Money Talks: Battle of the AI Bots

Summary

Slate Money – Money Talks: Battle of the AI Bots (Sept 16, 2025)

Host: Elizabeth Spiers
Guest: Gary Rivlin (Pulitzer Prize-winning investigative journalist, author of AI: Microsoft, Google, and the Trillion Dollar Race to Cash in on Artificial Intelligence)

---

Episode Overview

This episode takes a deep dive into the world of artificial intelligence, focusing on the current competitive landscape—including big tech’s “trillion dollar race” to dominate AI, the unique challenges facing start-ups, and differing philosophies guiding leading investors and founders. Journalist Gary Rivlin joins Elizabeth Spiers to discuss his new book and share behind-the-scenes stories of AI development, the culture of Silicon Valley, and the broader societal implications of AI’s rapid rise.

---

Key Discussion Points & Insights

1. The Trillion-Dollar AI Race: Incumbents vs Startups

• AI as 'Big Stakes':
  • Gary Rivlin describes the “mad dash to cash in on AI” and escalating stakes: “They don’t even really use the word billion anymore. It’s now the trillion dollar race because that really is the stakes here.” (01:14)
• Barriers for startups:
  • Discusses how big tech’s data, money, and Aqua Hires (talent acquisitions) make it nearly impossible for new players to compete:
    “It’s really hard for a startup to get access to that kind of money...How are two people in a dorm room going to have access to the data that you need for training and stuff?” (02:21)
  • Points to massive offers to individual AI researchers, citing a $250 million contract: “That’s like the highest paid baseball players kind of contract.” (02:58)
• The role of compute power:
  • Elizabeth Spiers points out that, unlike classic “software is scalable” logic, AI is resource-intensive at scale due to compute and data requirements. (03:14)

---

2. Inside Inflection AI: Personality, Emotional Intelligence, and Philosophy

• Origins of the book and Inflection’s vision:
  • Rivlin was inspired by an email from Reid Hoffman (LinkedIn co-founder) about launching Inflection AI, focused on making computers “speak human.” (03:48)
  • Hoffman backed OpenAI, is on Microsoft’s board, but sought to build a product prioritizing EQ (emotional intelligence), not just IQ.
• Building an emotionally attuned chatbot ("Pi"):
  • The chatbot Pi aimed to be engaging, friendly, and empathetic—sometimes to a fault:
    “It had really loved emojis at first. They had like, okay, chill folk. PI it was called. PI is the name of their chatbot.” (05:01)
• Balancing friendliness and utility:
  • The engineering challenge: finding the sweet spot between being too sycophantic and too bland. (06:46)

---

3. Human-AI Attachment and Public Response

• People get attached to bots:
  • Some users developed deep, even “romantic” bonds with bots (07:15)
  • Quote: “I think the future is going to be very, very interesting. Very, very weird because 10 years from now there are going to be people who have very deep, I call it romantic relationships. I mean, it’s already starting to happen, but I think it’s going to be far more commonplace.” — Gary Rivlin (07:36)
• Product changes disrupt relationships:
  • OpenAI’s switch from GPT-4 to GPT-5 frustrated users who’d bonded with previous versions. (07:03)

---

4. Hype, Disappointment, and the Reality of AI Progress

• Hype cycles hurt AI’s perception:
  • Rivlin criticizes tech leaders for overselling incremental improvements:
    “Hype is really hurting the cause of those advancing AI. I use AI all the time…It’s a tool. A calculator is a tool, A camera is a tool. It’s amazing…But as a co-pilot.” (08:27)
  • Contrasts ChatGPT’s understated, viral launch in 2022 with today’s splashy announce-and-disappoint cycles. (09:36)

---

5. Why Did OpenAI Break Through?

• History and missed opportunities:
  • AI as a field since the 1950s; machine learning broke through in the 2010s
  • Google had ChatGPT-like capability as early as 2020 but hesitated to launch, wary of PR disasters (e.g., Microsoft’s failed 2016 Tay chatbot):
    “The specter of Tay just kind of hung over Microsoft, Google, every big company. So… it was inevitable it was going to be a startup… because they didn’t have as much to lose.” (11:29)
• AI was already everywhere, just invisible:
  • Search, Netflix recs, Google Translate—AI has run in the background for years, but ChatGPT put it “front and center” where consumers could interact directly. (12:34)

---

6. Techno-Optimism, Pessimism, and The ‘Bloomer’ Middle

• Comparing Hoffman and Andreessen:
  • Rivlin distinguishes Mark Andreessen's hardline “zoomers” (full speed ahead, no regulation) from “doomers” (AI apocalypse), and situates himself and Hoffman as “bloomers”—optimistic but urging caution and regulation. (17:05–18:28)
• Trust and safety lost primacy:
  • As the stakes and arms race intensified, trust and safety teams at Google and others were deprioritized:
    “The trust and safety teams, like, oh no, we’re in a war against OpenAI… So a lot of people were shifted from trust and safety.” (18:52)
• Ethos of “fix it afterwards” in Silicon Valley:
  • “There is an ethos in Silicon Valley where you ship your product and then if it has deleterious effects somewhere, you fix it after it happens. And that’s how a lot of the really destructive things happen.” — Elizabeth Spiers (20:28)

---

7. Beyond Hype: AI, Electricity, and Climate Impact

• Data centers and climate:
  • The arms race strains electrical grids and may soon drive up consumer rates and cause brownouts. Investment in green tech isn’t keeping pace. (21:53–22:59)
  • “It’s really putting a huge strain on the electric grid...within the next year or two, consumer rates are going to go way up and we’re going to start to have brownouts.” — Gary Rivlin (21:31)

