Can AI be good? Josh Tyrangiel in conversation with Paul Ford

Summary

Author and journalist Josh Tyrangiel joins Aboard’s Paul Ford and Rich Ziade to discuss his new book AI for Good, a deliberate counter to the noisy polemics around AI. Rather than profiling the labs and billionaires, Tyrangiel went looking for the “downstream” people — teachers, doctors, and bureaucrats — who stumbled into AI to solve specific, tangible problems, often with no software background. The through-line is that there are no home runs in the book: progress comes from stubborn, sincere, non-technical people willing to grind through institutional friction, be the “jerk who says no,” and fight software vendors until the tools actually fit the work.

Tyrangiel walks through a series of concrete case studies from the Cleveland Clinic — AI scribe software that saves doctors hours a week (though half the volunteers never used it), and sepsis prediction software that flags patients red/yellow/green and, once integrated into the workflow, cut in-hospital sepsis mortality by 41%. He recounts Palantir engineers building a working patient-flow prototype in three days, and tells the story of physicist Christie Jungst, who left her field to build AI-powered communication tools for nonverbal children after her son was diagnosed nonverbal — synthesizing a library of 7,000 sounds into 80,000 and drafting off zero-shot translation research to attack a problem “we didn’t even know we wanted to solve.”

Looking ahead, Tyrangiel frames his optimism around an unlikely source: public anger. After 25 years of passive acceptance of technology, people booing Eric Schmidt at graduation speeches signal a healthier, less passive democracy that is finally forcing companies and government to reckon with consequences. He argues the labs are a “very different kind of intelligence” trapped in a bubble who simply aren’t thinking about what good AI should do — so it falls to everyone else to deploy it well. He closes on education, where he sees the real leverage in empowering teachers to escape “teaching to the middle” and personalize learning, and offers the everyday public a way in: photograph your dying houseplant, ask an AI what to do.

Highlights

”A thousand people who are alive because some people did some work”

“These are boring words, but these are words that ended up reducing the mortality in the hospital by 41%. So that’s a thousand people who are alive, more than that, because they put together a really decent AI bridge and software and people used it.” — Josh Tyrangiel, 16:33

yt-dlp --download-sections "*16:33-17:35" "https://www.youtube.com/watch?v=GHQM81L7ALI" --force-keyframes-at-cuts --merge-output-format mp4 -o "GHQM81L7ALI-16m33s.mp4"

”This shit mostly works”

“It to reduce it to the simplest terms, because this shit mostly works, right? And not that much stuff in the world does. And so the ability to get something that works and is really efficient and actually can improve your product and your ability to care for people, um, it doesn’t come along all that often.” — Josh Tyrangiel, 21:00

yt-dlp --download-sections "*21:00-21:55" "https://www.youtube.com/watch?v=GHQM81L7ALI" --force-keyframes-at-cuts --merge-output-format mp4 -o "GHQM81L7ALI-21m00s.mp4"

”We’re going to take your job, then kill you”

“But you can’t just other them and say they’re terrible and not recognize that the tech they’re making is actually really impressive. It is odd that their marketing messages are, ‘We’re going to take your job, then kill you.’” — Josh Tyrangiel, 24:00

yt-dlp --download-sections "*24:00-24:58" "https://www.youtube.com/watch?v=GHQM81L7ALI" --force-keyframes-at-cuts --merge-output-format mp4 -o "GHQM81L7ALI-24m00s.mp4"

”It’s great that people are angry”

“If you lived through the last 25 years of technology in America and weren’t mad, I’d be really fucking worried about you. Just be like what have you not been paying attention to any of this?” — Paul Ford, 31:03

yt-dlp --download-sections "*31:03-31:44" "https://www.youtube.com/watch?v=GHQM81L7ALI" --force-keyframes-at-cuts --merge-output-format mp4 -o "GHQM81L7ALI-31m03s.mp4"

”A problem we didn’t even know we wanted to solve”

“Living in a world where a million people who were just not connected at all might actually be able to lead richer lives, we may be able to communicate with them, like that’s a problem we didn’t even know we wanted to solve.” — Josh Tyrangiel, 39:00

yt-dlp --download-sections "*39:00-40:00" "https://www.youtube.com/watch?v=GHQM81L7ALI" --force-keyframes-at-cuts --merge-output-format mp4 -o "GHQM81L7ALI-39m00s.mp4"

”You’ve got a sick houseplant? Take a picture”

“‘Alright, you have a houseplant? Everybody’s got a sick or dying houseplant. Just go take a picture of it. Go to GPT or Claude or Gemini and say, “What is this and what should I do?”’” — Josh Tyrangiel, 48:54

yt-dlp --download-sections "*48:54-49:50" "https://www.youtube.com/watch?v=GHQM81L7ALI" --force-keyframes-at-cuts --merge-output-format mp4 -o "GHQM81L7ALI-48m54s.mp4"

Key Points

• The book is about people, not tech companies (2:42) - A counter to Karen Hao’s Empire; focused on the underdogs downstream trying to use AI for COVID response, disability communication, and other neglected use cases
• The Levine provocation (4:07) - Paul reads NYC Comptroller Mark Levine’s boilerplate AI paragraph to ask how genuine civic good actually gets started versus “big AI thoughts and not enough action”
• Progress depends on stubborn humans (6:32) - Success comes from the handful willing to be “the jerk who says no” to both colleagues and software vendors
• Most subjects are not technologists (7:38) - They often had no idea what AI did; they just knew they had a data problem
• Danny Hillis’s advice (8:13) - “Stop talking to the tech companies. Go find the tech in the world without the tech companies” — the origin of the book’s approach
• No revolution overnight (12:13) - The Cleveland Clinic, one of the best hospitals in the world, grinds out a 2.1% profit; change is evolutionary, not instant
• AI scribe software (13:31) - Phone app records the visit, fills electronic health records, sends labs, translates; saves hours a week, but half the volunteering doctors never used it
• Sepsis prediction (15:02) - Sepsis kills 350,000 Americans a year; Bayesian software flags patients red/yellow/green without making the final call
• 41% mortality reduction (16:33) - Over a year, doctors and technologists integrated the software into the workflow, cutting in-hospital sepsis mortality by 41%
• Some people never change (18:00) - Cardiologists shown a faster, better technique often refuse; realism about how fast the world will change
• Doctors have discipline control (19:05) - Paul’s hypothesis: regulated professions can refuse policy, so AI can’t be thrust on them — unlike do-gooder orgs that get scared off by bad news
• Don’t hate the game (21:30) - On Palantir: hate the players if you must, but the game is good; Ukraine has “no war without iPads and Palantir”
• Hospitals are hotels with no data (22:22) - Unlike hotels, hospitals don’t know when patients arrive or leave; Palantir built a working patient-flow prototype in three days
• The labs aren’t thinking about good (25:03) - Researchers are a different intelligence in a competitive bubble; deploying AT for societal benefit will be on the rest of us
• Nick Clegg on democracies (27:21) - AI exposes a weak spot for large democracies “incapable of having mature conversations”; the US sits uncomfortably in the middle
• Optimism from anger (31:03) - Booing Eric Schmidt is a sign people are finally active, not passive, about the technology being handed to them
• Quiet regulation returns (33:00) - After AI czar David Sacks left in March, the Trump administration signaled it might start regulating models over a certain size
• Christie Jungst’s story (35:43) - A physicist who left her field to build communication tools for nonverbal children after her son’s diagnosis
• Synthesizing sound data (39:00) - A library of 7,000 sounds became 80,000 via AI synthesis; drafting off zero-shot translation research; $79M in funding from Yale and others
• Journalism has a moat (43:11) - AI replicates the past, but good journalists tell people new things — a natural defense if handled carefully
• Scribe tools aren’t commodities (45:14) - Of five AI scribe tools, three required too much editing; one had two-minute review time and high accuracy
• Teachers, not students (51:42) - The education leverage is with teachers who turn lessons into labs; one 30-year veteran rebuilt her plan with a Khan Academy bot
• The middle is imaginary (52:42) - AI lets teachers escape “teaching to the middle” and personalize instruction across a class of 25

