping the industry. He highlights the importance of understanding information topology and the significance of recent research in AI.
- CEOs are often seen as unpredictable, especially during turbulent times.
- Focus should be on deeper industry forces rather than just immediate events.
- Understanding information topology is crucial in the context of AI advancements.
The Noise of Information
How does the concept of transformers relate to human information processing?
Shear draws a parallel between transformers in AI and human cognition, suggesting that just as AI tokens struggle to filter noise from signal, humans often get overwhelmed by information and fail to discern important news. He argues that our interconnectedness amplifies this challenge.
- Both AI and humans can struggle to filter out noise from important information.
- Interconnectedness in the digital age complicates our ability to focus.
- Understanding our cognitive processes can help improve information absorption.
The Role of CEOs in Long-Term Strategy
What is the long-term role of CEOs in guiding companies?
Shear explains that effective CEOs focus on long-term strategy rather than immediate reactions. They lay the groundwork for future success, and while poor leadership can lead to quick failures, competent leaders can sustain a company through challenges.
- CEOs should prioritize long-term strategy over short-term reactions.
- Good leadership is essential for navigating challenges in a company.
- The impact of leadership may take years to fully manifest.
The Importance of Leadership Selection
What factors should be considered in selecting company leaders?
Shear discusses the importance of leadership quality in companies, suggesting that while founders may have a selection bias towards better leadership, the focus should be on who is best suited to lead. He emphasizes the need for alignment in leadership to ensure effective decision-making.
- Leadership quality is crucial for company success.
- Selection bias may favor founders, but effective leadership is what truly matters.
- Alignment in leadership goals is essential for organizational effectiveness.
Understanding Alignment in Leadership
How does alignment among leaders affect organizational goals?
Shear explains that alignment among leaders is about finding common goals among various agents. The more shared goals there are, the more aligned the organization will be. He argues that achieving alignment is easier in smaller groups than in larger populations.
- Alignment among leaders is key to achieving organizational goals.
- Shared goals facilitate better cooperation and decision-making.
- Smaller groups are easier to align than larger, more diverse populations.
Transcript
0:00 you know that you cannot trust CEOs who wear leather jackets. >> Hm. All right. We've got Emmett Shear here, CEO, co-founder of Softmax. Of course, before that, YC, Twitch, and then the very brief role that you played in the OpenAI Sam Altman blip saga. I mean, it's a crazy day to have an AI conference in the midst of, you know, what's going to end up being like a seismic moment in in the AI industry. And you're someone who's lived in the middle of like a maelstrom of of AI, you know, history. What what's going through your mind right now as you kind of see things playing out, you know, with Jensen stepping down and everyone kind of just in very react take mode?
0:42 >> Well, you I mean, just as a starting point, you know that you cannot trust CEOs who wear leather jackets. >> Hm. >> They're just like they're just fundamentally like there's I mean, there obviously can be high performers, but like you never know what they're going to do. >> Yeah. >> but but really I think everyone everyone always pays too much attention >> -huh. >> to the to the shiny object and not enough attention to the deeper forces driving things along.
1:14 >> -huh. >> Like no one is irreplaceable. >> Hm. >> And and also, you know, of course, everyone people do make a difference. So things will change, but when I think about the things happening today that are that are, you know, seismic, it's the what we're learning about how information topology works and how the platonic realm is real. I don't know if people noticed that, but like there's there's increase incredible papers coming out. Michael Levin has a particularly good one about the platonic realm and platonic forms. But like that basically all transformers, all machine learning things learn the same it's it's all different, but the topology of the information geometry, the topology of the structures inside them is the same.
2:00 Even in the that are trained completely separately. >> Yeah. >> So, like the structure how the week days of the week relate to each other, how the how the months relate to each other is the same. That's that's weird. That's not what we expected to go find. And so like I'm really into that. Is I think that that discovery is really important going to change the course of the world. >> Right. >> And you know, probably probably changing leadership at NVIDIA does too, but like how do you know which scale to be paying attention to?
2:29 >> But your point is basically like the science and and the progress exists independent of an individual. That these sort of breakthroughs will end up being far more significant than who is directing the capital, which is essentially the job of the CEO. >> is it is it's almost certain at this point that what's more important than either of those things is you know, some some little startup just launched their thing today. Somebody at one of the big labs just figured something out that holy crap guys, have you seen this?