---

8. The Limitations and Risks of LLMs (Large Language Models)

• AGI is far off—LLMs have clear limits:
  • Cites Gary Marcus and the “stochastic parrot” criticism:
    “They know a lot about everything...but they don’t understand a word. They don’t have common sense, they don’t have reason.” (24:03)
• Autonomous AI is cause for concern:
  • “Another big fear I have is autonomous AI. AI in charge? No, no, no. It’s a tool. A human has to be in charge.” (24:39)

---

9. Jobs, Automation, and Economic Consequences

• Displacement and slow adoption:
  • AI likely to slowly but steadily shrink many professional teams; may eliminate more jobs than it creates:
    “The marketing team of 20…is going to be a marketing team of 15 or 10 or 8. Because you could use AI as sort of your junior member of your team.” (25:42)
• AI as a tool for research and brainstorming:
  • Rivlin’s personal uses: research assistant, editor, idea generator, but not for actual writing.
    “If I made a living writing press releases, I’d be worried because that is formulaic.” (27:05)

---

10. AI’s Usefulness, Limits, and Forgiveness

• AI is not perfect, and neither are humans:
  • The public is less forgiving of AI errors because we don’t trust it’ll learn the way humans do. (29:29)
  • “I’ve hired interns in the past and they’ve made mistakes too. Sometimes I think we’re unforgiving about AI, about the mistakes it makes. Humans can say really stupid things too.” — Gary Rivlin (27:40)
• Best role = Supercharged assistant:
  • Summarization, checklists, brainstorming; ultimately, it’s always human-in-the-loop. (29:11–29:52)

---

11. Creativity, Human Agency, and the Myth of AI as Creator

• AI gives creative ‘superpowers,’ but concept must come from humans:
  • “You, the human, still have to have the plots, the characters, the twists and all that. And I think that’s the way to understand it. You could be faster…but it’s still as an artist, as a creator… you’re still the one in charge.” (31:59)
• Memorable user experiences:
  • Spiers describes using a generative model to produce abstract angel imagery, which inspired but did not replace her artist’s creativity. (33:08)
  • Rivlin’s story: Google Gemini animated his son’s drawing and eerily grasped the story’s theme.

---

12. What’s Next? Cautious Optimism, Pessimism, and Busts

• P(doom) and long-term outlook:
  • “So this idea of P Doom, it’s not zero…it’s very, very low. These things aren’t even close to doing, able to do the kind of things we’re scared of. They’re really limited.” — Gary Rivlin (36:10)
  • Predicts disappointment and a possible “AI bust,” analogous to the dot-com bubble. Progress and mainstream adoption will take 10–15 years. (36:10–38:31)
  • Key worry: placing AI in autonomous decision-making roles before it’s ready.

---

Notable Quotes & Memorable Moments

• “I call [Reid Hoffman] a billionaire you can almost root for.” — Gary Rivlin (05:46)
• “If people use ChatGPT… It’s a conversation. You go to ChatGPT and it’s like, what should I do in this situation? And here’s five things to consider. They just give you bullet points. But with Pi, it was kind of fun… it would compliment you on your stupid jokes.” — Gary Rivlin (06:18)
• “People get very attached to their bots and do insane things.” — Elizabeth Spiers (07:03)
• “Hype is really hurting the cause of those advancing AI… Don’t tell us it’s amazing because we’re going to be disappointed because it’s not artificial general intelligence. It’s a slight improvement for most of us on the previous version.” — Gary Rivlin (09:17)
• “Don’t be evil. I thought that is the lowest bar in the world.” — Gary Rivlin (20:04)
• “The real way I use it, or maybe the most valuable way, it’s a great editor. I’ll tell it, something’s not working here. I need a transition. Give me 10 different transitions… But it gives me a lot of good ideas and it’s good at spotting mistakes.” — Gary Rivlin (28:36)
• “My favorite construct for AI is alien intelligence. We’ve never had anything like this. Again, it knows so much about everything…but doesn’t understand a thing.” — Gary Rivlin (29:59)
• “It gives us superpowers…but it’s you. The human had this idea, use this tool, shared it with another human who didn’t copy it.” — Gary Rivlin (34:01)
• “It is here. It’s going to have a big impact on society… Let’s figure out what’s the best approach so it would be more of a positive thing and negative… But we could debate whether television, the car, you know, the Internet are good things. And I think that’s the kind of debate we will be having. But let’s do what we can now to ensure it’s more of a positive than a negative.” — Gary Rivlin (35:49)

---

Timestamps for Key Segments

• 01:14 — Why start-ups can’t compete with big tech in AI
• 03:48 — Origins of Inflection AI and the push for emotional intelligence
• 07:15 — Human attachment to AI bots and product transitions
• 08:27 — Hype cycles and public disappointment with new AI releases
• 10:47 — Why OpenAI overtook Google despite big tech’s early leads
• 17:05 — Comparing Andreessen’s and Hoffman’s tech philosophies
• 21:31 — AI’s impact on energy use and climate
• 24:03 — Stochastic parrots: the limits of current LLMs
• 25:42 — Job displacement and economic impact
• 27:05 — How writers and researchers use AI as a tool
• 29:59 — “Alien intelligence” as a metaphor for AI
• 33:08 — Creativity with AI as a jumping off point
• 36:10 — P(doom), the ‘AI bust’ and the long view on AI’s impact

---

Summary

This wide-ranging episode explores the fierce competition to lead in artificial intelligence, the real-world limits of current AI technology, human behaviors and irrational attachments to bots, and the intertwined risks and hopes for a future shaped by AI. Through candid and often humorous anecdotes, Gary Rivlin demystifies the nature of AI breakthroughs and underlines the continued indispensability of human creativity, judgment, and oversight—even as supercharged “AI co-pilots” become part of everyday work and life. While the trillion-dollar arms race continues among giants, the guest and host urge listeners to prioritize careful, collaborative, and ethical deployment—before the AI tide becomes truly unstoppable.