Mentions

Companies

• Cleveland Clinic (12:13) - Central case study; nonprofit hospital running AI scribe and sepsis pilots at ~2.1% profit
• Palantir (20:32) - Peter Thiel’s company; controversial but built the Cleveland Clinic patient-flow prototype and supports Ukraine’s military
• OpenAI (25:03) - Referenced re: Greg Brockman and Sam Altman; and later solving Erdős/big math problems
• Facebook / Meta (27:00) - Where Nick Clegg became “concierge” of global relations
• Google (39:00) - Strong at zero-shot translation, which Jungst’s project drafts off of
• Uber (1:34) - Referenced re: Josh’s post-Businessweek television venture and “market is family feud” lesson
• The Atlantic (2:00) - Where Josh is now a staff writer covering AI
• The Washington Post (8:13) - Called Josh to be an AI columnist when GPT came out; where the book idea began
• Khan Academy (52:14) - A teacher rebuilt her entire lesson plan using its ChatGPT-made bot
• United Healthcare (33:12) - CEO killing brought up unprompted by Fortune 100 CEOs Josh interviewed

Products & Technologies

• AI scribe software (13:31) - Phone app that records visits and fills EHRs; five tools piloted at the Cleveland Clinic
• Bayesian sepsis software (15:02) - Prediction/detection tool that ingests patient data and flags red/yellow/green
• Electronic health records (13:31) - Introduced ~1999; turned doctors into grumpy clerks
• Zero-shot translation (39:00) - Translating based on sound waves rather than word-to-word; leveraged for nonverbal communication
• ChatGPT / Claude / Gemini (48:54) - Recommended for the “photograph your houseplant” onboarding trick
• Affective computing (35:43) - Rosalind Picard’s field of emotion-sensing computing at the MIT Media Lab

People

• Josh Tyrangiel (0:00) - Author of AI for Good; Atlantic staff writer, former Businessweek editor and Vice News Tonight creator
• Paul Ford (0:00) - Aboard president and host; former Businessweek writer under Josh
• Rich Ziade (3:45) - Aboard CEO and co-host
• Mark Levine (3:59) - NYC Comptroller whose AI fiscal-future paper opens the conversation
• Danny Hillis (9:00) - Credited as creator of cloud computing; advised Josh to leave the tech companies behind
• Eric Boose (13:31) - Cleveland Clinic general practitioner who led the scribe pilot
• Christie Jungst / Christy Johnson (35:43) - Physicist building AI communication tools for nonverbal children; son Felix, 15
• Rosalind Picard (35:43) - Runs affective computing at MIT Media Lab; connected Josh to Jungst
• Sam Altman & Dario Amodei (6:32) - Lab leaders cited as the “top level” of AI
• Nick Clegg (27:00) - Former UK Deputy PM turned Facebook global-relations chief
• Eric Schmidt (31:03) - Booed at a graduation speech about AI
• David Sacks (32:19) - Trump administration AI/crypto czar with 300+ AI investments; left in March
• Gina Raimondo (26:09) - Former Commerce Secretary putting together a “pretty good” AI plan
• Karen Hao (2:42) - Author of Empire, framed as the counterpoint to Josh’s book
• Steve Jobs (47:33) - Cited as the cultural translator of technology we now lack
• Rex (41:22) - Aboard blog editor, first audience questioner
• Clay (50:18) - CTO of NYU, asked the education question

Surprising Quotes

“Stop talking to the tech companies. Like, go find the tech in the world without the tech companies and you will find people doing interesting things with it.” — Josh Tyrangiel (recalling Danny Hillis), 9:00

“The Ukrainian people are incredibly brave. There’s no war without iPads and Palantir.” — Josh Tyrangiel (quoting a DOD source), 21:30

“The last thing they’re thinking about is what good should this do? And so that’s my sort of urgent call is like, well, we better figure it out, because it’s going to be on us to deploy it in ways that actually benefit our society, because they are just not thinking about it.” — Josh Tyrangiel, 25:03

“Data centers are like 1/15th as bad for the world right now as golf courses in terms of water usage. And that’s a fun little stat.” — Paul Ford, 34:39

“If they’re interested in the credential, there’s never been a better time to cheat.” — Josh Tyrangiel, 51:26

Transcript

Paul Ford: 0:00 So I’m going to kick us off. I’ll intro Josh and then Rich and I are going to ask him some questions and he’s going to answer them and then we’ll let everybody else ask a couple of questions and then before Josh gets to ask too many, we’ll cut you off. So, Josh, first I want to congratulate you on your timing because everyone is excited about a narrative of AI being really great for the world and booing people at speeches, it’s really just great. So you did it.

Josh Tyrangiel: 0:30 I nailed it.

Paul Ford: 0:31 You nailed it. And so this is actually part of the series. The series, this is AI for good. The next will be deforestation for good, extremely high electricity prices for good, your stepdad thinks he’s talking to you via ChatGPT for good, losing your job for good and lying for good. So those are all coming here. We actually do a lot of it in here, so but anyway, but they had nice civic themes. So most of you know Josh, I will attempt to capture just a little bit to set us up. He’s had a truly storied career in media, but his roots are in journalism and it’s like the really kind of sweaty baseball cap wearing listen to people complain for hours kind of journalism. And he was writing at Time magazine in 1999, which was literally a millennium ago, and steadily rose through the ranks to become the editor of Businessweek, winning tons of awards, mostly thanks to me. I wrote for him. And then he left to do something with Uber television, which I don’t think you’ve ever properly explained.

Josh Tyrangiel: 1:34 Oh, I explained it. He just didn’t understand it.

Paul Ford: 1:36 There it is. But then they let you actually enact what you were up to at Vice News Tonight and Vice HBO, which absolutely ruled. I don’t know if everyone’s seen it and won tons of awards and was kind of the best thing to happen in news in forever and which, you know, and then the obvious happened because you can’t have good things in this world. And then you became a columnist covering AI at The Washington Post. That’s where this book idea began, but you escaped just in time and are a staff writer at The Atlantic where you keep writing cover stories about AI. And if you want more details, I would advise you to check out joshtyrangiel.com, one of the last websites on the internet. Anyway, I love Josh, we love Josh. He genuinely helped me build my career in about 5,000 different ways. I would not be sitting here in this nice office with Rich and a business without him, frankly, and so I’m very grateful for that. And he’s an advisor to us and to our company and thank you for giving me Paul.

