AI for Good? The Case Against AI Doomerism

BIANNA GOLODRYGA: From the rise of chatbots to the push for automated warfare by the White House, artificial intelligence is becoming increasingly pervasive in our lives.

 

With little regulation, many fear its impact, including Pope Leo, who has teamed up with Anthropic to issue a document highlighting the need to protect jobs and truth amid this A.I. boom. On the flip side, some believe the technology is being harnessed for good. Author Josh Tyrangiel joins Walter Isaacson to share some examples.

 

ISAACSON: Josh Tyrangiel, welcome to the show.

 

TYRANGIEL: Thanks so much.

 

ISAACSON: So you’ve got this new book out, “AI for Good.” And I wanna read you something from the introduction that struck me. It said, “we are living through a moment in history that often feels cataclysmic. Climate change, extremism, institutional collapse, furious inequality. It’s natural, even rational, to assume a protective crouch against the future, yet my defenses were disabled by a glimpse into a completely different way in which we may live as citizens.” What do you mean by that?

 

TYRANGIEL: Well, I’ll take the first part first, which is the, the cataclysmic part. Right? I think AI has arrived in a very particular context, which is people are just drenched in existential risk and existential dread. And the people who are running these labs are not helping matters. They come in and they talk about you know, AI is either gonna cure cancer or mitigate climate change – which is great, but hard to believe – or it’s gonna doom human existence. And so that gets people to tune out right away. So that’s not helpful. The positive part is that when you separate the tech from the tech companies and you actually look at the ways in which it can help us solve meaningful problems, particularly in things like government and healthcare and education, it’s dazzling. Not always, and it’s not always easy, but the solutions that are available to us are completely different. And so what I found through the reporting is the tech can do a lot of things for us. We just have to mentally wrap our heads around what it can do and how it can do it. And it’s gonna take some work, but wow. When when you apply it to certain problems, it really does make a huge difference.

 

ISAACSON: Well, you say it could be dazzling, apply it to certain problems. Give me an example. I’m not sure we solved anything yet.

 

TYRANGIEL: Yeah. And so I wanna separate the tech from the hype. So I went to the Cleveland Clinic, right? And they were kind enough to just let me wander around, which I’m very grateful for. And they have a lot of AI pilot programs going on, and they’re all led by doctors. And, again, none of them are easy. They have a hundred years of established systems. They have patients with their various symptoms. They have doctors who hate changing their workflow, right? And yet they’ve made a lot of progress. 

 

So a big example is sepsis, right? Sepsis is one of the worst things that can happen inside a human body. It’s an out of control reaction to infection. And each year it kills about 350,000 Americans. More than breast cancer, prostate cancer, opioid addiction, combined. It’s very hard to detect because in its early phases, it actually just sort of presents like a cold or dehydration. And then before you know, it, it races away and it can kill you. So they did a pilot program and there was a human element to the pilot program, which is making sure that their clinicians are all much more aware of sepsis than they were. And then they did an AI pilot. And the AI pilot is software that is hooked up to every patient that comes in. And all it does is remind doctors through a prediction and detection engine that sepsis might be present, and it separates sepsis risk into three levels, and then it beeps. That’s all it does. The doctors are the ones who intercede. And yet over the course of a year using this sepsis prediction software, they reduced deaths in the hospital due to sepsis by 41%. So that is a thousand lives saved in part through the partnership of AI and doctors working together. So that’s one example, right? 

 

Another, you know we all know that, that healthcare is a terrible business for everyone, but the insurers, right? And for a hospital system, part of the problem is that hospitals are basically hotels. So they have patients, they have rooms, they have staff, food, linens, beds. And the key difference, the difference between profitability and hemorrhaging money is that hotels know when the customer is showing up and when they’re leaving, and hospitals don’t. And so what they did at the Cleveland Clinic is they worked with Palantir, very controversial company in some realms, but basically an enterprise software company. And they created a system to actually know when people are coming and going. And so what this software does is it hooks up to every data set inside the hospital, including electronic health records. And so when a doctor just makes a verbal note saying, this patient is likely to be released tomorrow the software knows it. And all of a sudden the hospital administrator can play the hospital like a video game, she knows when people are coming, she knows when they might be released. They’ve increased their transfer volume tremendously. They’ve cut down on emergency room wait times by 90 minutes. And I don’t know if you’ve watched The Pitt, but that’s a different show if you cut down on the wait times by 90 minutes. And so these are things that are just showing up, and they’re not revolutionary. They’re more evolutionary. But that’s just one example. And there are many –

 

ISAACSON: Wait, wait, Josh, what you’re saying, actually, I agree with. They’re not revolutionary. Those are evolutionary. It’s something my Hilton hotel could put in. Tell me, it seems somewhat disappointing that we’ve only gotten to that. I read in your book about the digital twin of a heart, that type of thing seems that’s a revolutionary leap.