2:56 >> Yeah. >> Or something that's I didn't even like almost certainly something I didn't list, right? But something we paid far too much attention to the one thing of the moment that like everyone is like like we're since the world has gotten so overconnected, right? Everybody's connected to everybody all the time, all all connectivity. >> Yeah. >> And it's actually it's funny actually. We don't softmax it. We don't softmax attention. Softmax of over attention means you're forced to sort of like not pay medium levels of attention to things.
3:26 You can only pay attention to the stuff that's really important. And then stuff in the middle gets pushed down. >> Yeah. >> And what that means is like you can pay attention to your to your family, to your friends, to your job. But sort of middle attention stuff that like oh, everyone's talking about this. It's too if you which which we do in a transformer. If you apply a softmax over the over your attention distribution, you you suppress all that stuff. You suppress this seemingly maybe important, but like not not more important than the actual important things.
3:56 >> Yeah. >> And the problem is if you don't do that the transformer the the the all the all the little tokens trying to figure out what they should pay attention to. >> Yeah. >> They they start getting it's noise. They can't hear the signal anymore because they're paying attention to all this noise. >> And so but you're you're aligning that basically with the way humans are absorbing, you know, seismic news. Is that like you're so focused on the one thing, you're sort of unable to you know, disambiguate the noise with the with with the actual >> We're We're all transformers in a little recurrent transformer layer. And I'm a token and you're a token.
4:27 >> Yeah. >> And we're attending we're attending to we we we both have a a query where we're looking out and we're attending to some set of the other humans. And we're also producing >> Right. >> some some information on some channels. And we're sending whatever our our value is down those channels. And like it's just like a recurrent Yeah. metaphor This is a metaphor, right? And kind of not, but mostly a metaphor. But like I think it's a productive one because what it used to be that we weren't all connected, that people in France connected people in France and people in America connected people in America. And even before that, it was people in Texas and people in in New York and you had local newspapers.
5:04 >> Yeah. >> We haven't had non-global attention for a while. So we're all in this big all all connected thing. >> as someone who's lived through essentially this, right? Like you were kind of a a part of a piece of this sort of media craziness that ends up, you know, not me mattering that much in in the matters for you obviously and you start your own company. We talked about that. But like what do you from having lived through this, like what what can we take from this particular moment that you think helps put it in context and makes it seem more rational and not not catastrophic or or or seismic?
5:36 >> The situation is normal. It's all up. >> Yeah. >> Like that's there's a word for it. It's a it's a snafu. It's like that that this is this is the way the world is all the time. There's never you're never in a time like this. Like in the the open AI thing, like it was really important to me and I think it was really important to probably to Sam and the people on the board and people who are in at at open AI was very impactful on their lives.
6:02 and like you have you know, Mira and Ilya going and starting their own companies and like like this is a >> Everything kind of fell apart and was like reconstructed over the course of a couple of days. >> And like well no but but it wasn't because then you have Thinking Machines and you have super intelligence and I and I and I go and start Softmax and now and now Sam's back but he's back with a new board.
6:21 Like a bunch of happened a bunch of things But for like for me for for the people who are involved and I'm sure I bet Jensen will have a different life now. I bet I bet the people close to Jensen will have a different life. I bet I bet Nvidia will have people who But if you don't if you're not directly involved >> Mhm. >> it's noise. >> Yeah. >> That doesn't mean it's not important but like what different decision can you possibly make as a result? Really nothing because you have no idea It doesn't really impact you. It's not actually about you.
6:51 It feels like it is but it's not. >> you think it's not really I mean you're someone in you know in the industry you've raised money Nvidia's obviously hugely influential in the directing of chips and obviously capital for backing a lot of these things. Do you think people are overestimating the significance of one person towards just the continuance of funding and the way that money goes to startups? >> I mean so Steve Jobs Titan Titan of innovation of industry like one of you know one of I think every entrepreneur's role models in a certain way everyone wants to be Steve Jobs. I think everyone can be I think I don't think I'm actually I I don't lead much like Steve Jobs does but but I would like to He's so cool right?
7:27 and he steps down and Tim Cook Tim Cook is is is is I I'm not no aspersions but Steve Cook Tim Cook is no Steve Jobs. >> Yeah. You would be the first to say it. >> and how's your iPhone? >> Still around. >> Yes and and and like I don't know you like I like I I I haven't found it changes I'm sure over a 10-year period whether Steve Jobs is leading Apple or Tim Cook is I bet it's the company's now in a very different place.