Josh Tyrangiel: 2:40 Truly my pleasure.

Paul Ford: 2:42 Yeah, it’s truly my pleasure. I don’t know if that was really felt when it happened, but so. But Josh has written a really good book about AI that’s for a very specific reason, it’s because it doesn’t have much about AI companies in it. It’s very much about the people using it and trying to do good. And in a funny way, it’s a counter to Karen Hao’s Empire… …which is a great book about how those companies came together and some of the ways that they are actually quite destructive. Josh is about the people—you’re about the people who are downstream, kind of the underdogs, and who are trying to use technology to enable something that previously has been really impossible to enable, response to COVID or communication for the disabled, or sort of all these sort of use cases that get kind of tossed away by society or downplayed. AI has been a really helpful mechanism. It’s good to get that story because it’s a real story and it’s a powerful story and, I mean, also I forgot to mention you’re from Baltimore and that’s I think very important.

Rich Ziade: 3:45 Yeah.

Paul Ford: 3:47 Yeah, so that’s good. And you love the Orioles?

Josh Tyrangiel: 3:51 I do.

Paul Ford: 3:52 Yeah.

Josh Tyrangiel: 3:52 But they don’t love me.

Paul Ford: 3:54 So, it’s a really great book and everyone should read it if you can score a copy. But we’re gonna open with a provocation because a couple hours ago…

Rich Ziade: 3:59 …Mark Levine, the comptroller of New York City, issued…

Paul Ford: 4:02 …hold on, I’m gonna wave it. Things are gonna get really hot in here.

Rich Ziade: 4:05 Get the rest from Mark Levine, the comptroller.

Paul Ford: 4:07 This paper on AI in New York City’s fiscal future, which I’m now gonna read in its entirety just to frame things. But it’s kind of what you’d expect, which is we’re either screwed or not, maybe a world of incredible bounty is opening to us. But I’ll read you just this one paragraph and ask you a question. So, Mark Levine, male speaking, I can’t do, I can’t do this voice, but, ‘That preparation for AI must begin now. New York City needs stronger fiscal reserves to withstand potential economic shocks. We need to modernize…’ …oh, it turns out we’re like low on money by the way, that’s part of this… …we need to modernize city government so agencies can deliver faster, more reliable, and more accessible services in an AI age. We need procurement and technology systems that allow government to innovate responsibly while protecting privacy, security, and fairness. We must understand which workers and industries are most exposed to disruption.’ And I’m gonna pause there because this is just a paragraph, like they just wrote it about AI and I’ve read this paragraph a thousand times. And I kind of just want you to talk about when you go and talk to people who are trying to do genuine civic good with technology, this paragraph won’t really do anything, right? How do things happen? How do you get an idea like, ‘Hey, we need to get AI into the schools,’ we need to get AI to help people who are disabled, or not just AI, just some, we need to help them and this is one mechanism. How does that get started? Because I think there’s just a lot of big AI thoughts out there and not enough action.

Josh Tyrangiel: 5:51 Yeah, I mean, I, one, thank you guys so much for hosting and for the lovely intro. Keeping my Baltimore roots. Yeah. So this is sort of the question not just about AI but just about civilization, which people have some doubts about these days. Um, how does shit happen that gets better, right? And it turns out when you’re in a place like government or a large healthcare operation or any of these places, they’re populated with some version of the same people you see at work every day, right? And everybody brings their whole self to work and sometimes you wish they wouldn’t bring all of it to work, but they’re all there and they all are trying to solve a problem. And it turns out that there’s a handful of people who actually are like, I’m not going to try, I’m just going to do it and I’m willing to make mistakes and I’m willing to be the jerk who says no, you have to change, and I’m willing to be the same jerk who says to the software provider, this doesn’t work, you need it to work the way we want it to work. And it’s crazy the more I dive into AI both at the very top level which is like, you know, the labs, the Sam Altmans and the Darios and those guys, and then at this sort of a ground level, success just depends on human beings, right? Like the technology’s really impressive, it’s really cool, it’s moving faster and faster. Um, I do not yet think it is beyond our control. I think it’s pretty convenient for lots of people in lots of places to say it’s beyond our control, sometimes for their own self-interest, sometimes not. But what I keep finding everywhere is like, yeah, you know, there’s a handful of pretty decent people who want to solve problems. Most of the people in the book are not technologists. In fact, they often had no idea what AI really did or does. They knew that it was good with data and they knew they might have a data problem. And so they’re like, all right, what do I have to learn? What do I have to do? And so one of the joys of writing this is spending time with those kinds of people who really like care about the world, are often, you know, interesting, dorky, but super sincere people. Um, and they’re so stubborn. And they’re so willing to fight and you’d say, oh, I love this. So like that was, that was the answer. It’s like you gotta find those people.

Paul Ford: 8:07 How did you as a journalist find them?

Josh Tyrangiel: 8:13 Um, well I’ll tell you a little bit about the origin story of this which is that basically Washington Post, um, when GPT came out they called me and were like, hey, we don’t know what’s going on, we could really use a columnist to explain it. And we had a little back and forth about like, well what is a columnist in 2023 at the end of the, wait, whatever, we got there. And so I went out and I started talking to people and basically every conversation was about money or how much the person running one lab hated the other person running another lab. And then I would sort of listen to that and it’s super, by the way, that’s pretty fun, right? When you’re writing and people hate each other and they’re willing to spend a couple billion dollars to prove it, you’re like, tell me more! Why do you hate that guy so much? Um, but at a certain point I would ask like… So what… I was just gonna do? And the answer like… you guys all are… I mean I’m just gonna take a guess based on the faces in this room that you’re all familiar with the show Silicon Valley? Oh. I mean it was basically season one where like, we’re here to change the world for the better. It’s like, oh okay. We’re gonna cure cancer. Oh cool. And then you talk to the other guys and they’d be like, this is the ad QA system. It’s like, this can’t be right. And so I talked to this guy named Danny Hillis who I’m also guessing some of the people in this room know as the creator of cloud computing, and he’s in his late 70s now and he’s just the wisest, sweetest guy and he was like, you just need to stop talking to the tech companies. Like, go find the tech in the world without the tech companies and you will find people doing interesting things with it. And, you know, I’m embarrassed to admit it hadn’t occurred to me at that point that one could actually do that. And as soon as I did, I started tripping over people. And so one secret of, you know, reporting in 2026 as some of you know is like, YouTube is awesome. There is not a single person I’ve talked to cold anymore. I generally know a fair bit of background because everybody talks everywhere all the time. And between podcasts and YouTube, you can listen. And with, you know, all of the AI tools, if I don’t want to listen I could read it. And what I found was like little green shoots of people trying things. And so I would go and pour water on them and feed them and like, yeah, sometimes you strike out, but a lot of the times I just found really cool stuff.

Rich Ziade: 10:42 I’ll work with you though. Oh go. Okay. Wait, I’m on the wrong page. Actually, I mean related to that, using me for that. Alright so list the podcasts… sorry, just got lost.