Transcript

[Sponsor/Announcer]

This message is brought to you by apple card. Get 3% daily cash back when you buy an iPhone with your Apple card at Apple. Subject to credit approval. Apple Card issued by Goldman Sachs Bank USA Salt Lake City Branch terms and more at applecard.com slate money is brought to you by Charles Schwab. Decisions made in Washington can affect your portfolio and every day. But what policy changes should investors be watching? Washington Wise is an original podcast from Charles Schwab that unpacks the stories making news in Washington right now and how they may affect your finances and portfolio. [e-mail] WashingtonWise.

[Elizabeth Spiers]

Hello and welcome to Money Talks. I'm your host Elizabeth Spires. And today I'm joined by Pulitzer Prize winning investigative journalist Gary Rivlin, who has a book out. It's called AI Microsoft, Google and the Trillion Dollar Race to Cash in on Artificial Intelligence. So Gary, introduce yourself and tell us about the book.

[Gary Rivlin]

Great to be here. I just had crazy timing right at the end of 2022. I had this idea like, hey, I'm hearing AI starting to get hot. Just before ChatGPT, before the whole world was talking about ChatGPT, I started contacting people in Silicon Valley looking for those trying to get rich. You know, the mad dash to cash in on AI. So that was that. That was my angle from the start. Let others kind of talk about the bigger issues. What does it mean that we're talking to a machine? Whatever. I just love the stakes. They don't even really use the word billion anymore. It's now the trillion dollar raise because that really is the stakes here. It's, it's, it's many trillions of dollars. Funny enough, I, I was paying attention at first to startups and I thought like, hey, who's going to be the next Google? Who's going to be the next Facebook Meta? And it turns out this rather depressing to me. I think the next Google is going to be Google and the next meta Facebook is going to be meta Facebook. But we could get into that.

[Elizabeth Spiers]

I mean that makes sense because they're all, you know, investing heavily and doing acquisitions in some cases incredibly high dollar Aqua hires in order to, I think vacuum up all of the AI talent in the Valley.

[Gary Rivlin]

Yeah, I mean the Aqua Hire for those who don't know is like, it's, it's buying a company largely just for the talent and there have been like multibillion dollar Aqua hire, so spending multiple billions of dollars just to get the two or three top AI researchers you want. But you know, across the Board. This is so expensive. It's like it costs hundreds of millions nowadays, billions to train and operate these models. It's really hard for a startup to get access to that kind of money. You, I mean Google has all this search data. How is two people in a dorm room going to have access to the data that you need for training and stuff? And of course, you know, these days, there was a news story a couple of weeks ago in the Times, New York Times that Facebook was offered someone or gave someone a $250 million.

[Elizabeth Spiers]

Yeah, that's contract.

[Gary Rivlin]

It's like, that's like the highest paid baseball players kind of contract. So you know, it's hard for a startup to compete in that realm.

[Elizabeth Spiers]

It's funny because a lot of people view technology as something that's fundamentally lower barrier to ent because software is scalable in a way that other businesses aren't. But with AI, it's very different. When AI people talk about compute, they're talking about a limited resource and the amount of investment you need to actually do this. So just to go back to your book, your book is primarily about Reid Hoffman's AI ventures and his primary company that he invested in is not a household name. It's called Inflection AI. Can you tell us a little bit about what happened there?

[Gary Rivlin]

It was an email from Reid Hoffman in 2022 that gave rise to this book. And he just, he's the founder of LinkedIn and co founder of LinkedIn and he said, oh, I'm co founding my first company. It was like dear friends. It turns out I'm one of his 2,500 friends, but so am I. You know, I was like, hey, I've co founded my first startup since LinkedIn and I had programmed earlier in my life. And like I just love this idea he laid out. Instead of us having to invent machine languages, instead of having to learn the computer's language, the computer's going to speak our languages and that just kind of like speaking human. I just loved that idea. And so Hoffman, he was one of the initial investors in OpenAI, which gave us ChatGPT. He's on the board of Microsoft. They've been a central player in AI. But I love the idea that he was coming up with a startup that wasn't paying attention just to IQ, but EQ emotional intelligence. And you know, they really kind of staked that out and you know, I, I played with it, it was rather extraordinary and I, I kind of think that's underrated these days and how important it is to winning the AI chatbot war is like, you know, one of the more popular ones is. Is Claude from Anthropic. And that's because it's kind of fun to talk to. It's playful. And so I spent time with the other co founders, engineers, you know, those working on this, trying to like get it just right. It was, you know, sort of whack a mole. You want it too friendly and then it's, it's glazing you, it's complimenting you, it's. It's just kind of insipid but you know, then if you kind of turn that down. So it was kind of really fun to. It had really loved emojis at first. They had like, okay, chill folk. PI it was called. PI is the name of their chatbot.

[Elizabeth Spiers]

How much of that do you think is really just a function of the founder ethos? Like I know Reid a little bit and he's kind of, he's teddy bearish, but will also be formidable when the situation calls for it. But he has, I think higher EQ than most billionaires. How much of that was really sort of built into sort of philosophical thing from the beginning?

[Gary Rivlin]

I like that description of, of Hoffman, I call him a billionaire. You can almost root for. But inflection came more from partner, the co founder, the main impetus for the company, the CEO, this guy, Mustafa Suleiman. Mustafa Suleiman is kind of royalty within the machine learning within the AI world because he was a co founder of DeepMind, the first great machine learning company. Goes back to 2010 and founded in London. Google bought it for like $650 million in 2014. And you know, I'll give them credit. Hoffman, Suleiman, the whole team, they really cared about personality, they really cared about safety. They really wanted a certain bot that had a certain sensibility that you could go to it and query and have a serious conversation about politics or life or I'm scared of this. And you know, as opposed to, you know, if people use ChatGPT, it's. They're changing, they're starting to really amp up the EQ. Especially with 4 and 5 where it was a conversation. You go to ChatGPT and it's like, what should I do in this situation? And here's five things to consider. They just give you bullet points. But with pie, it was kind of fun. You'd have this dialogue and hey, how you doing? And you know, you kind of chit chat and you compliment you on your stupid jokes. Again, it really was too much of.