7:50 >> Sure. >> For the next three years >> -huh. >> These things don't just stop on rails. Yeah. >> It's all on rails. You don't as a CEO if you're if you unless you are incompetent you are not steering the company today. You're steering the company years out. You're laying groundwork and and so >> yeah I probably does >> mean something. What does it mean? We don't know, and it will take years to find out. And why are you thinking about this? There's so much going on. There's so many things you could be thinking about. You're thinking about this because because the system doesn't have any brakes. Because there's it has positive Because it looks like there's three buttons on a on a Twitter post, right? It looks like a three There's a like, there's a retweet, and there's a comment. And you think that the retweet is a dunk, and the comment is a is is negative, and the like is positive.
8:36 >> Yeah. >> They're all yeah, more of this please, more of this please, more of this please. >> Yeah. >> And there is no inhibitory button. Well, there there is actually. If you double the triple dots and then don't like mute, block, less like this >> Very few people do that though. >> yeah. No, because it doesn't Because the system isn't set up There's a reason why it's behind the triple dots. >> Yeah. >> I think we should put a next to the like button. I think we should have a clown button.
9:00 >> -huh. What would that do? What what what would that what would that mean? >> It says less like this, but this is a This is a clownish thing to show me. Less like this please. >> Right. >> Don't not this. >> very rarely take the algorithm, you know? Like they kind of let it wash over them. They're very rarely directing, you know, I don't like this content, it's not relevant, things like that. I mean, I feel like people are so accustomed to >> directing it all the time exquisitely with their attention.
9:19 >> Yeah. I mean, you know, one of the things that's always fascinated me as a reporter in Silicon Valley is like, you know, the obsession with the founder and the belief that the founder is like the singularly most determinative aspect of the success of a startup. And like you know, we we have an interesting dichotomy here potentially with NVIDIA and OpenAI where, you know, Sam leaves. We never really get to experience what the company is going to be with you in control and with, you know, with >> you're never going to get that. I I was very clear going in. You have me at most 90 days and out. This is an interim thing.
9:48 >> You're here to stabilize the ship. >> here to like figure the the thought I of running someone else's 800 and rapidly growing company where you don't know the people and you're like if you're if I if I needed if I really craved money and power, I guess it would be appealing. But I've have I I have been plenty successful already. I don't need it that's a not a job I would have enjoyed. >> didn't have the megalomania to think like I'm going to come in here and make the company my own. Like that's a truly insane >> I I I went out there and I can make my own company my own. I don't I don't need to I don't I don't need to to to take the shortcut.
10:25 >> Right. Right. But anyway, but but back to the, you know, NVIDIA of it all, like if the company ends up maybe being fine, you know, it's like what he built was already so robust and the industry has so much kinetic energy and there's so much desire for the overall technology that like it's not going to end up meaning all that much in the course of things. Does that in any way change the way people view the actual significance of a founder and whether a company is founder led versus professional CEO led or >> It's so so baked into the Silicon Valley, you know >> Well, it's baked in because it's obviously true. There's no there is no NVIDIA without Jensen. Jensen is Jensen.
11:00 There is no Apple without Steve Jobs. You just look at Jobs coming back to Apple and the the the the, you know, utter directional change, right? Look at look at Microsoft under Satya and Microsoft under Ballmer. What you're really seeing is in some sense not the founder per se, but the leader. >> Yeah. >> Like the the CEO matters a lot in a company because companies are effectively they're like amplifiers for the CEO's will. >> Yeah.
11:23 >> Right? The CEO's imagination, CEO's will is being amplified throughout the company. And so who the person in that role matters a lot. historically founders are the the best because they have a sort of a a depth of understanding and moral authority that comes from it, but mostly because we're bad at hiring CEOs. And founders we're bad at firing founders too. It's just that we we 90% of the companies die. And so you've just you never never to find out.
11:50 And you and the ones who are poor leaders, the company just just just die quick fast. But if I put you in front of a new person in charge of a Fortune 500 company >> -huh. >> it's not going to die. So instead we're going to have the you get this thing where you have bad CEOs in seats for mediocre ones for a long time. >> Sure. >> And that why founders founders is just selection bias. Founders aren't better per se. They're just But they are actually because >> the ones that survived were naturally good leaders. It wasn't because they were founders, because they were a good leader.
