Josh Tyrangiel: "AI For Good"

0:00Your SoCal Honda dealers are honored to help the community. Our program started the first deaf and hard of hearing football team of its kind in our league. The helpful SoCal Honda dealers surprised us with brand new uniforms giving our athletes more than just jerseys. A community investing in their program. And we want to help you too with 1.99% APR on a 2026 Ridge line for well-qualified buyers. At your SoCal Honda dealers, made through contract limited bus finance, our SoCal Honda financial services, not all buyers will qualify to take delivery by 1126. Sometimes a little treat is your daily dose of joy. With the new Hershey's Crem bars, you'll discover your new happy place. It's classic Hershey's milk chocolate, layered with rich salted caramel or avocado creme. Fight into a more textured, more flavorful sweet treat worth talking about. Try the new Hershey's salted caramel or avocado creme bars, wherever you shop for snacks. And choose happiness with every bite. Hi, it's Michael Smirconish.

1:00Welcome to Book Club with Michael Smirconish. A collection of Michael's favorite interviews with authors from the last 30 years through today on the air on radio. What sets my book club apart is that I actually read the books. Book Club is now in session. Standby for good news about artificial intelligence. And I mean, lately the news about AI has largely been bad. It's been negative and really in two buckets. The first is the fear that artificial intelligence someday can slip its rails and turn against us. That was the story that was told from the anthropic researcher who quit and warned that the people building this technology, they themselves believe that it could wipe out humanity. And then Dariel Amade andthropic's own CEO said the industry needs to slow down open AI, open AI, then disclosed six incidents in which its systems hid errors invented data or moved files onto the internet without permission.

2:01The second is the one that I delivered a commentary on CNN last Saturday about because I think it's the more immediate threat, which is this that AI doesn't need to turn against humanity. In order to do catastrophic harm, it only needs people to trust it when it's wrong or to use it when their intentions are bad. And there was that CNN report last weekend that in the spring during the war with Iran, a US military intel report claimed that a Chinese ship in the Middle East was carrying nuclear weapon components. Armed US service members were preparing to board that ship planes were in the air only at the last minute that officials discovered that an analyst had used a chatbot to help write the report and the chatbot had gotten the cargo wrong. And then most significant to me, I still say this is the most under reported story. It was an anthropics own report describing in 140 pages how there had already been incidents where artificial intelligence had been used in the hands of people seeking to do bad things in Mali building a surveillance system using 25 million phone numbers.

3:11An actor in Yemen developing missiles operators in China building files on dissidents, including religious leaders. So that's the stage or at least that's the stage up until now and hopefully it changes with what you're about to hear in terms of leveling the playing surface. My next guest has spent the last couple of years going out and finding people who are actually using the technology to fix things things in government medicine classrooms and in ordinary human relationships. Josh Tear and Gal Josh Tear and Gal is the author of AI for good. A book that is describing how real people are using artificial intelligence to fix things that matter. Hey Josh thank you so much for being here I really appreciate your time I love to the book I'm looking to spread the word of it. And I guess it all begins with you being invited by a then Washington post editor to go and do some nerd cultural anthropology what's the origin of your look at artificial intelligence.

4:14Yeah that's right I mean one thank you for having me really appreciate it. I you know I've been a journalist for a long time I oversaw newsroom at Bloomberg and you know I had a working knowledge of artificial intelligence and its earliest phases about language prediction about preference. Prediction about preference prediction all the sort of basic algorithmic stuff that's kind of been with us for 15 years or so when chat GPT sort of exploded onto the scene I got a call from my friend David Shipley who was then the op-ed editor at the Washington Post and he basically described a kind of kind of like from a journalism standpoint which is our readers don't know what this stuff is our reporters are suddenly just inundated with things to report on that they don't understand. It seems like a pretty small world. It seems like everything is about to change simultaneously would you consider being a columnist about AI. Well that's interesting because you know as you know columnists have this like slightly different license than regular journalists to kind of step back ask different kinds of questions.

