Fire Is Technology – Sarah Rose Siskind 3/5

Sarah Rose Siskind and I start on who wins the data game — Apple, Amazon, Google — and why so many AI tools are features, not products or platforms, especially once a team has to switch with you.

We get into tech storytelling: the hunger for failure stories, her line that fire is technology, user-controlled algorithms, and real wins like speech recognition that finally listens and Be My Eyes helping someone find lavender pancake batter.

Transcript

Sarah: This like one random software. We’re starting to consolidate. But the giants aren’t necessarily the leaders. And my bet is this is the only bet I will take. Whoever has most of your data is gonna win. Apple, I think is a sleeper. I think that they are eventually gonna turn around and be an industry leader because they have everything on you. If you have an iPhone, they have everything on you. I think Amazon is also huge. And I think Google is huge. And I worry for companies that I really like OpenAI.

I think what’s good is people are like investing a lot right now into ChatGPT, but they never did that before. The way Gmail is indexing on the information, they’re slow rolling their AI inbox features because they’re worried about being creepy is my suspicion. But once that becomes really good, oh my God. How that’s gonna revolutionize email.

Pablos: I think you’ve got it mostly right. The- Thanks, B plus? Yeah, B plus for sure. Is that you could kind of just wall garden your people’s data into your space. And to get it, they had to, because the data- Right. And the application are kind of the same thing.

Sarah: Yeah, like the Apple store, the app store.

Pablos: App store is, has different dynamics than that, but like if you, but if you look at like a, like a real good example is like, I don’t know, like you’re, like if you’re playing World of Warcraft, like your character, all your stuff, your gold pieces, whatever, all in that game, you can’t take it with you somewhere else. Right. You’re stuck in there if you value any of it. And there’s just no way to export that. But data overall, now you could, you can just tell Gemini, go get all my shit from Otter. Right. And it’ll do it for you.

And because it can go and under, and it can reverse engineer the data structures and the file formats and figure out how to adapt.

Sarah: Gemini could? I feel like I would have to create a widget to, okay, yeah, go on.

Pablos: All these AIs are plenty good. Because the thing is, there’s a, we’re in this moment in time where what they’re really good at is that kind of stuff. So now getting out of Otter is easy. You just tell ChatGPT or Gemini, any of them, like, go grab all my conversations from Otter, whatever, re -transcribe them with a newer, better transcription AI, figure out how to do speaker detection in a way that Otter sucked at, import it into this place.

And so basically there’s a, we’re at this point where like getting all your data is now possible in a way that wasn’t true before you had LMS that could do that. That part I think is changing things, but you’re still going to need it to be stored somewhere. And Amazon and Apple and Azure have solved that for people. They store it for you. And you’re never going to be able to, like, I store everything here on a Mac, but it’s all backed up to those places.

And to the extent that I need, higher performance or other people need to access, whoever, I’m relying on all those services anyway. There’s a, so for software, I think it plays out, but these, there’s a, like this week, there’s a big kerfuffle and on the interwebs about not Otter, but like the next generation Otter was Whisperflow. And Whisperflow is like, live transcription for controlling your computer and, interacting with AI and apps.

Sarah: Which I was so hoping would happen when I was breastfeeding. Oh Because your arms are full 99% of the time. It’s a huge job. And if I could just be like, please open my browser and I couldn’t find a way. It was so frustrating.

Pablos: And now it’s all solved. And that was, months. And that was in December. That was in December. Right. In that window, that has completely been solved. But Whisperflow was kind of like the first attempt. Then there was Super Whisper. Then there was like, and now there’s like 150 knockoffs. Every day there’s another one. And it’s like Whisperflow for legal. Whisperflow for doctors. Whisperflow for vibe coding. Whisperflow for, like there’s basically, and anybody can make one. You and I could make one before this afternoon if we want. Because with the help of Claude and Codex or whatever, we could just go build that.

And so it’s not a business. And we used to have this, we’ve always had this problem in computers. We call that a feature, not a product. And now that whole cycle has moved up upstream. It’s not a product, it’s a platform. Right. And so at some point, it’s like, oh, Otter is on that border where it’s like, is it a product? Because I can totally replace this. Or is it a platform where I rely on it for more than just the feature set or more than just what it does? I’m relying on it. Store my data, long period of time, analyze it, connect it to other people. That kind of stuff.

Sarah: The difference that I’m experiencing that you don’t experience is that I have a team of people also using this software. There you go. And these are not, some of them, like some of them are very tech forward and are like super adopters, but a lot of them are not. And the behavior change that is so kludgy and slow of like switching software of like, different databases of having different ways of interfacing and accessing the info. It’s a whole thing. I’m not enterprise level, but I am dealing with a team.

