Paul Ford on AI

Summary

---

Podcast Summary: Alive with Steve Burns

Episode: "Paul Ford on AI"

Host: Steve Burns (Lemonada Media)
Guest: Paul Ford
Date: October 29, 2025

---

Episode Overview

In this lively, funny, and slightly anxious conversation, Steve Burns seeks reassurance—and clarity—about artificial intelligence from his longtime friend, technologist and author Paul Ford. Together, they wade through the buzz and buzzwords of large language models, exponential progress, and AI ethics, probing the future of technology, labor, and the reliability of the so-called "infinite answer machines" that are now everywhere. With humor and genuine curiosity, the episode explores whether our tools are outpacing our ability to wield them wisely—and what it means to be human in the age of AI.

---

Key Discussion Points & Insights

1. What Is AI? And Will It Eat Me?

• Steve’s Tech Anxiety ([00:41-04:07])

  • Steve opens up about feeling left behind by AI’s progress and asks the core question, “What is AI and will it eat me?”
  • He shares a story about smashing an MP3 player out of frustration—a metaphor for his anxieties with new technology.

• Paul’s Grounding Analogy ([04:33-06:00])

  • Paul defines a computer as “a clock with benefits,” emphasizing its binary, mechanical nature.
  • Quote (Paul, 04:40): “A computer does not love you.”
  • Historical note: The first "computers" were people doing math by hand.

• What’s a Large Language Model—Actually? ([06:04-12:51])

  • Paul illustrates that LLMs are not “thinking” but matching: They compress human knowledge, much like how a JPEG compresses an image.
  • Quote (Paul, 09:05): “It’s not thinking—it’s matching.”
  • He explains the idea of “vector space,” “tokens,” and the role of statistical prediction.

2. Trust, Bias, and AI’s Limits

• The Problem of Bias & Learning from Us ([12:51-15:45])

  • Steve raises concerns about AI learning from the "cesspool" of the internet.
  • Paul recounts early AI issues (e.g., all AI-generated CEOs being white and male) as a result of compressing biased training data.
  • Safeguards like content filters are in place at major companies, but they are not foolproof.

• Could AI Become Ultron? Paul’s Perspective ([15:45-18:02], [21:28-26:03])

  • Paul doesn’t lose sleep over “AI taking over the world,” explaining that while possible in theory, the reality is far more complex—there's no credible path from today's LLMs to artificial general intelligence (AGI).
  • Quote (Paul, 16:05): “There’s no real clear, obvious path from where we are today to artificial general intelligence… We just don’t have the tools.”

• Technological Hype vs. Reality

  • Steve expresses fear at the unpredictability of rapid progress (“We've made fire. Will it burn us down?”)
  • Paul characterizes much of the existential worry as “myth,” fueled by sci-fi and cultural narratives more than actual technical evidence.

3. The Cultural Mythology of AI

• AI Myths and Silicon Valley Narratives ([26:03-30:44])

  • Paul points out tech culture’s quasi-religious fervor for creating AGI, and contrasts disruptive “star baby” ambitions with practical productization from companies like Apple.
  • Quote (Paul, 26:41): “What you’re describing to me are myths… We needed myths to understand the change that was showing up.”

• Which Vision Wins Out? ([30:44-32:24])

  • Paul personally embraces AI as a tool for democratizing power, making previously expensive software accessible to more people.
  • He’s optimistic about the potential of personalized tools, but emphasizes: “That has nothing to do with it being intelligent.”

4. AI and the Future of Labor

• First Mile vs. Last Mile Thinking ([32:24-36:11])
  • AI excels at the “first mile”—summarizing, researching, generating ideas—but struggles with the “last mile” of human judgment, context, and creativity.
  • Quote (Paul, 33:42): “The last mile is very human.”
  • Steve observes: “Even in my casual use of AI, I can see that last 5% is where it kind of ceases to be amazing.”

5. Can We Keep Up? Regulation & Societal Adaptation

• Regulation & Safeguards ([37:40-39:43], [54:12-56:33])

  • Steve and Paul lament the lack of clear regulation and “adult supervision” in the US regarding AI. Europe is more proactive, with trade-offs in speed of progress.
  • Paul highlights the societal risk: AI adoption is outpacing our collective ability to adapt and regulate effectively.