5:18And so when I ended up doing for a year was writing about 2000 words every two weeks and I started as one does you know as a AI was it's hard to believe this but you know open AI at that point had 920 billion and fewer users than it does today so it was a very small thing I accepted the gig and within three days I was talking to Sam Maltman and open AI headquarters and as one does as one does I mean but it gives you an idea of how fast we're moving like these are not celebrities these were just people running weird like literal labs like these are research labs. And for the first six or eight weeks of the gig I was writing about the things that were very obvious which is you know just what happens and I was writing about money I was writing about Sam and Elon and Dario and how they know each other and all the the fundraising to the point that like I would sit down with people and the very first thing they'd say is all right well we can tell you about our next round but it's off the record and I was like I don't care like the last thing I care about is whether you're going to get rich.

6:27What I care about which nobody's really talking about is like what's this going to do like how is the rival this technology going to change the world for the better because every new every new tech arrives in the context of the last new tech right and the last significant new tech we had as a society with social media which for the first half of it we were like oh this is cool we're connecting this is great and the last half of it we've really been living with the consequences of essentially the same thing. And so I was really pointed with folks like what's it do so I ended up having this conversation with a guy named Danny Hillis and Danny some of your listeners probably know is like one of the great men of computer science of the 20th and 21st century is just a brilliant guy largely invented cloud computing and he's in his late 70s now and I'm zooming with Danny and just kind of venting to Danny about like I don't is this all ridiculous like is any of this worth something and Danny just kind of like I'm not going to be able to do this. And Danny just kind of laughed at me is like you you need to separate the tech from the tech companies because if you can do that you are going to discover that there are people out there like my younger self who really do care about problems and are adapting the technology in ways that you would never guess and all of a sudden I took his advice and like this world opened up to me that I kind of refer to as an AI counter culture it's happening far away from all the headlines it's happening in places where like you just

7:56wouldn't suspect it but hard headed stubborn caring people are using AI to make things better and so that's that's how I got there. Okay, the book is called AI for good and it's really not about the tech Titans right it's it's about these quiet practical showing little interest in vanity individuals who are already using it with a positive outcome. I'd like to run through some of the realms and have you give me the cliff-snote versions to allow listeners to know before they read the book exactly some of these good news stories for example. This is book club with Michael smirconish from serious exam. Hey, it's Dan Cummins. If you're into the weird the wild and the downright bizarre check out my podcast time suck. Each week I dive into shocking stories like the rise of the next year called the origins of conspiracies like QAnon and the San Francisco which killer murders with deep dives and dark humor time suck brings you the stories that'll fascinating make you laugh and fill your head with lots of strange facts new episodes drop every Monday join the cult of the curious follow time suck wherever you get your

9:04podcast. I'm Lovie done all American gymnasts and Vyuri athlete. When you travel and train as much as I do you find happiness where you are. On the mat or on the sand movement and comfort are essential that's why I live in performance joggers by Vyuri made from dream net fabric that's made of 89% recycle materials effortlessly soft and made to move as much as I do. My happiness starts here in the softest joggers on the planet. Get 20% off your first purchase at Vyuri.com slash Libby that's V-U-O-R-I dot com slash L-I-V-V-Y. Exclusions apply not only will you receive 20% off your first purchase but enjoy free shipping on US orders over $75 and free returns. Go to Vyuri.com slash Libby and discover the full versatility of Vyuri clothing. Exclusions apply visit the website for full terms and conditions. Welcome to the newest grocery outlet location your couch no parking lot no shoes required and the commute is approximately zero feet because now you can shop

10:15in the car or door dash and get your grocery outlet favorites delivered right to your door with the same great item prices as in store your couch has never been this productive. Grocery outlet big deals now delivered. I'm pretty confident talking into a mic. Hey, I'm doing it right now but home projects a second guess everything is that noise normal is that water damage. Who do I even call that's where thumbtack comes in just upload a photo or voice note and it uses AI powered search to match you with the right top rated local pro so instead of guessing you get clarity and can hire with confidence for your next home project try thumbtack higher the right pro today. Michael's for honest program listen weekdays at 90 am east on potez serious XM channel 124 and anytime on the serious XM app Josh Tear and go is the author of AI for good. I'd like to run through some of the realms and have you give me the cliffs note versions to allow listeners to know before they read the book exactly some of these good news stories.