[Elizabeth Spiers]

A thing to do. Sycophantic.

[Gary Rivlin]

Yeah, yeah, sycophantic.

[Elizabeth Spiers]

That's been a recurring criticism of some of the AI bots, especially whenever they lean into that. It ends up. People get very attached to their bots and do insane things.

[Gary Rivlin]

Actually OpenAI when they put out GPT5, they just said, this is what you're going to use. You're not going to use the past methods. And many people, including me, were very pleased because if you don't pay for it, you were using an old version. And here I was getting the cutting edge, best version of it. But a lot of people were really disappointed because they had developed a relationship with GPT 4.0 or 4.0 and so they got a lot of heat for it. Like wait, you know, you just overnight just kind of remove this relationship. I have. I think the future is going to be very, very interesting. Very, very weird because 10 years from now there are going to people who have very deep, I call it romantic relationships. I mean, it's already starting to happen, but I think it's going to be far more commonplace.

[Elizabeth Spiers]

There was a tabloid story yesterday about a woman who got engaged to her AI boyfriend.

[Gary Rivlin]

So mazel tov.

[Elizabeth Spiers]

I guess it's already happening and GPT5 has not gotten a great reception. And I think partly because of what you were talking about, but also because Altman kept talking about it like it was the arrival of general intelligence AGI. And then what people actually got was what seems to be a more incremental model. Have you played with it? What are your thoughts?

[Gary Rivlin]

Yeah, kind of the bigger point here is hype is really overhype is really hurting the cause of those advancing AI. I use AI all the time, all day for research assistant to bounce ideas off of. It's a pretty good editor. I use them all. I feel it's kind of my job. So. You know, Claude happens to be my favorite for writing, editing. But yeah, I've been using ChatGPT because the upgrade, I think it's a miracle. I think it's alchemy. I can't believe I could work with this thing. And it writes so fast and like, you know, it's an amazing tool, but the problem is the rhetoric coming out of Silicon Valley as, as if it's this Uber thing that's going to run our lives and make all decisions like, no, it's a tool. A calculator is a tool, A camera is a tool. It's amazing. It lets you do amazing, amazing things. But as a co pilot, we could get into the whole artificial general intelligence debate. But there was such a drum roll. Altman was, was, you know, promoting it on X and, you know, other social media, and it's just, it was inevitable disappointment. As opposed to ChatGPT when it came out November 30, 2022, I greatest flex in the history of technology. They didn't call a press conference, they didn't put out a press release. There was no press event. They just posted it with a research note and, you know, a few people put it on X. And it kind of took some time, took a few weeks before it became a big deal, before people, you know, a lot of people started using it. Now, of course, it's, you know, the Steve Jobsian, like, events where the great Oracle, in this case, Sam Altman, comes on stage and, you know, and it's just like, yeah, no, I would just go back to what you had. Just let people discover. Let people tell you it's amazing. Don't tell us it's amazing because we're going to be disappointing because it's not artificial general intelligence. It's a slight improvement for most of us on the previous version.

[Elizabeth Spiers]

What I find sort of surprising is that we use AI every day. It's built into most of the big software that we use, and it's been around for decades. I remember reading Marvin Minsky's book, you know, in 2007, and I think the real inflection point was really, it was ChatGPT being released because it was positioned as a consumer product. And I think it was probably the first one that people really played with, and they were astonished by it. Why didn't the other bots break through like that?

[Gary Rivlin]

That's a great question, that. But to back up a bit, like, the great surprise for me when I started doing research on this is like, AI has been around since the 1950s. The term artificial intelligence was coined in the mid-1950s, and there's been all sorts of advances, you know, key points, you know, kind of in the early 2000 and tens, they realized, like, wait, we shouldn't, like, code every possible alternative. That's not the way to create artificial intelligence. Let's have machine learning. Let's kind of. It's akin to the way a human learns. We'll feed it data, books, articles, kind of Reddit posts, whatever, and it'll learn like humans do with, you know, with feedback, like, yeah, that's not a great answer. It will learn over time. So that was a great breakthrough. But, you know, if I worked at Google in late 2000 and tens or early 2000 and twenties, I would have been going out of my mind. In fact, Mustafa Suleman was. And in fact, banging on the table like, this is amazing what we have. They had ChatGPT as early as 2020. They had. Wasn't called that, but they had a bot that could do exactly what we all discovered when we started playing with ChatGPT, you could do. But Google's a big company. Google brings in tens of billions, hundreds of billions of dollars. And they were worried. The PR arm, the marketing arm was worried, what will it mean? What happens if it starts misbehaving? Because there have been examples. In 2016, Microsoft put out Tay. It was one of these chat bots, trained, not a very good idea in retrospect on social media. And within 24 hours they had to pull it down because it was, you know, a Nazi saluting white supremacist and, you know, saying anti Semitic stuff. And so they had to pull it and say, you know, kind of the, the specter of Tay just kind of hung over Microsoft, Google, every big company. So in retrospect, it was inevitable that was going to be a startup, a startup like OpenAI that put this thing out because they didn't have as much to lose. And you put your finger on it, like Google Translate's been around since the mid 2010s. That's AI. You know, the suggestions we get from Netflix, you might like this movie based on your previous watching habits. That's AI. But that's in the background. What was astonishing, what is astonishing about ChatGPT and the other large language models, the chat bots, is you could talk to it. You see it working, you hear it working.

[Elizabeth Spiers]

Okay, we're going to stop right there and take a quick break for an ad and then we'll be back with Gary Riplin.