• The Real Danger: Humans, Not Robots

  • Paul underscores that the real risk isn’t AI “waking up,” but our inability as a society to process and adapt to the changes AI brings.

6. How Should We Use AI? What Should We Know?

• Don’t Trust—Verify ([49:49-52:44])

  • Paul’s core advice is simple: “Don’t trust.” Always verify AI’s outputs—use it when you can fact-check or spot-check its suggestions.

• The Oracle Problem ([41:00-45:54])

  • Steve worries about people treating AI as an all-knowing oracle, confirming biases and deepening divides.
  • Paul cites an example where tweaks to X.com’s AI resulted in it promoting white supremacist content, showing vulnerability to manipulation.

7. Playfulness & AI as a Cognitive “Power Tool”

• Hands-On Demonstration: Prompting ChatGPT ([57:44-73:45])

  • Steve and Paul live-prompt ChatGPT to generate podcast questions and then rephrase them using only one-syllable words, demonstrating both the tool’s capability and its inherent sycophancy.
  • Using playful prompts, like “Explain anxiety management as a squirrel,” Paul illustrates how AI can be a “translation engine” for ideas—helpful, novel, and playful, but not a substitute for expert knowledge.
  • Quote (Paul, 69:14): “It really is a kind of cognitive power tool if used well.”

• AI for Play, Not Just Work

  • Paul insists that creative and playful uses of AI (“how would Yoda bake brownies?”) are legitimate, reminding listeners that new technology is also for fun and exploration.
  • Quote (Paul, 72:12): “It is absolutely what it’s for. To me, I look at that and I’m like, you could get a grumpy 10-year-old to make brownies with you…There are applications here that are really new in the universe.”

---

Notable Quotes & Memorable Moments

• “A computer is a clock with benefits. The computer does not love you.”
  — Paul Ford ([04:40])

• “It’s not thinking, it’s matching.”
  — Paul Ford ([09:05])

• “We’re living in a frontier. You are not protected. Caveat emptor. Don’t trust these things.”
  — Paul Ford ([50:14])

• “The last mile is very human.”
  — Paul Ford ([33:42])

• “If you had an infinite answer machine, what would you ask it?”
  — Steve Burns ([76:20])

• On the power and risk of AI as an oracle:
  “How we respond to this technology, emotionally, I think is what I’m most afraid of. And if we view it as an oracle… and we’re using it in our current polarized shit show to confirm our biases, Dear God.”
  — Steve Burns ([45:54])

• On guardrails and regulation:
  “There are no safeguards in place.”
  — Paul Ford ([54:34])

---

Timestamps for Important Segments

• [00:41] — Steve’s anxiety & the question “Will it eat me?”
• [04:40] — Paul’s metaphor: “A clock with benefits”
• [06:14] — Large Language Models, explained
• [09:05] — “It’s not thinking, it’s matching.”
• [12:51] — The problem of bias and “compression of humanity”
• [16:05] — Is AI world-ending? Paul’s assessment
• [26:41] — AI myths and sci-fi influences
• [32:24] — “First mile” vs. “last mile” of AI & labor
• [39:55] — Humans’ inability to adapt: the real risk
• [49:49] — Paul’s advice: “Don’t trust.”
• [57:56] — Live demonstration: prompting ChatGPT
• [63:59] — AI rephrases questions in one-syllable words
• [69:35] — ChatGPT as Yoda, giving a brownie recipe
• [73:45] — Reflections: optimism, fear, and human responsibility
• [76:08] — Steve’s closing thoughts: “It still at least depends on what we do with it.”

---

Takeaways

• AI is a tool, not a person; it matches, not thinks.
• Technology can amplify both the best and worst of human input; much depends on how we use and govern it.
• Regulatory frameworks lag far behind the pace of AI adoption—caution and self-reliance are essential.
• Playfulness, creativity, and skepticism are vital for healthy engagement with AI.
• The greatest danger is not AI “waking up,” but humans using—or misusing—the tool without reflection or oversight.