 

TYRANGIEL: What we’re talking about as far as digital twins of hearts there’s a pair of doctors at Cleveland Clinic who are working on the ability to do AI assisted cardiac scan. And the idea here is that as opposed to every doctor starting fresh and doing their scans, and knowing where you are, you basically walk around with a twin of your heart and everything that’s happening to it. And at any given moment, we could run tests on that digital twin as opposed to having to run tests on you. So an AI model can run a test on an AI model, saving you a lot of heartbreak, customizing medicine. Now, they started this work about seven years ago. It’s taken time to work out bugs because as you would expect with AI, you know, you can do amazing things, but you gotta tweak it. You’ve gotta actually work with the patient, establish their comfort, but they have made a tremendous amount of progress. And so it’s these little evolutionary steps we’re seeing that are gonna get us to the revolutionary stuff. And it’s not that far away.

 

ISAACSON: You talk about Palantir, you mentioned it earlier in the interview. It’s in your Cleveland Clinic chapter. And you say, the very mention of Palantir causes people’s blood to curdle. Why?

 

TYRANGIEL: So Palantir on the left – Palantir was founded or co-founded by Peter Thiel. And Peter Thiel obviously is a very vociferous Trump supporter. He is in Silicon Valley, he’s a co-founder of PayPal. He’s very good friends with Elon Musk. I could go on about the reasons that the left hate him. The right doesn’t like Palantir because it really came in and challenged everything about the military industrial complex. And so Peter Thiel’s co-founder, is a guy named Alex Karp. He’s half black, he’s half Jewish. He’s a self-described socialist, a Kamala supporter. How the two of them are friends is a little bit of a mystery to me. But the combination has really freaked out just about everybody everywhere. And it does not help that the company is named after the mysterious stones in the Lord of the Rings. They cultivate a mystique. Okay? 

 

Now, the truth about Palantir is that what they do is almost comically dull. Okay? I just mentioned all the need to keep data clean and keep its infrastructure right. The chief architect of Palantir said, we are the mole people of Silicon Valley. We’re basically plumbers. We go in, we straighten out all the data pipelines, we clean all the data, and then we present the data on very clear dashboards so that someone running a company or a federal agency or a military operation can actually see what they’re dealing with and organize it and make decisions based on this data. And so I understand completely how Palantir has become a very political hot topic. What they do is very important and they do it very well.

 

ISAACSON: Yeah. But in New Orleans, for example, they were taking all the cameras all over town, doing facial recognition, putting that in the dataset, being able to follow people. Is that something that causes people to push back against AI?

 

TYRANGIEL: Absolutely, because if they are not brought into the process, if they’re not told why this might be good for society and why it might be good for themselves, the natural and completely understandable reaction is, I don’t want any of this. And so what we’re really talking about is we have a crisis of trust that we have earned in our society. If we are gonna get the best out of this material, we have to trust our institutions. Now, there’s a little bit of a Mobius strip in the logic here, right? Well, how can I trust institutions that are distrustful and use AI against me? And the answer is, we better figure out how to stop it somewhere. Now, I am, happen to be a big fan of government. You know, I’ve really enjoyed the last 70 years of peace and prosperity. And I do think that AI has a role to play in strengthening government and strengthening people’s trust in government. But two caveats. The first is it’s hard. It takes a lot of work. And the second is, no matter how good the tech, you still actually have to want to have a government for this stuff to work. Otherwise, AI can be just as destructive as it can be productive.

 

ISAACSON: Let me ask you the big question. Is AI gonna create more jobs or is it gonna reduce the number of jobs that humans do?

 

TYRANGIEL: So I, it’s a great question. A couple of months ago, I wrote a big cover story for The Atlantic about AI and the future of employment. And I’ll tell you what the economists mostly say. They say AI is a a general purpose technology, and they compare it to previous general purpose tech like electricity, right? Electricity came in, everybody knew it was great. It took about 40 years for the benefits to be felt across society. It changed a lot of jobs, but it increased productivity in America so much that it was worth it. Right? Now, the divide among economists tends to be – and a lot of Nobel winning economists hold to that. Younger economists I spoke with said they don’t think that their elders are misunderstanding the data. They think they’re misunderstanding the tech. And that AI, by its nature is smart machinery. And smart machinery can help roll itself out. 

 

So if the AI revolution in American life takes 10, 20, 30 years, we will have time with a natural rate of adjustment in labor to figure out where jobs move. There’ll be natural attrition as companies come online. And we’ll probably be okay. If it takes three to five years, the disruption will be significant. And the younger economists say it might be three to five years. And we might have a real crisis on our hands with unemployment rates, you know, rapidly going up, up to 10%, 15% in which –

 

ISAACSON: Well, wait, let me push back on that. I’ve heard them say that for the past three to five years. And you see it in all the press releases, all the anecdotes. Meta is laying off this, that, and the other. You look at the jobs numbers, it’s not there. Most recent job numbers, more employment. So if that’s supposed to be happening in the short term, where’s the data saying so? 