12:15 >> But the fact that they were founders, how you know. >> Yeah. >> Right. So so that's you know, it is and it isn't. I feel like the way to think about it for me is much less this question of like whether or not does it matter whether it's the founder led or not the founder led or this this thing. It's the question of who do we want leading our companies? >> Yeah. >> And like what would it mean for us to to have a good leadership in companies and like how do we align how do we get aligned?
12:45 >> Yeah. >> And that's sort of what Softmax is about actually to to to to you know, to talk about my own my own book for a second. It's like what Softmax is working on is not AI actually. We're an we're an alignment company. Now of course, the most important thing to be aligning right now is AI. That's a little bit of a joke, right? But >> Nothing's not AI. >> but but it really is from my point of view the question is you have all these agents, human agents, artificial agents, >> Yeah.
13:06 >> and we they all have goals. They're all headed in some directions. Their goals are to some degree the same and to some degree not the same always. You can't align something to humanity because humanity is not aligned to itself. So there's no there's no singular humanity thing to align to. >> Right. >> Or if there is, it's very very long term in a way that like no individual person would recognize as aligned. And nobody wants that anyways.
13:29 I don't want I don't want people who are aligned to humanity and not to my my city, my community, my people, my friends, me. >> Yeah. >> I don't find I like those people very much. >> Interesting. >> I find those people are dangerous. They think they know what's best for me and for humanity. And they don't care too much whether they're aligned with me about it. Well, I good that I don't I message received. I'm not aligned with you. Like alignment is not something that happens.
13:55 Everyone's Well, won't someone won't someone figure out how we're going to all those people out there, they're not aligned. All those agents they're not aligned. All those they're they're they're the problem. They we need to figure out how we can control their control them and how we can influence them and how we can change them so that they're aligned. To who? Well, me, obviously. And and then and then, but more than that, >> But how do you reach consensus on alignment?
14:22 >> You don't. You don't. You never will. >> It's it's a choice. >> It's it it's a it's a reality. Alignment is a reality. Alignment is not a choice. It's not it's not a judgment. You are aligned to the degree that you share the goals. You can be aligned and not know it. >> Right. >> Alignment is a is an objective quality or intersubjective, but it it is objectively accessible. Are these two people on the same team?
14:41 Well, I don't know. Are two of the people who are play for the Lakers, are they on the same team? Yes. I can observe This is not a not a subjective I mean, it's kind of subjective, but it's not really subjective. >> things don't go well if they're not. Yeah. >> Well, no, just like the two guys playing on the Lakers, they're definitely both playing for the Lakers. Like we can all objectively agree that's true. They're on the same team. So team is not an imaginary quality. It's a very real one.
15:03 And the problem the people who need to get aligned about AI are the human beings who are interacting with, building, training the AI. The alignment you seek Are you aligned with the people in your life? Are you aligned with the people in your workplace? Are you aligned with your community? Is your community aligned with its neighbors? >> Mhm. >> Are your friends aligned with each other? Are you part of your is your is your nation aligned with other nations?
15:31 But mostly it's like you kind of earn the right to participate in this question of like the alignment with humanity for me when when you have strong relationships and you have alignment in your life. And like that's always going to be partial. No one is perfectly aligned everywhere because they're not aligned with each other. You can't be. But the but the alignment is a is a starts here. >> Right. I mean, as you're saying, like as you as you go up the stack, as you widen the number of people, the ability to have alignment gets more and more difficult because there are individual thoughts and, you know, priorities and values. But your point with with Super Max is basically to say at this fundamental level, can we get, you know, alignment at its most basic founder foundational moment where that at least produces something that feels what true and valuable to the people that are making the most important decisions.
16:27 >> Alignment is a is this question of you have you have some N agents and to what degree is the intersection of their goals the same as the union of their goals? Is it is there their shared goal They have some set of shared goals and then some set of goals that are all their goals some of them which are shared and unshared. >> Yeah. >> How big is the shared group? If they if you have a lot of shared goals you're very aligned.
16:49 >> Right. >> What's easier getting 8 billion people and aligning them or getting one human and one AI or one human and one other human and aligning them? Well it's obviously easier when it's smaller. >> Yeah. >> Duh. >> Yeah. >> And and so the alignment problem starts not with your continuous extrapolated volition for all of humanity for all time for all species. It starts with the same way it starts with whenever a human child is born.