---

Episode Tone

Candid, humorous, and gently skeptical. Steve and Paul approach AI with curiosity, humility, and a bit of existential dread, balanced by genuine friendship and a willingness to play—and question everything.

---

Transcript

[Paul Ford]

Lemonada.

[Steve Burns]

Alive with Steve Burns. Hey, there you are. Come on in. It's good to see you. I have a question for you. Do you understand large language models and recursive learning? Yeah, okay, me neither. Like at all.

[Paul Ford]

Right.

[Steve Burns]

I, I, every time my laptop these days or look at my phone, I'm like, oh, cool, here's an object that is already a gajillion times smarter than me and is becoming a bajillion times smarter by the nanosecond. And I like, I like, literally struggle with Google Docs for real. I'm making tea. Do you want tea? We live in a freaky, freaky time where the machines aren't just catching up anymore, they are now sprinting ahead at an exponential pace. I'm talking, of course, about AI. It's like writing poems, it's passing the bar, it's deep, faking whole ass human beings. I mean, realistically. Here you go. How long before it starts just telling us what to see, what to hear, what to think? How long before it just swallows us whole, you know? And if I'm being honest, I don't even know what it is really. Not really. So I guess my question is, what is AI and will it eat me? Let's, let's go. Okay. Today's guest is Paul Ford. This is a guy who has been patiently, heroically explaining technology to me for the better part of 30 years. And he has done this despite overwhelming evidence that I am simply not wired to understand it. I get lost. I was focused. I asked the same questions over and over again. Without understanding. I get extremely easily frustrated. I was a very early adopter of MP3 technology and I once ceremoniously smashed an MP3 player to death with a hammer because I felt that it had betrayed me. And Paul stood by me, concerned, probably, but unsurprised. Paul is also an author. He wrote the legendary what is Code? He is a former tech advisor for the Obama White House, and he's the co founder of Postlight and now a new AI startup called Aboard. But most importantly, he is a calm, patient and brilliant friend who has walked me back from the precipice of digital despair more times than I can count. And he is here. Paul Ford. Steve, it's me. How are you?

[Paul Ford]

Can I come in later?

[Steve Burns]

Later, you can come in later, but.

[Paul Ford]

But not right now. Have I shown value at the. On the porch?

[Steve Burns]

Yes, you have to earn entrance into the clubhouse, but I'm sure that you will. It's lovely to see you. You look great, my friend.

[Paul Ford]

Thank you. My friends. Good to see you, too.

[Steve Burns]

So you know why you're here?

[Paul Ford]

We want to. We.

[Steve Burns]

We want you to assuage my fear. We want you to assuage my anxiety about. Mostly about AI and mostly about this sort of exponential curve that we are all finding ourselves in, whether we understand it or not. So if it's okay with you, I'm going to ask you some very basic tech questions, as is our want.

[Paul Ford]

Absolutely. I'm here to help.

[Steve Burns]

All right, so what is a computer, and does it love me?

[Paul Ford]

A computer is. I once wrote that a computer is a clock with benefits. It's just a thing that, like, it just turns the wheel, and every time it turns the wheel, a new little teeny, tiny thing happens. But when you turn the wheel millions and millions of times per second, those little things add up. And you can add numbers, or you can put pictures on a screen, or you can browse the web. It's just. It's a machine for doing tiny things over and over and over again. The computer does not love you.

[Steve Burns]

Ouch. It's interesting that you describe a computer like, mechanically. Weren't they mechanical at first?

[Paul Ford]

Well, they were. Well, I mean, initially, computers were people who just did math on a schedule, like in NASA. But the. No, they were mechanical. They are mechanical. It's just the mechanical part is sort of lots of little virtual switches flipping, like, literally kind of light switches, but billions of them all at once. So, you know, if you could flip a light switch, like, millions of times per second, you could be a computer.

[Steve Burns]

And how fun that would be. So it really is just. Yes, no, on, off. It really is just splitting binarily this way, that way, this way, that way, this way, that way, over and over and over again. I like the clock with benefits idea. That's from. What is code?