11:22For example, let's talk education tell me about Peggy buffington the superintendent in Hobart Indiana and what's going on there. Yeah, so Peggy is what I think I describe in the book is she looks like loretta Lynn but she dresses like Johnny cash and she's a celebrity in the town of Hobart. I mean, she's just a hard ass. Who everybody knows cares more about your kids and anybody else in the city she just does. She also happens to be really into technology she in turn to apple in the 90s she really is integrated tech her whole life and so. This pilot program called con me go which is out of Khan Academy is an AI tutor right and this is very early days in the AI revolution so we're talking like. 2024 she's getting a call from South Khan and he's saying you know Peggy I'd really like you to pilot this very controversial tech and because Peggy really runs the town she's like you got it. I believe our kids need to be experimenting with tech I know ed tech is disappointed before but we're going to do it and so she brings the pilot into this school district and what they discover and and part of the reason that sell went with Peggy is because Peggy does not.

12:36Like hold back and so a couple of weeks and she's like you got problems like the bot can be bullied our kids can bully it into giving them answers. It has all sorts of issues around safety it's promising she's like I see our high flying kids really getting to it but you got issues and so Peggy works with Khan Academy diligently lots of follow up lots of bugs happening which is really like what all this is about is like that. The tech has great capabilities but you have to hammer it into place to the problem you're trying to solve the problem she's trying to solve is getting kids to practice their homework more and getting teachers to be more creative and so they're working together month after month and over the summer so the first pilot is about 12 weeks over the summer she realizes you know what really is going to make this work is teachers. I have to do professional development for teachers I'm going to hand select some teachers who are going to figure out how this can help them in the classroom and not every teacher did it and not every student thrived with it but what I witnessed was a couple teachers who really figured out they had a they had a very specific problem which is it post covid kids just have their their laptops open with multiple tabs there barely paying attention they're sneaking looks at their phone.

13:53And one teacher in particular told me like we have to reinvent education this model whereby I hold the knowledge I give you the knowledge I test you on the knowledge it does not work for the way they live their lives and so she and several others used one of the teaching tools in con me go basically so which by the way con me go is aligned to each states curriculum. And has a master list of how students are doing she basically learned that she could take all of her lectures and turn them into these active teaching sessions so quiz became a lab and she would bump around the room she pair people up she would put into con me go. I need to test people on the principles of style chemistry which is a really interesting mix of chemistry and physics basically and I here's what I have I've ziplock bags I vinegar I baking soda I X Y Z. And she would work with con me go to create this lab and then she would work with con me go to structure and scaffold each lesson so that first she would tell them the instructions the kids would go into con me go when they got to a certain place they would stop call her over to inspect their work.

15:01They could go on kids who are falling behind she could spend more time with and it became this sort of interesting creative teaching assistant and you know listen I don't know that much about style chemistry I know a lot about what joy. Well it was an it was an air bag simulation I was fascinated by the and here's something that I want to stress for people so that they appreciate this this is not the artificial intelligence that you describe in those several chapters loosely tied to the con academy it's not a teacher replacement it's an enhancement of what teachers are able to do. Yeah and there's listen I didn't go into this looking for an angle or a thesis I went into this as a reporter trying to figure it out and what I found across government and health care and education other realms is it it's very rarely replacement technology it's augmentative it works very well when you have a specific problem to solve but what's absolutely indispensable is a hard headed human being who insists on making the tech work for the problem and not vice versa.