[Sponsor/Announcer]

This podcast is brought to you by Progressive Insurance. Do you ever think about switching insurance companies to see if you could save some cash? Progressive makes it easy. Just drop in some details about yourself and see if you're eligible to save money. When you bundle your home and auto policies. The process only takes minutes and it could mean hundreds more in your pocket. Visit progressive.com after this episode to see if you can save Progressive Casualty Insurance company and affiliates. Potential savings will vary. Not available in all states.

[Saks Fifth Avenue Advertiser]

Sleep Money is sponsored this week by Saks Saks Fifth Avenue makes it easy to shop for your personal style. This season. Fall is here and there are so many new fall arrivals that you're going to want to wear again and again. There's a great new relaxed Prada blazer. There are Gucci loafers you can take from work to the weekend. It is incredibly Easy to visit Saks.com and find new arrivals from your favorite designers. I kind of love the shirts from Commes des Garcons. I can't always afford them, but it is definitely always there on my Inspo board. And once in a blue moon I might even buy one. Saks makes shopping feel customized to you. They have in store stylists. They have Saks.com showing you only what you like to shop. They will even let you know when arrivals from your favorite designers are in or when something you love is back in stock. So find inspiration for your personal style every day at Saks Fifth Avenue.

[Sponsor/Announcer]

This message is brought to you by Apple Card. Each Apple product, like the iPhone, is thoughtfully designed by skilled designers. The Titanium Apple Card is no different. It's laser etched, has no numbers and it earns you daily cash on everything you buy, including 3% back on everything at Apple. Apply for Apple Card on your iPhone in minutes, subject to credit approval. Apple Card is issued by Goldman Sachs Bank USA, Salt Lake City branch terms and [e-mail].

[Gary Rivlin]

This podcast is brought to you by Progressive Insurance. You chose to hit play on this podcast today. Smart Choice Progressive loves to help people make smart choices and that's why they offer a tool called Auto Quote Explorer that allows you to compare your Progressive car insurance quote with rates from other companies so you can save time on the research and can enjoy savings when you choose the best rate for you. Give it a try after this [e-mail] Progressive Casualty Insurance Company and affiliates not available in all states and situations. Prices vary based on how you buy.

[Sponsor/Announcer]

Slate money is brought to you by Charles Schwab. Decisions made in Washington can affect your portfolio every day, but what policy changes should investors be watching? Listen to Washington Wise, an original podcast for investors from Charles Schwab to hear the stories making news in Washington right now. Host Mike Townsend, Charles Schwab's Managing Director for Legislative and Regulatory affairs, takes a nonpartisan look at the stories that matter most to investors, including policy initiatives for retirement, savings, taxes and trade, inflation concerns, the Federal Reserve, and how regulatory developments can affect companies, sectors and even the entire market. Mike and his guests offer their perspective on how policy changes could affect what you do with your portfolio. Download the latest episode and [e-mail] WashingtonWyse or wherever you listen.

[Elizabeth Spiers]

When you started this book, I think you were thinking about writing or Started to write a book about Mark Andreessen. And you decided to do this instead. So I'm curious about how you contrast Andresen's version of techno optimism with say, Reid's or do you think that there's a material difference between them?

[Gary Rivlin]

It's funny. So I was thinking in the fall of 2022 and I was unfortunately right. Like wait, you know, kind of the base, the center of a lot of the right wing stuff was Silicon Valley and Dreeson had a big cloud forum. Peter Thiel and these are folks. So I've been covering tech since the mid-1990s. I read for Wired, New York Times, it was covering these folks and you know, they would all talk to me. In fact, it's all my fault. I. I did the first big profile of Peter Thiel. I put him in the new magazine in 2004, 2005. You know, it's funny, I always found him such an interesting guy. Obviously we have very different politics, but it was something like, you know, he's principled and I kind of feel like all of that's gone out the window.

[Elizabeth Spiers]

Well, Nick Denton found him interesting too, and that's how this whole thing started.

[Gary Rivlin]

Yeah, I'm not the only one apparently, you know. But getting to your, your question, it's like Hoffman has a really interesting construction. There's the Doomers and those. I feel like they're kind of watch too many Hollywood movies. They're just fearing like laser eyed robots subjugating humanity. Like we have a lot to worry about in AI and that's so far in the future. Let's worry about the stuff now. Jobs, surveillance, AI and warfare. I have a long list of those. So there's the Doomers on one side and then the other. And this is where I put Mark Andreessen and much of Silicon Valley, much of those working on AI right now. Zoomers. Anything that gets in the way of AI is a crime against humanity. Because AI could do so much good in medicine and science and education, et cetera, that any regulation, even trust and safety teams within the corporations themselves, the creators with these models, that's the enemy. Reid Hoffman, and I put myself here is somewhere in the middle. I mean he's closer to a Zoomer than a Doomer and he calls him a bloomer and that's like he. Again, I put myself in this category. I see great potential in AI. I think if we're not stupid about it, by the way, that's a big. But it could be a big net positive. There will be Negatives, like with any new technologies. But I think this could be a real positive. But let's be smart about it, let's be deliberate about it, let's have guardrails in place. Let's really listen to trust and safety. One of the interesting things. So in 2022, when I started this, trust and safety was front and center Every. In fact, OpenAI was created in the mid 2010s because they worried about Google only looking at the profit motive for AI. It's too important to leave it to the, you know, for profit. And obviously that went by the wayside with OpenAI, but it went by the wayside with Google. Google in the late 2010s promised they would never let their AI used for surveillance or warfare started. 2025, they dropped that. The trust and safety teams, like, oh no, we're in a war against OpenAI. We're in a race, an arms race against OpenAI and the other makers of these LLMs. And so a lot of people were shifted from trust and safety to let's be.

[Elizabeth Spiers]

Well, they went from don't be evil to be a little bit evil, selectively here and there if necessary.