Paul Ford: 10:57 I can jump in here Paul, the monster of this is… I’m gonna talk about us for a second, we go into large buildings to try to solve problems. And it’s… you could have the shiniest, most glistening piece of technology that just glows in your pocket and people show up. And as I was reading your book, as I’m reading the book… I was like, there’s not a lot of tech in here. It’s all people. And it’s… some people are terrible, some people are inspiring, but it’s all people. And tell me a little… and just to turn it into a question… there are no home runs in your book.

Josh Tyrangiel: 11:49 No.

Paul Ford: 11:50 Some are like, oh I guess that kind of went okay. And they’re all… they’re pretty much universally inspiring in the way they’re committed to even the little little patch that they’re going to focus on. Talk a little bit about…

Rich Ziade: 12:00 The work because everybody’s like, ‘I- I- look what I did in a minute, in a day, in three days. I recoded this thing in a minute.’ And then your book is a- it’s a lot of grinding.

Josh Tyrangiel: 12:13 Yeah, well, I mean, we’re all conditioned to expect the revolution just to arrive and everything’s different overnight. And this is not actually the way anything works. Because as you say, there are stubborn human beings along the way and the systems that we’ve made over the years. So, you know, I went to the Cleveland Clinic, I spent a lot of time there. And it is like one of the greatest healthcare institutions in the world. And it’s so great that every year they grind out a 2.1% profit.

Paul Ford: 12:43 So they’re in a terrible business. Like healthcare is terrible.

Josh Tyrangiel: 12:46 They have a really inspiring CEO, um, and you know, I was like, ‘So you got a bunch of AI pilots going?’ He was like, ‘Well, we have a bunch of pilots. And we don’t have AI people here. We have doctors here. And so if you want to work with us, first of all, we have this position that we’re not trying to become the Lightsaber of health. We know who we are and we have the belief that we need doctors to run our technology in partnership with technologists. So why don’t you just wander around?’ Which was like great. And what I found was that there- there are some small singles. So just a simple example’s like, you know…

Rich Ziade: 13:29 Scribe software, right?

Josh Tyrangiel: 13:31 Which is basically an app on your phone. You go in, you visit your doctor. And what used to be a process of them sort of like, at least in my case, like if my doctor knew my name, I’d be shocked. But basically they would just examine you and they’d grunt a little bit and they’re all typing into their electronic health records, right? Which which came about in 1999 and basically that’s what they do. They’re clerks. And they fill out paperwork and they’re as grumpy as clerks. And there’s a reason, right? There’s a reason there’s all this attrition. So at the clinic, they found this guy, Eric Boose, and he’s just the loveliest Midwestern fellow and he likes tech, but he’s a- he’s just a general practitioner. And they came to him and said, ‘Would you lead the pilot?’ He was like, ‘Absolutely.’ So they tried five different scribe software and what it does is the doctor puts their phone down, presses a button, it records the conversation with your permission. And when you’re done, it’s transcribed it, filled out all the electronic health records, sent all the labs in, it can translate this recording into whatever language you want. And they had five different pilots they were running with by- I think it was 500 doctors who had volunteered, right? So good news, dear Lord, for the people who use it, it really works. It saves them several hours a week of the worst kind of work. Um, patient satisfaction’s through the roof. Bad news, half the doctors who subscribed to it never used it. And he would chase after them and be like, ‘Hey, we’re running the pilot. We need you to use it.’ And they would just be like, ‘Yeah, I forgot.’

Rich Ziade: 15:00 No, I don’t want to do that.

Josh Tyrangiel: 15:02 And so that’s like the push and pull at a very small level. And then at a much bigger level, I mean, I… I bet everybody here probably knows what sepsis is, but it kills 350,000 Americans every year. More than prostate cancer, breast cancer, opioid addiction. And it’s basically your body reacting to infection in such an overreaction that it causes massive inflammation. And so the Cleveland Clinic, which is, again, the best, loses about 4,000 patients a year to sepsis. And so they… after COVID had calmed a little bit, they were like, “This is ridiculous. Like, we… we’re right at the standard for all these other hospitals, but we’re the best. What do we have to solve?” Again, they do two things. The first is they talk to all their people and they say, “Hey, we’re going to do something about this. And we need you guys to be alert.” And the second thing they do is go get some pre-pioneering software by a company called Bayesian that’s just sepsis prediction and detection software. And it’s pretty simple. It takes all the data from a patient, which can be their electronic health record, it can be their pulsometer, their oxometer, everything, any piece of data they can get. And all it does is alert the doctor, is this a red, green, or yellow, right? Is it one, two, or three?

Paul Ford: 16:27 It’s not making the final call.

Josh Tyrangiel: 16:29 It’s not making the final… it just raises the flag.

Paul Ford: 16:31 And that’s… and that’s the thing, it’s…

Josh Tyrangiel: 16:33 It’s the Cleveland Clinic, doctors run everything. And initially, they came in and said, “No, this is gonna… this is gonna be the protocol, it’s gonna beep, and then we need whether it’s an ICU nurse or someone to respond.” And the doctors are like, “No, it’s not gonna work that way, because you don’t know what sepsis is. You know what sepsis looks like in an algorithm, but we know that sepsis looks like X, Y, and Z.” And so you have these two people, both trying to solve the same problem from completely different places. Over the course of the year, they figured out how to integrate the software into the workflow. These are boring words, but these are words that ended up reducing the mortality in the hospital by 41%. So that’s a thousand people who are alive, more than that, because they put together a really decent AI bridge and software and people used it. And they didn’t use it grudgingly. And you’re like, oh, right, so that’s not revolutionary in the sense that today we just printed out a, you know, cure for cancer, but that is a lot of people who are alive because some people did some work.

Paul Ford: 17:35 And that wasn’t 100 percent… I mean, sure some doctors still didn’t use it. It wasn’t fully…

Josh Tyrangiel: 17:40 Some doctors still don’t use it. Some people hear the beeps. And… and there’s like throughout the clinic and… and so what I write about too, like, you just aren’t going to change some people even if you have a better result, right? And so in the cardiology world in particular, um, I learned a lot about doctors and… and man, people really hate their cardiologists. But they are just… they just don’t change. Frozen in a very particular kind of behavior, and you can show them a better, faster technique, and they just won’t do it because they’re used to doing things a very particular way. And by the way, as I’m sitting there thinking and hearing this, I’m like, well, I kind of do the same thing, and everybody does. And so it gives me hope that we have these people out there who are willing to actually be the battering ram to make progress, but it also gives me some realism about how fast the world’s going to change.

Rich Ziade: 18:30 What is it? I actually have a very good relationship with my GP because he would, when he found out I worked in technology, he just started complaining about the healthcare system every time I saw him. And so like over the years, we’ve actually gotten to know each other to the point that now I go in and he says things like, ‘Man, after we go legal high, I just went all in because this job’s so stressful.’

Paul Ford: 18:54 Wow.

Josh Tyrangiel: 18:56 Which you don’t really want to hear from your GP.

Paul Ford: 18:58 You kind of sound like you should be charging.

Rich Ziade: 19:00 I know a lot about… one of his kids is just kind of a knucklehead.