16:07Okay, let's go to another of those hard headed individuals you're going to help me with the pronunciation of his last name Thomas at the Cleveland clinic he's probably related to me say it again. Tomas Mahalyevich. Okay, he is the CEO of Cleveland Clinic which has right it's an 80,000 person one of the best health care systems in the world. He's a former cardiac surgeon who really he has come up through some hardship and he got to the top of Cleveland Clinic and basically AI comes about and he had some experience with it and he served they invited me in they're like look we don't have it perfected but we believe we have two problems here. One is that health care needs to be better and the other is it needs to be cheaper right and so there are a lot of companies around the world that you think AI is going to be this silver bullet and doctor. Tomas Mahalyevich sat me down he's like it's not a silver bullet but what it can do is like we have an obligation to improve care and if we can save money we have an obligation to take those savings and plow it right back into improving care.

17:10And so they ran these pilots around this hospital on everything from back end things to actual care and what I discovered is you know I met the CTO who they recruited in from out of a completely different round but he's at the you know late in his career he's a really good CTO. And health care is hard because one you know the paperwork the company like it's very hard to make a dent in the culture of health care and two something that he discovered and I discovered is you know doctors are pretty stubborn there you have to wait a long time to be a doctor you get a very specific kind of temperamental training. And so he was the CTO is looking around it's like I have this mandate to experiment with AI but I don't have the right people to really do it I have maybe eight ten people in this system I need each and every one of them to work and so I'll give you just two examples of things that really made a huge difference right. Number one when it comes to profitability. I'm including clinic runs about two percent margin which sounds terrible but in health care is like pretty great but part of the challenge is that hospitals and hotels are basically the same thing with one difference right there I'll have rooms and beds and food and linens one one one knows who's made a reservation and the other has no idea who's coming through the door.

18:29Exactly right and so with no predictability you have no way to run a hospital particularly well and so they realize this and they went to a bit of a good place. So they're a very controversial company called Palantir and basically said can you build a software that will predict what is going to happen with our bed situation with our transfers. You know the clinic makes a lot of money from people who transfer in from around the world because of the quality of care but it can't help them if they don't know if there's a room and so very swiftly. Palantir comes in and Palantir was subservient to the people running the system and within a couple weeks they were able to stand up a system. Palantir is a very bad example of how to leave them clinic to suddenly predict quite accurately where rooms are going to be how they could get people in and out. The thing that just stunned me is that almost like within weeks emergency room wait times went down 90 minutes and if you've like watch the pit it's a totally different show if you reduce the wait times time.

19:29So that's like one back end thing and then on the other you know sepsis is this is a very quiet thing in the United States but it kills 350,000 people. A year in the US and it's basically the bodies response dysregulated response to an infection right it can happen inside the hospital outside the hospital. The challenge is that it looks like almost everything else it has no unique signal until it is advanced to the point that you can't stop it and so Cleveland clinic in 2021 realized they lost about three 4000 people a year. Two sepsis which is totally standard for the industry but this is one of the best hospital systems in the world and they decided this cannot this cannot be anymore so they did a variety of things one of which was bringing sepsis prediction software which had been developed by a young woman who's at Hopkins and who had a nephew who got a sepsis. And it's really it's a very simple system it doesn't tell the doctors anything but as these three flags basically a green light red light yellow light right. Green light you don't have to worry about sepsis in this patient and the way it works is it's hooked up to all the biometric information coming through it's scanning tons and tons of noise looking for signal yellow light maybe you need to check on it red light we have a problem right so they implement the system and across the hospital and particularly in the ICU where people are giving off signal all the time it's kind of a pain in the ass and so this program is overseen by doctor and a nurse and the two of them are just like listen in our hospital.