[Gary Rivlin]

But don't you. I always thought it funny. I covered Google going public. I met the two founders when they were a 40 person company in like 2000. I always said, don't be evil. I thought that is the lowest bar in the world. Like if there's a homeless guy in the street, don't punch him. I don't be evil. Like, yeah, okay, that's, that's easy. Like try to be good. I, you know, I always said that was kind of such a low bar and they're even sometimes failing by that low bar.

[Elizabeth Spiers]

It kind of is, but not for, from a perspective of I don't even think it's malicious intent. But there is an ethos in Silicon Valley where you ship your product and then if it has deleterious effects somewhere, you fix it after it happens. And that's how a lot of the really destructive things happen.

[Gary Rivlin]

I mean, this in a way is my single biggest worry with AI it's in the hands of too few people and it's people who have proven that they're not very trustworthy. And so I do believe Google has the best intentions, but they also care more about winning this race. It's a trillion dollar stakes. It's the future of the company and I think they rationalize a lot in the name of that. In fact, another big worry about AI is, you know, it's on A collision course with climate change. These data centers, they're unquenchable. They're just building more and more and more of it. It's really putting a huge strain on the electric grid. But so meanwhile, you have Google and Microsoft and Meta have these climate neutrality promises. So they're investing money in green energy and stuff, but that's just an accounting trick. It's like there's a fixed amount that's being produced at this point. And so they are really, because of their desire to win the AI wars. I think within the next year or two, consumer rates are going to go way up and we're going to start to have brownouts and, you know, kind.

[Elizabeth Spiers]

Of restrictions in some areas. There are demonstrable proof that rates are going up in areas. And it's interesting because, you know, Elon Musk is, you know, re headquartering X in Texas, which already has a very fragile grid that's been largely privatized. And, you know, if their data centers are there, that seems like a recipe for disaster, right?

[Gary Rivlin]

It's starting to happen. I'm just saying I think it's going to be widespread. You know, it's interesting. I hadn't quite realized this before doing the research. Like the last couple of decades, you know, electricity, electricity demand has been pretty flat. And so, you know, prices might creep up because they have to upgrade equipment and all. But now largely with AI, but electric, vehicles, whatever, a host of things, demand is spiking. We're slowly adding capacity. And, you know, now Trump, with his big ugly bill and other actions, is taking away the financial incentives for wind, solar, wind, battery, geothermal. And so it's making it where a lot of that stuff is being stopped at the exact same point where a lot of that stuff should be accelerating ahead.

[Elizabeth Spiers]

There's a school of thought that says part of the problem is that when we think about AI and where the money is going, it's all going into large language models. And there are other types of AI that might be promising. Gary Marcus, who is a very, I would say skeptic, but also enthusiastic about AI, likes to argue that part of the problem is that we won't get to AGI until the models are really able to generalize in a broad way. And so he argues for a kind of model that's a neurosymbolic model, which ostensibly could be less of an energy suck than the LLMs that we are now all taking for granted are core de facto AI. Where do you see that going? Or are we so invested in LLMs. Now that's just the inevitable route that we're taking.

[Gary Rivlin]

Well, I'm some of the more interesting stuff coming out of Google in recent weeks is around image models, like either video or still images. So there's plenty of money being put into that. To me, where I really agree with Gary Marcus and people a lot smarter than me about AI also agree, like machine learning, large language models brought us so far. But we really need another breakthrough or two before these things really can attain anything like artificial general intelligence. You know, this is this idea that a machine, a model could best humans at more or less everything, brain wise. And so because you see the limits, I mean the phrase I love Professor Bender at University of Washington, stochastic parrot. And that's just this idea like these things are amazing. They know a lot about everything. They have a PhD level knowledge on history and subspecialties of science across the board, but they don't understand a word. Yeah, they don't have common sense, they don't have reason. And you know, another big fear I have is autonomous AI. AI in charge? No, no, no. It's a tool. A human has to be in charge. We have the common sense, we have the idea, we have the goal and this thing can help us get there. But I really fear with some of the rhetoric and of course it's a cost saving for corporations, they always like that, that we're going to have autonomous AI when it could do really stupid things, even though it's in quotes. Smart.

[Elizabeth Spiers]

My, my favorite recent example of this is I was looking for foods that are high in potassium and so I, I just googled how much potassium does a banana have? And the AI response gave me a number, but also said, you know, this is x percent of the daily value, therefore you should eat 14 bananas a day. And that was the kind of reasoning that you get. It's entertaining until you realize how many companies want to incorporate AI for actual decision making, especially when they believe that they can replace very high cost humans.

[Gary Rivlin]

Exactly, exactly. Now we're getting to another fear. You know, jobs, I think it's much more of a slow roll than people think. Like corporations want to save money, but corporations are conservative, they're testing, they're piloting. Let someone else do it because it's expensive right now. So I think it's going to be five, 10 years, maybe 15 years before this really kind of infiltrates every aspect of society. You know, schools, business, social relations. It's happening, but I think it's going to be a long time. But it is happening. So the marketing team of 20 and some big corporations going to be a marketing team of 15 or 10 or 8. Because you could use AI as sort of your junior member of your team. Like give me 15 illustrations of this idea, give me 100 ideas. And so I think AI really is coming for the job. It will create new opportunities, new job categories, just like the Internet, just like every technology. But I think it's going to eliminate far more jobs than it creates.

[Elizabeth Spiers]

So how do you use AI in your everyday life?

[Gary Rivlin]

Well, the first thing I did when I got this idea for a book is write me a 5,000 word book proposal for a book about AI. And you know, actually I learned two things. Like it's much better read, it has a much better memory than I had. So they're like little insights or factoids, data points I could use. But the other thing is like it can't write for anything. It's very formulaic.

[Elizabeth Spiers]

It flattens a lot of cliches, a.

[Gary Rivlin]

Lot of, you know, if I made a living writing press releases, I'd be worried because that is formulaic. But it's worth, I don't know, GPT9. You know, maybe eventually it's going to be able to write a novel that impresses us.