[Paul Ford]

Yeah, that's right. Yeah. Yeah, that's right. Well, you sort of set me up for that, which I greatly appreciate.

[Steve Burns]

Yeah, you know, you're welcome. You're welcome, Paul. Now, what is a large language model?

[Paul Ford]

Okay, this is a hard one, and I'm a professional technology explainer, and no one's nailed it yet.

[Steve Burns]

Okay.

[Paul Ford]

But I'm going to give it a go. Okay. I'm going to. I'm going to try something here. I want you to just count from one to ten. Just do that for me.

[Steve Burns]

1, 2, 3, 4, 5, 6, 7, 8, 9, 10.

[Paul Ford]

Now, and if I'd asked you to count from, like, one to a hundred thousand, it would make a bad podcast, but you could do it, right?

[Steve Burns]

I Think I could? Yeah.

[Paul Ford]

Okay. You are a computer in this moment. You have an algorithm and you're running it. And the algorithm is every time a number pops into my head, add one to that number and say the number out loud. That's what computers do. They're good at it. And they've done it for decades and decades and decades. And we kind of understand how this works now. You asked me the same question, but I will be an LLM, a large.

[Steve Burns]

You, you. You will play the part of a large language model. Okay.

[Paul Ford]

I am, I am actually physically a large language model. All right?

[Steve Burns]

So I'm prompting you right now, ask me.

[Paul Ford]

Ask me the same question.

[Steve Burns]

Call GTP. Count from 1 to 10.

[Paul Ford]

That is a great question. You did a great job asking it. When. Okay, count two. Okay, I'm going to take that count. And two are separate. And one and ten are separate. Okay, I'm going to count. I know is in number space. I know. The numbers and count go together. And the other, the 1 and 10 are numbers too. They go together. What matches when. If I was to actually produce something based on that input, I might get a five. Wait, you know what? Actually five and seven are all in that space together. We'll get those. Guys, let me go to another layer and have it take a look at. Sort what I just produced. Okay, I could do better here. I could have a six on and on and on. And then it might. Then it'll spit out. It'll sort of glue it back together at the very end and it will say, 1, 2, 3, 4, 5, 6, 7, 8, 10. Like, you know how sometimes it misses things, it's not doing the math, it doesn't execute, takes what you put in and then it enters vector space. I'm sorry, but it does. It actually. It's sort of like it enters this really high order mathematical model of all the language it's ingested in the past. And it then uses that to figure out and predict what typically might come after somebody asks for 1 to 10 to be counted off. And because so many people have asked for that in the past, it tends to have lots of strong connections between the numbers and the tokens and the words count too. And it can usually produce something pretty good and pretty accurate.

[Steve Burns]

That didn't make me feel better at all. What I heard you saying was a lot of a computer inferring things from things that we told it previously. So it's learning from our collected learning and then we learn from it, and then it learns from our learning. Like the snake Eating.

[Paul Ford]

You know how JPEG images get real blobby edges if they're really compressed?

[Steve Burns]

Kinda, yeah.

[Paul Ford]

I mean like you've seen like crappy Internet images that are really squeezed.

[Steve Burns]

Yes.

[Paul Ford]

That's what it's doing with all human knowledge. It's squeezing it down into like things that fit on one hard drive. Like everything on the web gets compressed into a couple hard, Big. Yeah.

[Steve Burns]

Oh, this makes sense to me.

[Paul Ford]

Yeah, yeah, yeah, yeah, yeah. And so when you ask it a question, it's sort of statistically decompressing all knowledge by linking what you've just asked. Those little tokens that you've put in, it splits them up and it goes. And it finds things that kind of match and then it produces the statistically most likely next response. But that is not thinking. Right. It's a database query. Yeah, yeah, yeah, yeah.

[Steve Burns]

A really good and detailed, enormously fast and complex database query. And does it work a little bit like audio compression in that sense where.