Paul Ford: 19:05 Anyway, onward. What’s interesting, okay, we did a healthcare event not too long ago here. And doctors in general, compared to many other kind of do-gooder communities, are relatively tech-friendly, in that they like the scribe when they use it and so on, and they’re not scared of AI. My hypothesis there is that doctors ultimately have a lot of control over their own discipline. They literally can’t screw up or they get thrown out. There’s regulations. And if they don’t like a policy, they have a lot of mechanisms for saying, ‘No, I’ll not enact that policy.’ So AI can be handed to them. Maybe they could use too much of it, but ultimately it can’t be thrust at them without them saying, ‘No, I’ll do it this way.’ And I think that’s also a little bit true with lawyers, it’s true with other very regulated spaces. But as you get wider and wider… and when I’m talking from experience, because what happens is do-gooder orgs come to us and they’re very curious about AI because they want the money. They want the save because they can push that money. Software is so expensive and they’re often spending hundreds of thousands of dollars on things like Salesforce for casework, for constituent management and stuff, and it’s a bad use of that money and they know it. But then they actually get scared because the news about AI is so bad and the internal team doesn’t want to do the pilot and they don’t want to mess with this stuff. And I’m finding that that is something that’s very, very tricky to triangulate because… and so I’m curious what you think about that because I do think like we have some pretty bad actors in this space and we have people where like I think ideologically most people who might work at a do-gooder or civic org are just going to be very opposed. Like I know Greg Brockman at OpenAI or Sam Altman or you have a lot of Palantir in the book and I have a lot of big non-profit people love Palantir because it really helps them but most people do not. Most people are like ‘that guy’s wacky and I don’t want him anywhere near me’. Just be devil’s advocate for a minute. Make a case for why a do-gooder org should let this in.

Josh Tyrangiel: 21:00 Um, yeah, I mean, it to reduce it to the simplest terms, because this shit mostly works, right? And not that much stuff in the world does. And so the ability to get something that works and is really efficient and actually can improve your product and your ability to care for people, um, it doesn’t come along all that often. And so I’m sure some of you remember the phrase, you know, ‘don’t hate the player, hate the game’. We should reverse it.

Rich Ziade: 21:27 Definitely.

Josh Tyrangiel: 21:30 Definitely hate the players. Don’t hate the game. The game is good. Like, you know, Palantir, I I totally understand why people are like, ‘oh, I would prefer not to support the company that Peter Thiel co-founded with the CIA.’ Right? Yeah. And then you go in and see what they do, and it’s like, well, on the one hand, you know, a DOD person basically said to me, look, the Ukrainian people are incredibly brave. There’s no war without iPads and Palantir. This is not fair. And so recognize that as incredibly powerful. On the other hand, they do a lot of that, but they also are at the Cleveland Clinic and they ended up doing just very quickly kind of God’s work. So I mentioned they make the Cleveland Clinic, which is a nonprofit, basically makes no money.

Paul Ford: 22:22 Part of the problem with hospitals is that they are basically hotels. Except-

Josh Tyrangiel: 22:26 -because they have patients, they have rooms, tons of staff, food, linens, all of that, right? Hotels know when people are coming and when they’re leaving. And it turns out that’s a really important part to making money. And hospitals have none of that insight. And so basically they the CEO realized, this is a problem. It’s a problem not only because it’s keeping us from paying people out quickly and we’re hemorrhaging cash and we’re a nonprofit, but it’s also a terrible service problem. Because if you go to the ER, you wait endlessly because nobody knows when anything is becoming available. So they talked to Palantir, Palantir sent over two people in their mid-20s, listened to the problem, built a working prototype in three days in this woman’s office. It’s run by a hospitalist, who is a very particular kind of doctor. And they piloted it and over the course of a few months, they basically were able to increase all of the transfer traffic rapidly. And they reduced the ER times by many minutes.

Paul Ford: 23:32 So, and if you’ve seen the Pitt, like, that’s a different show, there they’re not punching people. They get out of there a little bit faster and like, this stuff works, right? And so at some point we have to come to grips that, um-

Josh Tyrangiel: 23:46 -and I’m gonna I’m gonna get into some third rail territory because I think it’s important, right? We’re all having these negative reactions to the people running the labs for good reasons. They’re billionaires and they basically want to be trillionaires. They all So, and I want to say this—and I feel comfortable saying this in this room—they’re very much alike, and they’re very different than a lot of us. And I’m not trying to other them, I’m trying to make a point about the fact that the very unique type of intelligence that it takes to be an AI researcher and an AI leader maps to a different kind of deficiency when it comes to explaining what you do, understanding empathy. And so we, if we decide that those two things can never meet and we’re not going to meet it halfway, well then we’re going to get some terrifying AI. If instead we realize, ‘Oh, this is very useful if it’s controlled properly and we get what we want out of it,’ then we’re going to do just fine. But you can’t just other them and say they’re terrible and not recognize that the tech they’re making is actually really impressive. It is odd that their marketing messages are, ‘We’re going to take your job, then kill you.’

Paul Ford: 24:58 It’s not great.

Rich Ziade: 25:01 Yeah, I mean, it hasn’t helped our sales process.

Josh Tyrangiel: 25:03 Yeah, well it’s also wild, as a… you know, this book’s been out for a week and I’ve heard from three or four people at these labs who are like, ‘Wow, I mean, we’re just tracking this and people really seem to not like AI.’ Oh, are you just tracking it? And the answer is—I mean, listen to me, this room full of smart people—we know the answer, right? They’re, one, a very different kind of intelligence; two, they’re in a bubble and competing with each other and the money is enormous and the pressure to be there is enormous. But the last thing they’re thinking about is what good should this do? And so that’s my sort of urgent call is like, well, we better figure it out, because it’s going to be on us to deploy it in ways that actually benefit our society, because they are just not thinking about it.

Paul Ford: 25:58 Let me ask you a question about a through-line. Is there anybody in government who you think is doing a good job of thinking about this? I mean, you’ve been around DC—is there anyone…

Rich Ziade: 26:08 Me!

Josh Tyrangiel: 26:09 Yeah, obviously Mark Vid is number one with a bullet. He is controlling us right now. He’s got a printout right there. I don’t know, I mean, I wish I had really great definitive answers. I don’t. When I talk—I did a cover for the Atlantic about AI, the future of employment and basically I was like, ‘So, Washington, what you thinking?’ And there was a lot of like, ‘About what?’ Okay. Gina Raimondo, who used to be the Commerce Secretary, really smart, talks to everybody, is putting something together that’s supposed to be pretty good—that’s one. There’s a lot of like 700-page books coming out of Europe, like nice binders.

Paul Ford: 26:58 Yeah, and by the way, like, God bless Europe, right?

Josh Tyrangiel: 27:00 of covers like you know there’s a guy named Nick Clegg who who’s really talk about bridging two worlds. So Nick Clegg was the Deputy Prime Minister of England not 10 years ago in the Lib Dems and then he became like concierge of Facebook right? And I was just like one dude global relations… Yeah, and by the way, like one of the things…

Paul Ford: 27:19 Global relations.