[Elizabeth Spiers]

But can it write a joke? That's been my bar. Can it write a good joke that comes out of nothing?

[Gary Rivlin]

That's the Reid Hoffman would say. Yes, I've read the jokes that he thinks are good. I'd say not quite, not quite yet. It gets the idea of a joke but you know, so that's my long winded way of saying I don't use it for writing. But you know, it's a really interesting research assistant. Oh, what about hallucinations? Well, I use the ones that footnote.

[Elizabeth Spiers]

Yeah.

[Gary Rivlin]

And so like, oh, that's a really interesting quote. That's a really good data point. And then I go, oh good, CNBC has an article that person did in fact say that quote. So I, you know, I need to verify. But then again I've, you know, hired interns in the past and they've made mistakes too. Sometimes I think we're unforgiving about AI, that the mistakes it makes. Humans can say really stupid things too. Like 35 to 40,000 humans die every year in the US because of auto accidents. There's rarely a headline about it. But if a self driving car killed somebody, it would make a headline, that kind of idea. But so I use it for research. It's really good for like a sounding board I've used it like I have to have a tough conversation with one of my brothers, you know, let's play it out kind of thing. I used PI. My son was facing a serious surgery. He's fine. It went great, but we were really worried about it. And like, I'll give pie credit. It asked all the right things, it expressed concern. It. I don't know if it meant anything to me, but it did get me thinking about things that I might not otherwise have been thinking about. But the real way I use it, or maybe the most valuable way, it's a great editor. I'll tell it, something's not working here. I need a transition. Give me 10 different transitions. And like, it's never like I cut and paste and say use that. But it gives me a lot of good ideas and it's good at spotting mistakes. And like, oh, you have an agreement, you know, your verbs and nouns aren't agreeing here kind of thing. One other thing, I didn't list it. It's great at summarizing, which is like there's a 20 page paper like, oh my God, I don't have all afternoon, but can you summarize the key points here? And it's, it's, it's very useful. Limited, but useful.

[Elizabeth Spiers]

Yeah, that's kind of how I use it. I think of it as an administrative tool. So last week I, I needed to set up an LLC and some corporate stuff and said, give me a checklist from easiest to most difficult and then another one that says the order of operations. I should do this stuff because I would spend eight hours down rabbit holes trying to figure it out and get it wrong. But I also think about what you've just said. I think people are less forgiving with AI because they don't have any guarantee that it will learn any lessons from mistakes. Whereas with humans we kind of assume that if you fuck up really badly, you're going to internalize it and do better next time. And we don't necessarily have any evidence that AI does that without explicit training.

[Gary Rivlin]

Right, right. It's kind of deliberate that it's not learning from its mistakes. They train these models, but then they kind of freeze them in time. And so some, most nowadays are connected to the Internet. So you could ask it about something that happened in yesterday's news, but the model itself, and I think this is a good thing if it was just learning on its own, that's autonomous AI and that's where something bad can really happen. So that is one way that, that they're Largely being conservative in their approach in that. But it also means, again, they're not human. In fact, my favorite construct for AI is alien intelligence. We've never had anything like this again. It knows so much about everything, in a way, no human being on the planet. In fact, it is artificial general intelligence at all. It has such a broad range of knowledge about everything. You could ask about contracts or creating an llc, whatever. And it, you know, has strong domain knowledge but doesn't understand a thing. And I think we need to figure out how to use it. It sounds so authoritative. And you know, when someone sounds so authoritative, like, oh, we'll just trust that entity, that person in the past. Now this AI, that's not the way to use it. It's like, again, I think it's this great tool for me, for you, for writers. But that doesn't mean, oh, someone who's not a writer can then write what we write. The human element is in fact, right after ChatGPT came out, people were sharing like ways they used it. And one of my favorite examples is someone asked it explain Marx's theory in the version of a Taylor Swift song lyrics. And it was brilliant, it was funny, it did such a great job of doing it. But it's the human who had the special idea. Like, it just did what the human conceived. Some people think like, oh, no, what a filmmaker is like. You can't say like, make me a Martin Scorsese film. You, the human still have to have the plots, the characters, the twists and all that. And I think that's the way to understand it. You could be faster, but you, but it's still as an artist, as a creator, as a thinker, as a boss, you're still the one in charge.

[Elizabeth Spiers]

So why do you think people, when they think about what can I use AI for, what can I replace? They immediately go toward human creative things like writing or creating art. Why is that the first thing? It's probably the thing that AI is least capable of being helpful with.

[Gary Rivlin]

I'll confess I've had that same idea. I, I wish I was a good draw. I wish I was an artist. And you know, all you have to do is conceive the idea and, you know, one of these models can see it through. I'll steal a term from Rita Hoffman. It gives us superpowers. You know, he's never been good at languages, but in his voice he could deliver a speech or it can deliver a speech in Italian and Chinese and, you know, a multitude of languages. So I think we're all Kind of amazed by this idea of a superpower to write, to create and stuff. I mean, no one really dreams like, wow, I hope this bot could help me do my spreadsheet more quickly. I do think that is where it's going to be used, but I think that's where everyone's mind goes because, like, oh, it'll make me good at writing and I've never been good at writing kind of idea.

[Elizabeth Spiers]

Yeah, I think that's true. I used it year or two ago. I have a tattoo of a seraphim. It's very abstract. And so I used the biblical description, dumped it into a generative model just to see what it would spit out. And it spit out so many just crazy things that I wouldn't have visuals that I couldn't, my human brain could not have thought of. And so I sent a bunch of them to my tattoo artist and said, not asking you to copy this, but you know, they all look a little demonic and creepy and I like it. And so he, after looking at the generative stuff for inspiration or just an idea that would be wildly different than his preconceived notions about what this was supposed to look like, he came back with something that was incredible and did not look like any of the AI stuff, but it just took that one prompt to get him to think about it in a different way. And that's what I think of as the potential for it.