[Paul Ford]

You'Re getting, you know, I'm stretching it here. Yeah, a little bit. I mean, it's sort of like in that things are all smushed together, but they're smushed together in a different way than you might smush audio. Okay. They're organized into numbers and put into this weird database. I mean, this is merciless stuff. No one knows how to explain it. Your poor audience. Right. Like, this is hard.

[Steve Burns]

Yeah.

[Paul Ford]

But it's not thinking it's matching.

[Steve Burns]

It's not thinking it's matching. It's matching the best possible results from your prompt from your query.

[Paul Ford]

Right.

[Steve Burns]

But that sounds like a subjective. That sounds like a subjective measure. It's saying. Because the concept of best, does it mean the most likely or the most statistically, how is it deciding what best is?

[Paul Ford]

Well, there's a couple things going on. I mean, if we're going to get really nerdy, there's actually a factor called heat. Right. Like how hot, like how much? Or temperature. Sorry, temperature refers to sort of how creative and random it's allowed to be. So you can kind of get more and more consistent results out of these things by taking the temperature down. The other thing it does is it evaluates its own output using classic computer stuff. Like it's like, hey, does this look right? Using other kinds of pattern matching. Like it's not just one thing. You've got LLMs and LLMs are these kind of databases and they're many layered and they're very complicated and I'm going to be very frank, like I'm good at this stuff. I don't fully understand them. I'm researching them. But yeah, like it's, it's hard and it's, there's a reason why there are super geniuses getting lots and lots of money at these giant companies. But from there, right, you then can do all the other computer stuff that you do, like, you know, add some. You could sharpen the image or blur the image. You could check the text with a normal grammar editor. You could say, hey, if there's not a period at the end of the, go ahead and put one in. And that doesn't require LLMs. I could have done that for you 20 years ago. So it's a mix of like the things that you used to do all the time with computers and then this new weird thing, but because it talks to you and draws pictures of like squirrels as astronauts, it's so new, it's really confusing.

[Steve Burns]

Yeah, it is. And what concerns me just, just even from what you're saying, is this concept of it reflecting what we, what we've told it, what we've learned and what we've used the computer for, which makes me think that it is learning from the global interweb, which is essentially a post apocalyptic cesspool of flat earth theories and incest porn. And if, if we are, if we are doing this sort of feedback loop of learning with that and we are still the filthy input on one side, aren't we doomed? Aren't we making a misinformation golem and giving it the nuclear codes and the keys to our car and our Social Security number.

[Paul Ford]

Well, let's unpack that slightly. Right? So maybe is a reasonable answer here. So first of all, I don't know if you remember, but when AI was starting to get in the news, there was a whole lot of stuff about bias. And the reason that like, you know, you would ask it to make a CEO a picture of a CEO, and all the CEOs were white dudes who looked like they could be named Jeff. And why is that happening? It's because so many pictures of CEOs that had slurped up happen to be white guys named Jeff. And so, so, and it was like, because it does all this kind of averaging and normalization and kind of compresses things, it misses a lot of nuance. And so it just goes like, all right, well, that's what a CEO looks like. And that, that raised a lot of issues. And so a lot of flags and a lot of layers have been put in to try, especially by the Big companies to try to mitigate that. Right? So if you asked, if you go ask AI to plan a crime, perpetrate a genocide, things like that, it's going to throw up a tent. It's going to go like, I can't really do that for you. Really sorry. And, and that is because they've put these filters in, and some of the filters are probably at the LLM level, meaning that, like, they ask it a question about, you know, could this be used to harm someone? And it, and it goes into its big blurry brain and it goes like, yes, that actually sounds pretty harmful. And then they pull that out, they flag it and they say, okay, let's stop there, let's not keep answering this question. So you've got this world where people were really raising these flags. They're still not happy. The people who are raising the bias flags, they're like, I don't think we've gone far here. Like, we're slurping up the random Internet. We're taking in all the biases, we're taking in all the harms. You've also got a lot of good. You've got Wikipedia, you've got a lot of books. Maybe they pirated the books, maybe they paid for the books. Who knows? Not clear. You've got YouTube videos on and on. But yeah, it's slurping everything up and it's trying to make sense of it.