And it just really reminds me of this like proverb that is from China and Confucius and Africa. I don’t know where the proverb comes from. Okay, sure. But it’s like, if you want to go fast, go alone. If you want to go far, go together. And this is like very, never has like a phrase so perfectly encapsulated vibe coding. Where it’s like, we are prototyping at the speed of light, but the enterprise aspect is so lagging. That the differential is making me go crazy. Right.

Pablos: I get it. Everybody’s experiencing that right now. All that can be not worried about because there’s a short window in time a year from now. None of that will matter. And so if you think about the progression of computers, they went from, mainframe to, bedroom size to desktop to laptop to phone to, we’re trying to get it in your pocket. Now we’re trying to get it on your face.

Sarah: We’re trying to get the computer, which is we’re trying to get rid of the computer in your mind. Right. And for your team members, we’re trying to get rid of them having to learn anything about how the computer works. They just talk to it the way they talk to a team member. And it’ll start to understand them over time. It’ll never forget.

Pablos: It’ll coordinate and it’ll be helpful. And so that’s where we’re trying to get to. There’s a bunch of like nerdy substrate stuff about how you handle access control and permissioning and all this shit across enterprises and blah. That’s all going to get solved and you’ll just have a team that talks to each other. And talks to AIs the same way. It’s such a beautiful future. This gets into. Right now it’s you and me and nerds. Next year it will be teams and enterprises. I hope you’re right. Because, so I am obsessed with this new field that’s developing that’s called tech storytelling because I think it’s what I’ve been born for.

All right, good. Because I think the chasm between the tech, the tech communications and the popular understanding of tech is so wide. It is a chasm we are going to fall into and it’s an abyss. Wow. If you look at Gallup polls about people’s feelings about AI, it is decimation. Not only AI, all of technology. People are becoming obsessed with issues around screen time, social media and children, health and tech creating, endemic problems without talking about the benefits. People love a story about a Waymo or a cruise vehicle accidentally hurting somebody.

Sarah: It is, there is a thirst, a ravenous hunger for stories about tech failing us. And if you look at any sci -fi story about the future, it always ends up with some dystopia. About the tech turning on us. There’s some exceptions. Totally. But for the most part, it’s technology is anthropomorphized and it fails us instead of what usually happens, which is some asshole uses technology to do something like that. There is this chasm that to me is just so based in ignorance because people have misdefined technology.

And this is something I have learned deeply from you, which is that people think of software when they think of technology. And the biggest thing I say to correct them is fire is technology. And we became the species we are today because of fire. Our gut can now no longer digest raw meat. We’ve changed our whole physiology around fire and we have way bigger brains because of it. We spend like no time chewing like cows do because of fire, which is a tool, which is technology.

And if you can just open your mind so that software is just one part of technology, you’ll realize that Homo sapiens won because of tools and adaptability. Good. This is my mission in life right now. And everybody gets fire, by the way. That shit can actually kill you. Chachi BT can’t kill you right now. Can’t make paper clips. No, fire is like the worst. What’s going on with Canada? What’s going on with the Pacific Northwest? Fire is terrible sometimes. Exactly. And I think that but it’s a but we’ve developed a more mature relationship with it. And we take it for granted.

We understand it’s a tool used for good and evil. And it is so frustrating that people can’t grok that with all tools. I know it’s frustrating. And it’s hurtful to me right now because of the social environment I’m in and my personal optimism and excitement about the world we’re building. And so it’s very tough but I’ve had these pockets, these moments. To reconcile it, I think the reason I feel all those same things but what gives me, what I take heart in is that I see it as this life cycle and this developmental stages that you have to go through.

Pablos: And I think the, we’re at very early developmental stages with these, with AI and robots and so people have, and arguably we’re at early developmental stage still with screen time.

Sarah: We will, in a hundred years, no kid is gonna have a screen time problem. Maybe because screens are obsolete but we’re gonna have a, we’re gonna have to go through these developmental stages to build a more mature relationship, figure out how much is too much. In some sense, it’s exciting that people are pushing back on screen time with their kids because it means they’ve.

Pablos: It’ll evolve the technology. Because we overdid it. Our kids got fucked up by it and we need to… I don’t know that, I don’t know that, like, cutting it off completely is the right idea. Maybe a little is good. I think parents should have some volition in choosing what their kids are exposed to and how much… No, you’re totally right. And we gave all that up too much. I tend to be too contrarian to the environment I’m in but you’re totally right. Yeah, I’m the same way. I get it.

Yeah, I think that one of the big things is, like, the screens and the algorithms really are trying to optimize for attention. We’ve over -indexed on that. We’re going to have to correct. I think Jack Dorsey has a really good suggestion which is to make the algorithms come to the surface and we can choose which algorithms we want not off of our lizard brains but off of a combo of our System 1 and System 2 thinking.

A way I think we should describe that is, like, what went wrong with the algorithms is Facebook controlled all the knobs and dials instead of just giving them to the user. Right. Let me choose how to improve my feed.