Josh Tyrangiel: 27:21 Yes. And so he’s since stepped down, which has freed him a lot. But he, you know, we were talking about AI and governments and he’s like look this really exposes a weak spot uh for large democracies that are incapable of having mature conversations. Does that sound familiar to anybody? Um but it’s really it’s actually a competitive advantage for two kinds of nation states. One, the really small adorable European nation states that are able to have conversations and that they’re going to appoint a great finance minister who’s going to write a report and then everybody’s going to just do it right? And they’re going to smile and they’ll they’ll lead the happiness rankings for another century. And then the other group of countries that’ll be well off are countries that have no conversations and implement the policy on top of their people. And so we do we if we want to get real like we have to recognize we’re in the middle and that’s not great. I mean China actually has this very effective deep fakes policy like the Chinese internet policies or the AI policies are so fascinating because it’s truly from above and very effective and very critical of a lot of things and we’re critical of but they’re just like no we’re not going to have this nonsense.

Rich Ziade: 28:31 So just for for future purposes when recommending a lord as ball, reminder, he loves the Chinese internet. And today’s catering sponsored by Power Tea. So…

Josh Tyrangiel: 28:45 It’s early. It’s real early with this stuff.

Paul Ford: 28:48 Oof, yeah. But what I appreciated about the book is you didn’t everybody make so many… people have been vomiting predictions since the beginning of this thing, right? ‘We’re about to be here and just pack it up,’ right? It’s very dramatic sweeping statements about what this stuff’s gonna do. And you went out there and the stories are humbling. They’re not the most ambitious change the world stories, are they? Or…?

Josh Tyrangiel: 29:30 I think that what we’re seeing is these are evolutionary tactics that will lead if we’re serious about them to revolutionary progress. But it’s a little bit like you kind of get the engine going and you figure out what we actually want it to do.

Paul Ford: 29:40 So let’s have an event in six years. Right?

Rich Ziade: 29:43 That’s optimistic. And you wrote a bestseller called AI for Bad, are you Josh Tyrangiel?

Josh Tyrangiel: 29:47 Yes, unfortunately grandma, I wrote that book.

Rich Ziade: 29:51 It’s a long ways down.

Paul Ford: 29:53 Okay, yeah.

Rich Ziade: 29:55 Let me frame the question less cryptically. When transformational tech lands, humans are…

Josh Tyrangiel: 30:00 …really bad at wrapping their heads around it. Tends to get out of our, slip through our fingers, kind of be a runaway train and then we got to clean it up, right? This is, this is I was one of my questions here is did you write this too soon? Meaning these are vignettes and positive stories, but if this is a race and there are no leveling factors in the world, right? There are people who are trying to do good things and then there’s the power of this technology and what people tend to do and it happened with social media and now we’re kind of down the path and trying to put it back, but it’s too late, etc, etc. As you look ahead after you’ve been on the ground are you feeling like, oh, this could end up in the same place or do you feel like, are you optimistic? Like you took an optimistic swing here, right? And how do you as you look ahead, how do you think about it?

Paul Ford: 31:03 So, look, I’m optimistic for a couple reasons, one of which is just I am generally optimistic and I believe that people can be led to the good things despite the current environment. I really do believe it. The other reason I’m optimistic is because people are so mad. And so if you what I mean by that is like if you lived through the last 25 years of technology in America and weren’t mad, I’d be really fucking worried about you. Just be like what have you not been paying attention to any of this? And so when Eric Schmidt talks about AI and gets booed at a graduation speech, I’m like, yeah, where is this… oh, my bad. Right?

Josh Tyrangiel: 31:44 Did they know what they’re booing?

Paul Ford: 31:45 I actually do think they know what they’re doing.

Rich Ziade: 31:48 What are they doing?

Paul Ford: 31:49 They are booing people like Eric Schmidt telling them you need to change everything about your life plan because me and a couple of guys, we got something cooking. And that’s what they’re booing. Do they care about the parameters in the model? They do not. But to me what was so desperate about the last let’s call it 15 years of both technology and American life was just how passive it was.

Rich Ziade: 32:18 Yeah.

Paul Ford: 32:19 And so passive about the technology, passive about the fact that you’re looking at it, passive about the fact that you kind of suspect that the people giving it to you were not good. Some of them were made into massive films about how not good they were and we still were like… And so I think it’s great that people are angry. And one of the mild pieces of optimism I have is, you know, I had been talking and writing about the hilarity of David Sacks as the AI czar in the Trump administration who is also the crypto czar, um, who also has 300 something plus investments in AI companies, right? So David Sacks left in March…

Josh Tyrangiel: 33:00 And two weeks ago, the Trump administration quietly was like, “Yeah, we might start regulating models over a certain size.”

Rich Ziade: 33:07 And they’re doing that because people are booing speeches and getting really angry.

Josh Tyrangiel: 33:12 And so, for the first time in a long time, I’m a little bit optimistic that there is the proper level of fear from government and from companies. And I will also tell you, without getting too bummed out about this, but like, I talked to a lot of Fortune 100 CEOs for that, you know, story about AI and future employment. Without—I didn’t bring this up, but several of them brought up the killing of the United Healthcare CEO.

Rich Ziade: 33:41 I don’t like that world.

Paul Ford: 33:42 That got dark fast.

Josh Tyrangiel: 33:45 Um, I don’t want that world. They don’t want that world either. But I do think in the last, let’s call it, two or three months, there is this dawning recognition among people of power that you can have all the money in the world, somebody’s still going to get you. And we live in a culture where everybody has a phone and everybody has a gun. That’s not sustainable. And so, I’m—I’m optimistic because of how dark things have gotten. But it’s been dark for a while in this realm. And what I sense is that, oh, okay, people are beginning to reckon with the fact that we have a government that’s not legislating things or regulating things, we have a populace that’s getting very angry, and we have people who realize they have no product without those people.

Rich Ziade: 34:34 So, I don’t know, that’s a really scenic way to saying why I’m optimistic, but I am a little bit.

Paul Ford: 34:39 Exactly. One of the things I think about a lot is, people come to me about this subject a lot and one of the things that comes up a lot is data centers. And data centers are like 1/15th as bad for the world right now as golf courses in terms of water usage. And that’s a fun little stat. I don’t make it—I don’t say that because they’re right to be angry. Like, it’s just like, they’re pissed off and it’s a folk narrative about, you know, they’re putting huge boxes up and either ICE is either going to put people in them or they’re going to put data centers in them and either way is really, really bad. And I think like, this is how we’re—people are finally starting to express that they’re frustrated about a non-power to democracy. And I think that’s good, even though it’s awkward sometimes.

Rich Ziade: 35:20 Let it be known that the author of AI for Good is recommending that everyone get a gun.

Josh Tyrangiel: 35:26 No! That’s not what I said. That is not what I said.

Paul Ford: 35:30 Just check off the list of Fortune 100 CEOs he wants to shoot.

Rich Ziade: 35:33 I was going to ask this question at the last, but I want to ask this one last. I want you to tell the story of Christie Jungst. Yeah, just an awesome story.