 

TYRANGIEL: Exactly. And even the, even the skeptics about the future will say it’s not yet showing up in the data. What they will tell you is that it is inevitable that by the time we get to the end of the year, they expect to see significant change. When I spoke to Fortune 100 CEOs, who are the people who, you know, employ most of America, they were also concerned, and they were concerned for different reasons. They obviously report to Wall Street. They’ve made these huge investments in AI, and they’re worried that they aren’t yet able to show growth as a result of that implementation. Because as we’ve discussed, it takes a little time. It’s not as easy as people say. It’s not a silver bullet to just flip a switch, and AI comes on. And so they’re concerned that what they will have to do to satisfy Wall Street is cut jobs. And so they think a bit of a self-fulfilling prophecy has come along where if they don’t cut those jobs, it’ll be their job that gets cut. And then what they’re saying is, you know, quietly, we wouldn’t mind it if Congress helped us here, if Congress regulated AI, if Congress put in more money for job retraining. Because if we lay these people off, and all of us do simultaneously, there’s gonna be a convulsion in society.

 

ISAACSON: Sal Khan’s been on this show often – from New Orleans and helped build Khan Academy because the school system was closed there – has now made a deal with Open AI from the very beginning, even before ChatGPT was released. And you went to watch how chat GPT is working in terms of K-12 education. What did you find?

 

TYRANGIEL: So yeah, Open AI and Khan Academy essentially became partners. So they worked with Chat GPT they stripped it of some of its biases. They stripped it of some of its ability to talk to kids in inappropriate ways. They trained it so that it actually became a tutor, didn’t give the answers away. It was very intense process, very strange process to work with an AI. And then they launched a thing called Kahnnmigo in classes. And what they’ve discovered over the last year or so, and I saw this in classrooms, is it for kids it’s fine. You know, kids have their laptops open with lots of tabs open. And for a lot of the kids I spoke with, they either said, oh, I didn’t realize that was AI, it’s fine. Or they resented it a little bit because they don’t like AI and they want their teachers to teach them. 

 

The magic of it, which I was surprised by, and I think Sal’s surprised by, is that teachers have found it incredibly useful that you can input a lesson plan into Kahnmigo and use its teaching tools. And essentially say, look, I’ve been lecturing kids on this material and I’m getting blank stares. How would I turn this material into a lab? How could I get people to activate in the classroom, talk to each other, and then scaffold the learning so that when they have accomplished something, they call me over and I can test it. And what I saw in this classroom in Indiana was a teacher who was really waltzing with Kahnmigo. She told it what she wanted to do, she gave it her lesson plan. She told her what materials she had in the classroom in order to make a lab, and then she would go back and forth with it. And so she sent me some homework the night before I attended, and I opened the Google Doc and was like, oh God, I’m not prepared to do chemistry. And then I walked in and I saw a joyful classroom of 10th and 11th graders, and that is not a common thing. 

 

And, you know, the world did not rearrange itself as a result of that one chemistry class, but it gives me hope and it gives me optimism that people who care about the things that we care about are using the tools in ways that are leading the charge.

 

ISAACSON: In the history of the digital revolution, there’s really been two strands. One of them you can call the Ada Lovelace strand, which she talks about the symbiosis, the partnership of humans and machines. You see that with Doug Engelbart when he invents the mouse and the easy to use interfaces. Steve Jobs is part of that strand. The other strand is sort of the computers will go off without us, the Alan Turing machines will be able to think on their own strand. And you get a lot of people these days talking about this singularity. What is your view? Will this proceed as a partnership of humans and machines, or will the machines eventually leave us behind?

 

TYRANGIEL: We have to dictate which way we want it to go. If we crouch in a defensive position, because AI seems hard, because there’s lots of other existential risk out there in the world, and we let the makers of the technology tell us how to use it, we are likely to get lots and lots of automation with lots of profits rolling up to those companies. If, however, we do get involved and we insist on uses that we care about that are collaborative, everything that we’ve talked about today involved a machine either alerting someone or helping someone see something. But it was always a human in the loop. It was always collaborative. If we can insist that that’s a best case use, then I think we may really well get gains.

 

But it’s, this is gonna be contested territory. And what I would say is most important is that if you don’t want to get involved, if you don’t wanna make those decisions, there are plenty of people who will make them for us. And the last time that happened was the social media age, and we saw how that ended. AI is so much more powerful than anything social media has to offer. The benefits are so much greater. So I would really encourage people to use the tools and begin to insist that this is the way they want them to be produced, and they want them to exist in the world. Otherwise it could go completely the other way.

 

ISAACSON: Josh Tyrangiel, thank you so much for joining us.

 

TYRANGIEL: Thank you, Walter.