17:15 First you're aligned to your own body so you can move your eyes and head and like you know and then you align to your to your mom and your dad >> Right. >> and then you align to your family the larger family and and to friends. And the reason you align in that order is not because I mean exposure but also just because that's the order of difficulty. First the first alignment is an alignment with one other being with yourself and then with and then with one other being and then with two other beings.
17:42 >> Yeah. >> And so that's why Softmax's approach here our our approach to alignment is bottoms up. And what does it mean to have a bottoms up alignment approach? It means to create an environment where a human and an AI can collaborate together and where we generate a lot of human AI collaboration where you're in a you're in an you are in an alignment loop where your goals and their goals are together and changing over time. So you're in an attractor an orbit.
18:09 And then scale that up. And once you have that then you can have a human and an AI and another human and an AI or two AIs or two humans and you can have you can imagine this a a and larger community >> Yeah. >> of of people sharing goals, working together, and that is the training data. >> Yeah. >> That's the training. Do you want to train alignment? Well, you then you need that's the better lesson. You must have training data where people attempt alignment and you get differential variational outcomes where sometimes you succeed and sometimes you fail. And you need the whole thing.
18:39 >> Right. Right. I mean, it's a tall task, but clearly an important one and one that I think people are still trying to I mean, when it comes to the safety and long-term success of these agents and tools, that is the one that needs to be addressed, right? Yeah. >> is not does Claude obey you? Does Chat GPT obey you? The question is does it share your goals and do you share its goals? >> Right. >> And the most strongest evidence that you have about whether or not something is aligned with you is whether or not it acts like you're aligned with it.
19:06 >> Right. >> And the AI's what we the AI alignment that we're looking for like humanity needs is that you you probably chat with what's your favorite model? What do you chat with right now? What do you What's your >> instinct. I'm I'm interacting more with that than with yeah. >> are you aligned with instinct? >> Right. >> I mean >> the question we should all are For me, it's like it's like Claude. Am I aligned with Claude? And like that's that's where the alignment is in the individual interactions and then in the group ones. But first it starts with us.
19:35 It starts with every individual person. >> Thank you so much. Appreciate it. Lots to talk about, lots to think about and appreciate you stopping by. >> Yeah, thank you.
Summary
- Shear believes that CEOs, particularly those wearing leather jackets, are often unpredictable and that no one is irreplaceable.
- He highlights the importance of understanding deeper forces in AI rather than focusing solely on individual leadership changes.
- The topology of information in AI systems, such as transformers, reveals surprising consistencies across different models, which Shear finds significant.
- He argues that the real advancements in AI will likely come from startups and smaller teams rather than established leaders.
- Shear critiques the overemphasis on individual founders in Silicon Valley, suggesting that leadership quality matters more than whether a company is founder-led.
- He introduces the concept of "alignment" in AI, advocating for a bottoms-up approach where human and AI goals can converge.
- Shear posits that alignment starts on an individual level and scales up, emphasizing the need for collaborative environments between humans and AI.
- He concludes that true alignment in AI is about shared goals and mutual understanding, rather than mere obedience.
Questions Answered
What insights does Emmett Shear have about the current state of AI leadership?
Emmett Shear emphasizes that while CEOs can be high performers, they are often unpredictable. He notes that attention is often misplaced on immediate events rather than the underlying forces shaping the industry. He highlights the importance of understanding information topology and the significance of recent research in AI.
How does the concept of transformers relate to human information processing?
Shear draws a parallel between transformers in AI and human cognition, suggesting that just as AI tokens struggle to filter noise from signal, humans often get overwhelmed by information and fail to discern important news. He argues that our interconnectedness amplifies this challenge.
What is the long-term role of CEOs in guiding companies?
Shear explains that effective CEOs focus on long-term strategy rather than immediate reactions. They lay the groundwork for future success, and while poor leadership can lead to quick failures, competent leaders can sustain a company through challenges.
What factors should be considered in selecting company leaders?
Shear discusses the importance of leadership quality in companies, suggesting that while founders may have a selection bias towards better leadership, the focus should be on who is best suited to lead. He emphasizes the need for alignment in leadership to ensure effective decision-making.
How does alignment among leaders affect organizational goals?
Shear explains that alignment among leaders is about finding common goals among various agents. The more shared goals there are, the more aligned the organization will be. He argues that achieving alignment is easier in smaller groups than in larger populations.