Sarah: I know that you can make the best one that gets the most attention out of me but I might have other priorities because I’m getting shit in my feed that I don’t want right now. Just give me the knobs and dials. A hundred percent and I think, like, theoretically, you can improve your algorithm on Instagram and Facebook. You can say what you want to see and don’t want to see but it’s so clunky and not actually working. I don’t know that you can do that. Yes, you can. On X, for example, it’s, like, you can mute certain words.

Pablos: You can make your algorithm for you or following. That, I think, is great. But it’s not because, like, I don’t want politics in my X feed. I know that’s insane. I know that’s, like, going to 16 handles and being, like, I don’t want fro -yo but, like, I’m trying to make it so I, like, during 2020, I wanted to not hear, I just wanted jokes and the racial reckoning was happening so I tried to mute certain words like Trump and, like, Well, you couldn’t do it then.

Sarah: And I tried to mute words like white because people were always talking about, like, white people do this, white people do that and I just wanted jokes and animal gifs. And so, of course, now I don’t hear about, like, the endangered white tiger because it has the word white in it. It’s the most clunky way. Yeah, but that’s all, again, that’s a stage we went through one where we turned it all over to the platforms. They’re algorithmically controlling our minds. Now X gives you, like, one knob out of ten, let’s say. And it’s too much trouble to bother for most people.

But I think that’s much better. That at least gives them deniability. They can say, hey, your feed is controlled by you, not us. That’s where we want to land. And even if it’s buried six pages deep in the settings, at least it’s possible. No, but it needs to be better. I don’t have to explicitly say I want white tigers, but it would make it through because an LLM could make that kind of… That is my favorite thing about AI is inference because, like, so I have worked in robotics and tangentially in LLM since 2018.

Pablos: And one of the biggest problems was, like, you can search for, when you search in your email for, like, itinerary, you might have accidentally been sent something, you were sent something called agenda and you didn’t use the exact right word to find agenda. You used itinerary. Now AI understands not what you say but what you mean. There you go. Inference. And to me, that has opened up a huge world because English thrives as a language because we have the most distinct number of words. We have, like, 750,000. The second runner -up has, like, 250,000 distinct words.

Sarah: But the downside of that is because there’s so many ways to say the same thing with minor differentiations, Boolean searches become really difficult. Right. And so AI has allowed English to be this hegemonic language to preserve the nuance in our language but also to allow us to search better, to understand, what we mean. It’s awesome. I want to get back to the tech storytelling thing because I have had some wins. Yeah, the first win was, like, This is important to Me, fucking, too. And they’re not as many as I would like.

But it’s, like, the first one was that win with the women who are 60 plus. The second win is, like, so I’m doing this show on PBS that’s about, it’s, like, a daily show that’s not really daily that celebrates STEM. It’s, like, a comedy science show. And we did this one segment where we have a correspondent who has cerebral palsy and is a comedian. She’s awesome. Her name’s Tina Frimmel. And she’s using a new version of ASR, automatic speech recognition, for people with speech impairment. And she talks about how she has a huge social media following.

And whenever she tries to transcribe her sets, her jokes, for social media, the transcription sucks because she’s got a speech impairment. And she jokes that, like, she has a lot of dirty jokes and she’s, like, if I’m gonna be canceled, I wanna be canceled for something I actually said. Which is so good. And so we introduced her to this new disabled -focused disability tech, like ASR. To transcribe her speech. And she had the best quote about it. She was like, I’ve been talking at my tech for years. This is the first time it’s learned how to listen.

And I was like, if we could just let so many people know that can happen. My aunt is paralyzed. And if she, she’s now in her 80s, if she had been born just, like, a decade, two decades later, like, the world that would be open to her for smart prosthetics, exoskeletons, it, luckily, it’s not something she dwells on. But the way, like, this amazing service called Be My Eyes, like, so I did this web series for Cloudflare and one of the segments was about this service called Be My Eyes, which is a service for blind people.

Where they used to have a FaceTime call with a stranger who would volunteer to help a blind person pick out an outfit, go grocery shopping, et cetera. And it’s a great service, makes everyone feel good, but it’s not 24 -7 and sometimes blind people want to do things they don’t want another human being seeing. Yeah, sure. I’m going on a date or, they might be embarrassed.

Now AI is incredible. Yeah, of course. At identifying objects. And they can pick out, like, we had this blind guy go to a store, a grocery store that he’d gone to for years using the new Be My Eyes product and he was like, I love this pancake batter. I didn’t know that they had lavender pancake batter and I thought, that sounds awesome. I want to try that. What an interesting flavor pairing. I’m getting, like, goosebumps even thinking about it because it sounds small but it’s so important. I thought, why am I choking up? It’s just, like, it’s so meaningful for people who have lived a less than full life.

Transcript

Links

Thanks, we need good co-conspirators.