Josh Tyrangiel: 35:43 So, um, at the sort of darkest moment of, uh, parasocial relationships with AI, you know, when a bunch of people were going way down the rabbit hole and emerging with like, scary eyes. Um, I was talking to Rosalind Picard, who runs the MIT Media Lab and is sort of the saint of affective computing… of a thing called affective computing and sensors. And so I was like, so what’s um what’s interesting and what how do we combat the human connection problem with AI? She’s like, ‘oh you need to meet Christy.’ So Christy Johnson is a woman who’s a grew up in India and just got in love with physics and is a physicist. Um her husband is also an astrophysicist and was on the team of four that took the first picture of a black hole, right? So these are not dumb people. So Christy um has kid and she’s like, ‘something’s wrong with my son’ and the doctors are kind of gaslighting her. ‘He’s fine, he’s beautiful, a joy.’ And after a year, um they finally get him tested and he has a genetic deficiency and there’s seven other children like him in the world. He has severe autism, uh epilepsy and he will never speak. Like completely nonverbal. He makes vocalizations, but he will never speak. And there are about a million kids in the US who are nonverbal and many many more worldwide. So they grieve over this and cope with it, um they begin to go on with their lives. And they move to Boston and what she decides is, ‘I’m going to study physics? No. I’m going to figure out this problem.’ And so she uh and she’s incredibly smart and ambitious. She shows up at a book event that Rosalind Picard is doing in Cambridge and says, ‘I’m totally qualified. This is the problem I want to solve. I have no advanced degree because I had to drop out of my master’s program. Uh I want to come to the Media Lab.’ And Rosalind Picard accepts her to the Media Lab. And she then spends her entire PhD figuring out what is the map of this problem, right? And she has a very smart intuition which is this could be a data problem, but I have I’m missing two things. One, the kind of algorithmic power to crunch that data problem and two, any data. And so she’s like alright, she’s studied AI enough just enough to know that it’s moving and she thinks, well what if these two problems can converge? I’m going to focus on the data. So she starts putting out the message in the autistic family community, ‘I’m doing this and I’m trying to collect video, sounds, anything that I can turn into data.’ And she’s grinding and then again like it’s back to these people who are just too stubborn to let a problem linger. And she’s grinding and little tiny audio wave sample by little wave sample she begins to collect data from people. And then she begins to figure out how you want to structure that data. What what do we want to get from a person?

Rich Ziade: 38:51 Sometimes it’s hunger data like sounds, just movement?

Paul Ford: 38:54 She starts by getting everything it’s too much.

Josh Tyrangiel: 38:56 Yeah. And it’s really rough on the kids and ultimately what she wants is sound. Sound is is easy, it’s light, it doesn’t take… lot of room in your memory and so she begins to collect sound and then she begins to get some brains and get some grad students and AI comes along and one of the things that she can do with the sounds is synthesize the sounds, turn them into synthetic AI, she can replicate the sounds and so a library of 7,000 sounds becomes a library of 80,000 sounds and gets bigger. And then on the other side of things, you get developments like zero-shot translation, in which you’re not translating language word to word, you’re actually translating it based on sound waves. Now Google, which is really good at zero-shot, is not worried about this problem, but she can draft off of that technological development, scale it, convince grad students to work on the problem and so ultimately where I met with her and hung out with her family, her son Felix now 15 years old, he’s beautiful, he’s bubbly, couldn’t care less about any grown-up nonsense. Has his own world that he lives in, the problem is that if he’s not with his parents or a caregiver, he can’t be anywhere and his parents can’t be anywhere without him and that just limits everybody’s life. And where she has now gotten to is a massive library of sounds of non-verbal people. Um, they are running constant experiments and modeling experiments at Northeastern where she teaches. She got, seventy, since the book was written, she got 70, 79 million dollars in funding from Yale and some other places and it hasn’t been solved, but it might be. And I don’t think it will be that much longer. Again, the tech is moving fast and it’s really good and so living in a world where a million people who were just not connected at all might actually be able to lead richer lives, we may be able to communicate with them, like that’s a problem we didn’t even know we wanted to solve. And so she’s extraordinary, but that’s the kind of stuff that gives me some hope that if we could just get enough access to the good stuff, we’ll be in a better place.

Rich Ziade: 41:08 Awesome. Place here.

Paul Ford: 41:11 Let’s, let’s go for three or four questions and then we can have our toast and we can eat all the food and drink all the drinks. Does anyone have any questions? Rex, your hand went up early. Yeah. Hey Josh. Um, I’m Rex, an editor at the blog. Um, I think it’s the table of contents then is mostly a really heavy subset of science and medicine and I think that’s very valuable for people like me and perhaps other people in the room where you may have a large number of group chats out there with doomers and you’re frequently dropping links to them saying what the future could be for AI. And today people probably saw at OpenAI, OpenAI’s… I solved some big math problems, father.

Rich Ziade: 42:02 I mean, I solved it first, to be honest.

Paul Ford: 42:04 Big talker.

Rich Ziade: 42:06 Page 32, I’m sure.

Paul Ford: 42:07 Yeah. Erdos problem on it, whatever this mathematician here, you can help me with that is. And I dropped that link in my like group chats today. So I look forward to kind of like using this as more ammunition in those scenarios.

Rich Ziade: 42:23 But I think in a certain sense though, the world of science and medicine’s easy in the sense of it’s a group of people who are evidence-based. So they will, doctors will mostly take on things that work. I mean, you shown cases where they won’t, but they’re better than most people at it’s like, if this thing’s going to work and it helps people’s lives, they’ll probably adopt it. And that’s unlike other realms where that’s just not the case. And so my question is kind of, there’s another world you come from, which is media and journalism, very different points of view on whether or not this is good. And I was curious if you have anecdotes in the book about that or just feelings overall on how it affects media culture.

Josh Tyrangiel: 43:11 Yeah, I mean, I didn’t get into media just because I live it and didn’t think it’d be all that interesting to people. But I would say that most journalists are very annoyed. Um, they’re not annoyed about the tools, they’re annoyed about… I mean, media has been a challenge for a while, and now they’re competing against AI sloth. And they’re seeing the consequences of that. I don’t think they’re hard-headed about its potential uses, like there’s good AI data journalism going on. There’s bad AI glaci-ers is going on. Um, but to your point, it’s like, yeah, listen, science, medicine, evidence-based. Um, education and government are not. And so that’s the other places where I really spend some time. Like, um, it’s hard. It’s really hard. But there are heroes in there. And so I, I, I think you’ll find them everywhere. Um, but I, I wanted to go places where you could actually see progress, and journalism for better or worse is just not making a ton of progress with AI. And look, I, I’ve advised some people, journalism’s going to be great for AI. Uh, I mean, largely because it’s new, right? Like AI is trained on information that already exists. It’s really good at replicating things out of the past. If you’re a good journalist, you’re constantly telling people new things they don’t understand. And so there’s a nice moat around it if you’re careful. Not every journalism CEO thinks that way, but like, I think it’s going to be good for us.

Rich Ziade: 44:41 Other questions.

Paul Ford: 44:45 Uh, okay thanks. You were talking about the AI being used for transcription for doctors at the Mayo Clinic or the Cleveland Clinic, whatever it was. And um, that was a plot point on, on the pod that it didn’t work well, right? And I find the same thing when I use AI… I… it saves me a lot of time and it’s very good, but it doesn’t save me as much time as it could because I have to check what I got. So, do you see or what do you see a time where we could just copy that we go ahead with great success?

Josh Tyrangiel: 45:14 Um, I do think what — so the Cleveland Clinic used five different AI scribe tools, right? And their presumption was, oh, well, this is going to be a commoditized product, they’re all basically going to be the same. Turns out they’re not. And so for three of them, the doctor spent way too much time editing and reviewing and they’re like, this isn’t worth it. So there’s a pilot. But they found one where the review time was like two minutes and the accuracy was through the roof and then they tweaked it some more. And so I actually do think we’re — we’re in a rough period where people are going to trust it, but over the next handful of years, this stuff will get good enough that you will be able to rely on it pretty much all the time. Um, I think that’s progress. But it — no denying we’re still going to be in rough patches with certain models, certain kinds of software.