[Steve Burns]

If we didn't have that little governor that people put on that to keep it from acting poorly, could it reflect our darkest angels and become Ultron and become a genocidal megalomaniac super computer that wants us gone or something? Make me feel better, Paul. Make me feel better.

[Paul Ford]

No, I, I. Let me tell you something. I don't stay awake about this, okay? I'm in this world. I've been in this world on and off, and I've been paying attention to this world for 25 plus years. I used to do AI work when I first knew you, like in my 20s. And this idea that the computer will become intelligent and take over the world is not a new one. It's been going around for decades. It's been in science fiction, it's sort of been all kinds of places. And what I can tell you if you said, hey, could we eventually make a simulated brain, hook it up to all sorts of things and have it go to town and it might get out of control? I'm like, that is within the realm of possibility. Just as we can create humans who go out of control. Like Things are physically embodied. Who knows? But if you ask me, what is the path from the smushy databases to the superintelligence in charge of everything, I don't have an answer for you. And no one else has one that is absolutely credible because they'd be predicting the future. And so there's like, there's no real clear, obvious path from where we are today to what is called artificial general intelligence. People are on multiple paths, saying they think they're going to get there. There's a lot of venture and a lot of marketing associated with this story, but I don't now, you know, this is going to come out a little bit after we record it. So who knows? We could all be sheltering down as Skynet, you know, lasers us. But I don't see the clear path. And we're decades in and the path, it turns out that humans are still very complicated and that we're kind of here accidentally, like evolution has kind of left us here. This would be something very different. It would be very intentional that we would create it. And so we're not there yet. We just don't have the tools.

[Steve Burns]

Okay, so wouldn't it be, like, great if you could find stuff that you're looking for right away, like a parking spot or a pair of jeans that actually fit a short king or the beginning of the roll of clear tape, that thing? Well, while you may never instantly find these things, if you're hiring, you can find qualified candidates right away time and time again with ZipRecruiter. And today you can try it for free at ZipRecruiter.com alive. ZipRecruiter's got that powerful matching technology that works super fast to help you find top talent. It's really no wonder ZipRecruiter is rated number one hiring site based on G2. Makes sense. So if you want to know right away how many qualified candidates are in your area, look no further than ZipRecruiter. Four out of five employers who post on ZipRecruiter get a quality candidate within the first day. And right now, you can try it for [e-mail] alive. Again, that ziprecruiter.com alive. ZipRecruiter. The smartest way to hire. So it's getting cold. Air is getting kind of windy and dry and brittle. And I heat my house mostly with this thing right here. And that really dries the air out, too. And I. I do notice it on my skin most of the time. It tends to, like, dry out. And get cracky and stuff. But not anymore, really, because I started using One Skin. This. This stuff right here. OS One Face, the essential moisturizer. I believe this is their most popular product. And I gotta say, it works, right? I've been using it for a couple weeks now, and I think my skin looks pretty good. I'm feeling a little bit like a snack baby girl. Yeah. And I've also tried. This is the other one. I like OS One Eye topical supplement because I get these bags under my eyes because I'm Italian. And this stuff also really seems to help. Their products are awesome. And they're backed by five published clinical studies. And they have over 10,000 five star reviews. I would like to have 10,000 five star reviews. It's all about their OS1 patented peptide. It's the first ingredient to target senescent cells, which I didn't know what those were, but they are, in fact the root cause of wrinkles and loss of elasticity.

[Podcast Advertisers/Guests (e.g., Julia Louis-Dreyfus, Chelsea Clinton, Sarah Amos)]

Right.

[Steve Burns]

So anything I can do to fight back at age 52, I will. So this season, don't just moisturize. Support your skin at the cellular level with one skill skin. For a limited time, you can try one skin with 15% off using code ALIVE at OneSkin co. That's 15% off OneSkin co with Code Alive. And after your purchase, they will ask you where you heard about them. Please support our show and tell them we sent you. Try One Skin today.

[Paul Ford]

Okay.

[Steve Burns]

So you don't think the hype is real right now?

[Paul Ford]

You know, I think that an amazing and wonderful new technology popped up that no one ever expected.