20:59I got 30 things beeping in any given time you're telling me I have a yellow light doesn't tell me why so they go back they work on explainable AI which is tell me why you're flagging tell me what I should be looking for and gradually the system really improves and so over the course of one year what they measured across the system was a 41% decline in mortality now was it all the software no the doctors were more sensitized to looking for sepsis but when I talked to doctor my heart was a lot more sensitive to the doctor's high average about this who's like what I mean it does somebody wanted to dispute that this is good like obviously this is good and so is it perfect no but AI doesn't need to be perfect to be good it's better than what we have and we're going to continue to work with it and so in both of those ways very quiet I don't think anybody coming to the hospital knows this stuff is happening but in one hospital system a thousand people saved in a year is like that's a big deal. Josh Tear and go he's the author of a brand new book it's called AI for good how real people are using artificial intelligence to fix things that matter we've given two examples so far we're not given the whole book away for free but we talked about AI being used for positive influence in education in medicine and now Josh I go to chapter six you write the following I was beginning to worry about my premise that AI could make government more effective and restore the bond between citizens and institutions and I'm going to do that.

22:27Veterans I 먹는 train I like to tell what is happening and how long I was living in I'm just asking questions and I'm like no but it's the way I was actually benefit All right that's what I was looking at and I was looking to get guys really great just going to do them right in the kitchen I'm goingadı knew on the job and kind of doesn't know what he doesn't know. And so he's working in environmental sciences, which means that he's in charge of recycling. He's in charge of water quality. And one day, like a trash guy comes in and says, Hey, I got this note. I got this email from somebody who says it can help improve our recycling. And so Cliff, so it sounds good. Turns out it's two guys from Saskatchewan. They're software engineers focusing on AI, but they just so happen to have come up through the trash industry. They worked in landfills, they worked in hauling. And what they've invented is basically AI image recognition trained on the difference between things that are recyclable and things that aren't. And you're talking about hundreds of thousands, if not millions

23:31of images taken in lots of different light, train the model, attaches to a recycling truck. And so as the recycling is poured into the truck, they're scanning each image. And if it's recyclable, no problem. And if it's not, they take a picture. And then this is where it becomes public policy. They just send because of GPS is on the truck. They send the homeowner a cart. And it says, Hey, did you know that this wasn't recyclable? In the future, try not to do this. Now, this is a small problem for municipality. But you know, when you do recycling in East Lansing, which has about 40,000 people, the city bails all of that recycling up and sells it to somebody. And if it's contaminated, they have to buy it back. And it goes to the trash, right? And if you can get it right, the city can make more money and you can actually help the environment. And so Cliff is like, doesn't know much about city government. So yeah, no, sounds great. Pilots, you know, $8,000 versus let's try it. And what was amazing to me about this is I, you know, I've been through city government. My wife,

24:35ran New York City Transit, she ran the MTA. And the notion that you would just kind of quietly adopt a program and let it loose into a city, you would never do. I mean, I told a couple of my friends about this and their eyes rolled back in their heads like, how is this going to work? But there is a really important lesson, which was one, the tech worked. I mean, it worked phenomenally to the point that there was almost no error rate. You have a high education, you know, population because it's right near Michigan State University. They want a recycle. They see dramatic improvements. And I was like, Cliff, how did you pull this off? He's like, well, maybe the lesson is we didn't tell everybody it was this big AI perfect, right? We didn't talk about surveillance. We didn't talk about the tech. We just said, this is going to make something better. He got two phone calls in the entire program. One was from a woman who was upset because she left, she thought her meal case, you know, meal kit was recyclable and it wasn't. One was from a guy who just wanted to be clear that it was his neighbor's dog's bag of poop, not his that had been

25:36attempted to recycle. And so, you know, the lesson that I saw, because I did a lot of case studies and a lot of work inside the federal government, which is obviously, as you cover every day, highly contested territory, right? Some people don't want AI to make the federal government more efficient and better. Some people are desperate for it, but don't want to think about compromising federal jobs. You know, you talk about the left and the right being at odds. The middle is where the competency is. And so if you can just remind people in the civic space that your civil servants are just trying to get you a better outcome. You don't talk as much about AI. You might actually be able to accomplish things. And so I was just kind of amazed. The story's sitting here, you know, Cliff is like a minor celebrity on the very small world of municipal recycling people because this program really worked. But these are the kind of things that give me a ton of hope. And by the way, they don't require open AI. They don't require anthropic. It's not about the biggest model with the most threatening power. It's actually a pretty small model. It doesn't, you know, there's nothing that really can go wrong with it other than it has an error in your