Rich Ziade: 46:13 Back to Kim and Matt right there.

Paul Ford: 46:17 Okay, hi. Um, I’m wondering if you see a way to resolve a couple of the the gulfs that I that I feel like I see in this world. So the stuff that you’re describing in the book seems very sort of like tools for specialists who really know what they’re doing, and that’s where the good stuff is happening. But I imagine most Americans uh don’t see that, they see not just slop but sloppiness, right? Like AI is everywhere, it’s the future, it’s all over, but most people’s interactions with it on a day-to-day basis are not going to be, you know, high-end tools and stuff like that. And that’s replicated as well in the economic world where there’s these, you know, trillion-dollar valuations uh that are perhaps outsized compared to uh, you know, a scribe program at Cleveland Clinic or, you know, these more practical applications. And I’m wondering how like is there a way to get the public’s expectations to the right size? Or, you know, are trillion-dollar valuations and people’s day-to-day interactions with the more quotidian mess of AI actually accurate?

Josh Tyrangiel: 47:33 Well, I mean I think they’re two separate things that have gotten sort of melded into a hairball, and I think it will be fine if we can disentangle it, right? One is it’s America, people just keep score with money, right? And so when you hear trillion, everybody wants to know more. Even those of us who don’t think there should be individual trillionaires, people just go to that, right? Um, we could debate— The valuations. It’s kind of irrelevant to me, like, those of us who worked at Uber business before it broke out came from outside of it. There are a handful of us here. We all learned that basically the market is just family feud, right? Like it’s not what a thing is worth, it’s what 100 people think it’s worth. And so like, I don’t get too, I don’t over-rotate on that thing. I think what you’re getting at is that we’re missing a key connection piece, and I’m going to bring him up even though he’s cliché, but like, you know, Steve Jobs was pretty good at explaining technology to normal people. We actually don’t have that in culture right now. Paul Ford’s pretty good at it.

Paul Ford: 48:38 I’m not…

Josh Tyrangiel: 48:39 No joke. Like Paul describes Aboard…

Rich Ziade: 48:40 Yeah.

Josh Tyrangiel: 48:41 No, there’s just a, we are missing that in culture because government for all sorts of reasons is hostile to explaining complicated things, because people’s attention spans are much shorter. So I’ve been asked by, by, let’s call them normal people, which this room is really not full of normal people, God love us all, but like, ‘Well, what am I supposed to do with it?’ And that is like a real question out there in the world. And what I say is like, ‘Alright, you have a houseplant? Everybody’s got a sick or dying houseplant. Just go take a picture of it. Go to GPT or Claude or Gemini and say, “What is this and what should I do?”’ It gives you an idea of the fact that like, this is really smart, encyclopedic knowledge, powered by tons of processing, and you could control it to do things for you. And for the first time, people are like, ‘It can do that?’ It’s like, ‘Yeah, you know, turns out saying “My AI is multimodal” isn’t very helpful to people. Saying, “Oh, you know what you can do is take pictures of stuff and ask it questions.”’ And members of my family are here. They don’t have a lot of confidence when I take pictures and fix the washing machine, but I did it. And so that to me is kind of what you’re getting at, is what’s missing is like, ‘Yeah, we should maybe do some more of that.’

Rich Ziade: 50:12 Alright, that’s it, let’s go to Clay. Alright, let’s go to Clay and then we got to call it because everybody’s wiggling in their chair too, so it’s time. Clay.

Paul Ford: 50:18 No, just, I was just going to say, I’m… sorry, thank you. I’m in education, where we have both a lot of the worst and best use cases. I know you have written a lot about that. I wonder if you could just say, if you wanted to be an educational institution that was committed to getting on the good side of these tools, what would you advise?

Josh Tyrangiel: 50:32 Um, I think the secret is teachers not students. Um, students, you know…

Rich Ziade: 50:40 So Clay, in addition to also being one of the gifted communicators to technology, is the CTO of NYU, which is a small, independent school, just adorable, just tiny with no footprint, the Estonia of schools. Just to digress for a second, it is the largest private landowner in New York City.

Paul Ford: 51:00 Crack bread.

Rich Ziade: 51:01 Yeah.

Paul Ford: 51:02 Upper. Half the church had to sell.

Rich Ziade: 51:05 Cause it’s because the church had the business unit. Exactly. Well, and NYU is only too happy to stay.

Josh Tyrangiel: 51:10 I think what I saw in a couple schools is kids are going to be kids and they want to educate themselves because they’re not interested merely in credentials, but in identity formation, they’re going to learn. If they’re interested in the credential, there’s never been a better time to cheat. By the way, it can happen in one kid, right? There’s a kid who might really be into math, but if you put them in an English class, they’re like no thanks, write me some Pride and Prejudice essays. So I think the way to attack it is through teachers. And what I saw in the best use cases were teachers who over particularly the last six years realized, I’ve taught through COVID when they were all remote. Now they’re back and I know they have multiple tabs open, I know they’re sneaking looks at their phones, and this traditional model where I hold the knowledge and I speak it to them and then I force them to repeat it back to me either through a quiz or a lecture or whatever, doesn’t work. And so the there’s one teacher in particular who basically took her entire lesson plan—she’s been teaching 30 years—went to this Khan Academy bot that was made with ChatGPT and basically said, I want to make everything a lab. Everything. I want kids moving, I want them participating, I want them talking to each other and I want to scaffold the learning so that at key points they call me so I can see how they’re doing. And what she found was that her job was so much better. The kids enjoyed each other, it was a social room, and and it had no impact on her ability to test who was doing what. The biggest thing, and this is where I’m most excited in education, is like every teacher will tell you they have to teach to the middle, right? Which is pretty obvious. Like if you’ve got 25 kids in a class, you can’t teach only the smart ones, can’t teach only the dumb ones. But there’s nobody in the middle. The middle is an imaginary space. And so what this one teacher was telling me was, no, now I actually get diversification in the way I teach. I can group people. I can see that these guys are limited and they’re going to learn only so much today. And so getting that blend of personalization and creativity, I was wowed. So I think it’s the teachers who are going to be the ones who really need to take it on.

Rich Ziade: 53:30 Awesome. All right, let’s get a round of applause for Josh. Alright, what’s your next book on that?

Josh Tyrangiel: 53:51 Oh, dear God. I’ll stay on for a bit after the first because I worked in publishing for a bit. For publishing nerds here. After the first week when you’re like, ‘Oh, Jon Stewart and the Morning Show and, you know, a couple extras,’ then you’re like looking at your calendar and it’s like, ‘This can’t be a podcast!’ So I think it’s like the six weeks of like, ‘Wait, is this just—this is just a Zoom. I’m just on a Zoom with one guy. This isn’t even a podcast.’

Paul Ford: 54:23 Speaking of which, you recorded with us at our podcast, the June 24th?

Rich Ziade: 54:28 June 24th.

Josh Tyrangiel: 54:30 Alright. Thank you very much.