[Steve Burns]

See, that's what frightens me, man. You know, it's this idea that it's a little conflagration, you know, and everyone's like, ooh, we've made fire. Look at us. We've made fire. And will it burn us down? You know, it seems like a thing that could spin out of control. I mean, you said earlier, could it have all of this nefarious will one day and become a genocidal super robot that wants to destroy the world? You said, yeah, it kind of could. Just like a person kind of could. But a person couldn't contain all the collective wisdom of all of humanity at the same time. While it wants to do that, a person couldn't think that quickly. A person wouldn't have these supernatural seeming abilities that an AI would.

[Paul Ford]

There are many people who are really, really worried about this. They tend to be on the younger side and live in San Francisco. Yes. I just can't make myself worry about it. Okay. I get where you're coming from. I get why you've heard about this, Everybody has. But when I look at where we are and what computers can do today and where we would need to get and the safeguards that would need to fail in order for that to happen and unlock, and the kind of agency that the machine would need and the kind of guardrails that would. It would have to be something really intentional, right? It would have to be like somebody saying, unleash the hounds. And the hounds are, I have an enormous set of computers all ready to go, and we're going to hack into all the systems and we're going to, like, really mess up the world. And you're in a kind of like, bad Mission Impossible territory at this point. Like, it's just. There's reasons that that isn't real. It's not. It's. It makes a very clear story, but it just doesn't line up with how the world works, where people can get resources and what can come together. So I just, like, there's all. I mean, you know, literally the worst people in the world have nuclear weapons right now. And that is probably scarier, but we're not as scared about that because it doesn't have the sort of creep factor of the computer waking up one and going, you know, I see you, Steve, you know, and that if, you know, it calls you, it's like, what are you doing? And you don't need to tell it because it already knows. It's watching through the security camera. That is just. That's really great material, right? And it really locks in and it's scary. And like, anything you're in the zone of, like, the truly possible nuclear war is possible. The super pandemic is possible. Scary things are real. And we've lived through scary things in the last decade at probably a relatively high velocity. I think you'd have to kind of go back towards World War II to get a really big scary things at that level, right? And big scary things continue to happen, but again, like, they can happen, but the steps aren't there. Like, nobody really knows how to get there. So it's cool that they're saying, hey, let's be careful. It's cool that. But a lot of the dialogue about the super intelligent future also ties in with, like, how do we stop human extinction? We have to become star citizens. Like, there's a whole kind of like, deep cultural ethos that, you know, then ties into, like, crypto and like, all that stuff. It sort of all fits in a piece of like the super digital human is coming. Get ready. And it's almost a religion. And like any religion, it has a vision of heaven and hell and sort of like kind of what you're getting downstream from a lot of this stuff is them saying we have to really all get religion or hell will be real. And religion in this case is let's fund our genius AI people and give them the money to build the super intelligent computer so that they can control it because we trust them. That's the vision here. And you're getting this thing like seven levels removed about spooky AIs that are going to take, take all your money and send it to Mars, right? Yeah.

[Steve Burns]

You know, I, I feel coded in a way to fear AI. You know, it's. Yes, we, we've. This has been foretold over and over again, especially in, in our childhoods, right? This has been foretold. We watched countless movies about this. And the lesson is always the same, right? The lesson is always be careful what you build, you know? And it feels to me as a passive consumer of all of this, that, oh shit, we did it. We built the monster.

[Paul Ford]