26:40recycling. But it's sitting there as an example of like, man, we can do a thousand things like this in the country today to make our lives better. You're booked for me. And this is our final subject. And thank you for being so gracious with your time. Your book for me was a reminder as to what a wasted opportunity doge was they were not warriors for inefficiency. They went looking for fraud and your word choice. They were dilatants and vandals. I think that's accurate. The federal government thing seems so rich in terms of opportunity. And you tell some encouraging things about even the IRS. I think in the book somewhere it says we've implemented more software changes in the last two years than the previous 20 years. But talk to me briefly about the federal government and the opportunity for good that AI can do to reduce inefficiency. Yeah, I mean, I'll just get briefly technical, which is that AI works best when you have data, right? You just need data. And the federal government's primary existence is to count, it counts

27:44things in a variety of realms, it keeps that data. And so all of these interactions with the federal government that are so frictionful and that make you question whether the institution is in the bureaucracy really know what they're doing can be accelerated and improved by the right application of AI. And so I spent a lot of time inside the IRS, which is a highly bemoaned institution, and yet critical to our survival, right? Without revenue, we have no country. And so what they did very quietly with a couple of just absolute warrior samar eyes is they modernized a bunch of the systems without drawing too much attention to it because they didn't want anyone to cripple their efforts so that AI can now come in and actually optimize for you. And so they were on the verge of turning over the individual master file. They were just about done with it, modernizing it so that AI could really it's work when dogerived. And the last of my interviews with the commissioner and other people were right before the election. And what happened after the fact is that doge,

28:44which, you know, I will tell you, a lot of people in government had some hope for because it's not like the people who really are working inside government think it's working great, but immediately became clear this was not about operational efficiency. This was about destruction. And so a bunch of them have hung in there. A bunch of these folks are still there to pick up the pieces. And if we ever get really committed about using it in a safe and productive way, like I'm convinced we can improve things rapidly. But we have to actually ask ourselves, do we want a country that works? Or do we want a country that proves that it can't work? And until that premise is solved, no tech is going to help. I'm going to read a paragraph from the book and in the epilogue and then I'll let you go if you want to just respond to this or amplify more importantly, you write, stay close to team human. This means supporting AI that amplifies human judgment, not replaces it. A teacher uses AI to customize lessons for 30 different students is different from a company using

29:46AI to eliminate teachers. A doctor using AI to spot patterns and scans is different from an insurance company using AI to deny claims. Support the former. Be vigilant about the latter. The question isn't whether AI is involved in our daily lives. Its utility is undeniable. And the regular improvements are worthy of awe. The question is who it's empowering. Final thought. Yeah, listen, it still holds. And I think it will hold for many, many years to come because software is only as good as the people using it and implementing it. So my hope is that we get through this moment pretty quickly and get back to trusting, you know, real people who want to help us solve our problems. Josh Tierengell, the book's terrific. And this is such a leveling of the playing surface. I can't thank you enough given all the negativity. I'm not saying unwarranted. Those concerns are warranted. But the positive aspects have not been told. And that's why I'm glad that you were here. So thank you. Thank you. I really appreciate it. The book is called AI

30:49for good. How real people are using artificial intelligence to fix things that matter. 99 cents. Ice cold. Perfectly refreshing. At extra mile, get a large fountain drink for just 99 cents with extra mile rewards. Refresh for less. Grab yours while at last. Feel your day at extra mile at select chevron and texico. Limits and terms apply. Real talent is defined by what people can do, not where they learn to do it. So by stopping at the education section of a resume, you might throw away the perfect tire. Skills first hiring helps you see talent others miss. Like more than 70 million stars, skilled through alternative routes. Let their story unfold and gain a competitive advantage because hiring managers who start with skills are 60% more likely to find a successful higher higher skills first. Learn why at tetherpapercealing.org. Brought to you by Opportunity at Work and the Ad Council.