[Gary Rivlin]

And this is perfect. Again, it's, it's you. The human had this idea, use this tool, shared it with another human who didn't copy it. I mean, none of them sound usable, but it helped him do something better or something that pleased you. I, I had an interesting experience just this past weekend. My 16 year old son likes to draw and Google has this thing they're, they're promoting right now. Like take it. They said for your 7 year old, he's 16. But you know, take a drawing, take a, your kid did, and we will animate it, we will turn it into a story just based on that one static picture. And like, we're just amazed because he had written this short story for English class last year about an artist who is so into his drawings that he melts into it and loses himself. And how is he ever going to get out of his drawing? And that was the story that Gemini, Google's Gemini came up with. And like, it just like was spooky, like, wow, I guess you're a great artist that was in there. But you know, I mean, this is some of the Alchemy this is some of the miracle nature of it. Again, this whole argument, is it a good thing, is it a bad thing? Should we halt it? We can't. It's innovation. You can't stop the tides. It's coming. I just wish we'd stop having these more esoteric things. It is here. It's going to have a big impact on society. Schooling, jobs across the board, warfare, surveillance. Let's talk about it. Let's not debate whether we should pause this thing. It is here, ladies and gentlemen. Let's figure out what's the best approach. So it would be more of a positive thing and negative. I guarantee you bad things will happen with AI just like, you know, the Internet who thought like oh, social media. And that's gonna, you know, help to set us each at each other's throats because it just amps up outrage kind of thing. I mean bad things happen. But we could debate whether television, the car, you know, the Internet are good things. And I think that's the kind of debate we will be having. But let's do what we can now to ensure it's more of a positive than a negative.

[Elizabeth Spiers]

So what's your both optimistic case and your pessimistic case? And where do you put your P doom or your probability that we all just get destroyed by the AI?

[Gary Rivlin]

Yeah. So this idea of P Doom, it's not zero because that I guess is possible. An asteroid is going to hit Earth and destroy us all, you know, and so it's not zero, but it's very, very low. These things aren't even close to doing, able to do the kind of things were scared of. They're really limited. So I covered the dot com years like mid-1990s and the rise of the Internet and you know, a lot of similarities, the money being thrown at it, the hype. And of course there was a big bust and I, I kind of think more or less the same is going to happen here. I, I don't know. There's going to be a bust on the order of the dot com bus because that was a lot of publicly traded companies. These are private companies. But you know, the world is going to be disappointed. A bad thing or two is going to happen. And so then the question is, what's on the other side? You know, let's again, let's just use the Internet.comyears like okay, a lot of great ideas were thrown out there. No, you can't get rich overnight. It's going to take a long time before we have the critical Mass for social media and that kind of stuff, or, you know, E commerce. Oh, we had to figure out how you pay for credit cards and, you know, trust and all that kind of stuff. And I think that's what's going to happen with AI that, you know, it's going to be a 10 or 15 year trajectory here that, you know, as people kind of struggle with it as they improve it. Another pet peeve of mine is I respect Gary Marcus. He's really interesting. But I feel like he's someone who embodies this, like, wait, look, it's making this stupid mistake. It's terrible. Like, okay, well, it's a young technology. They're improving it every year. There's a lot of smart folks, a lot of money being thrown at this and it's going to get much better. You know, some people read Hoffman, Safa Suleiman said, you know, the hallucination rate's going to go way, way, way, way down. I don't know. Other people say that'll never be solved for, but like, again, this idea that we're unforgiving, like it's a thing that's being created. It's like judging the Internet based on 1997 website. You know, like, everything looks like Craigslist, you know, kind of back then you really couldn't do much with it, but okay, well, now of course, we see what you could do with it. I think that's going to be true with AI. I do fear that people are going to be too impressed with it. They're going to put it in charge of things. And I think that, you know, in the short and medium term, I do not want autonomous AI on anything that matters. Long term, maybe we could argue, but I still think I don't want AI in charge.

[Elizabeth Spiers]

I would agree with that. Well, thank you, Gary. This has been a great conversation and we're pleased to have you.

[Gary Rivlin]

It was a lot of fun. Thanks for having me.

[Elizabeth Spiers]

And that's it for Money Talks. Thank you to our wonderful gift guest Gary Rivlin for joining and thanks to Jesmyn Molly of Seaplane Armada for producing. We'll be back on Saturday with our regular edition of Slate Money.

[Saks Fifth Avenue Advertiser]

Slate Money is sponsored this week by Saks. Saks Fifth Avenue makes it easy to shop for your personal style this season. Fall is here and there are so many new fall arrivals that you're going to want to wear again and again. There's a great new relaxed Prada blazer. There are Gucci loafers you can take from work to the weekend. It is incredibly Easy to visit Saks.com and find new arrivals from your favorite designers. I kind of love the shirts from Comme des Garcons. I can't always afford them, but it is definitely always there on my Inspo board. And once in a blue moon I might even buy one. Saks makes shopping feel customized to you. They have in store stylists. They have Saks.com showing you only what you like to shop. They will even let you know when arrivals from your favorite designers are in or when something you love is back in stock. So find inspiration for your personal style every day at Saks Fifth Avenue.

[Sponsor/Announcer]

Slate Money is brought to you by Charles Schwab Decisions made in Washington can affect your portfolio every day, but what policy changes should investors be watching? Listen to Washington Wise, an original podcast for investors from Charles Schwab to hear the stories making news in Washington right now. Host Mike Townsend, Charles Schwab's managing director for legislative and regulatory affairs, takes a nonpartisan look at the stories that matter most to investors, including policy initiatives for retirement, savings, taxes and trade, inflation concerns, the Federal Reserve, and how regulatory developments can affect companies, sectors and even the entire market. Mike and his guests offer their perspective on how policy changes could affect what you do with your portfolio. Download the latest episode and [e-mail] WashingtonWyse or wherever you listen.