We didn't. Not yet. What you're describing to me are myths because strange new entities that, you know, literally at that point, kids are in, they're playing Missile Command and Pac man and they've gone somewhere new and no one's ever seen them go there before in the arcade. And there are a lot of movies and myths about, you know, video game players. Like the Last Starfighter, where he plays the game and then he gets recruited to join the intern. Yeah, great, great, great 80s cheeseball kid. Like Star wars ripoff, right? And so we're watching these things because we needed myths to understand the change that was getting, that was showing up. Like your bank account is this number and you don't go to the teller anymore. You go to a machine and the machine says, put stuff in. And now it's an ATM and you don't need humans. And that's weird, right? You're living in the 80s and that stuff is strange. And you need stories that people can understand. And they're sort of. And the stories, they're Hollywood stories, they get wacky. And so you are downstream of that. And then you are seeing this new thing that shows up and the interface to it. It was a blobby database. This technology has been around for decades, frankly. And then around2014 or so, people were like, hey, if we do it this way we get really interesting results. We can use them for all sorts of things. And then in particular, Sam Altman and OpenAI strapped chat to it. They're like, what if we use chat as the interface to this GPT thing? And that's ChatGPT. And version three started kind of chatting because it was doing the things we described before, the statistical stuff, and boom, it seems human. And I think everybody saw that. And you got to understand the incentive structure here. Everyone is really excited by this. They've never seen it before. It's a novel new technology doing things no one has seen. There is money pouring in. Altman is the single best connected person in Silicon Valley. He seems like a relatively young, sort of like kind of sweet, goofy guy, but he's at the absolute center of all of venture capital, all of that world, and he has the best story to tell. And in the end, he's always been very motivated by creating, creating artificial intelligence. He sees it as his destiny. And so you put all those things together and you think about the stories he starts to tell himself about why he's here, what he should do next, what it all means. Okay, at the same time, you've got, I'm going to give you this is, we're going to go in the weeds for 10 seconds. But if you look at what the big AI companies are doing that are talking a lot about artificial intelligence and artificial general intelligence and then you contrast that with what the giant platform multitrillion dollar companies are doing, like Apple and Google and also Micro Apple and Google and Microsoft are saying this is a product. And Apple in particular, this is going to kind of live on your phone. We'll have a server version, it'll be really easy to build with. You can have lots of apps with AI involved. They're not waiting, they're not saying, we better don't even bother because the artificial intelligence is coming. The multi trillion dollar Tim Cook led. Apple is saying this is a cool tool and we're going to bring it into our giant world where it will be one of many things that you use to do computer stuff. And we'll clean it up, we'll make it tid. We're going to only index wholesome things. Ethically, they say this and it's going to be nice for you. It's going to be, you're not going to have all this scary stuff. You'll just be able to use this as a tool. Right? And so, okay, contrast that to the people saying that this, you know, the star baby is being Born on Nvidia server racks. Which one do you want to believe? The giant capitalist platform company that is making choices for its shareholders? Or do you want to believe the pseudo religious but very, very success LLM builders over here? It's a choice. You don't have to choose one or the other. You could choose neither. You could say, this is all really bad and we should put it back in the box.

[Steve Burns]

What do you choose? What do you. Where. Where do you fall on that? Where do you end?

[Paul Ford]

I'll tell you what I love about this software. I love software. I love it. I'm a big nerd. You've known this about me forever.

[Steve Burns]

Yes.

[Paul Ford]

You know, I used to carry around programming books even into my 20s.

[Steve Burns]

Yeah, it was bad.

[Paul Ford]

It was bad. It was. It was rough. Took me a long time to get married, but. But you did great. I. Everything worked out just fine. But it might have taken me a little bit longer because I loved the computer, too. And so basically, I'm at a point where I can do something for somebody at 20% of the cost that it used to be much, much, much, much faster. Because AI can do so much of the, like, Lego assembly work and it can get stuff on the screen that works a lot faster. And so there's a lot of talk about jobs going away, but I think there's this other future where everybody gets their thing. If you sit down and you go, I need a better tool for managing the questions that I'm asking these guests on my podcast. You can have that. You need a. I want to track dog walking. I want to. All these things that, like software, you don't only have to. You have to, like, go find a specific app, and if it didn't exist, you were out of luck. I think that's coming. And I think sort of like specific tools and ways that people can have more power in their life can be a huge side effect of this weird new technology. They can talk to it and they can get what they want. And that is really, really exciting to me. That has nothing to do with it being intelligent. It is about taking power away from the really big old companies and giving it to people using this new framework and hoping it works out for the best.

[Steve Burns]

That's encouraging. Do you see a future where it does kind of, in a large and meaningful way, replace labor, like, replace a lot of what we do. And if so, what is your plan? What are you going to do with all your free time?

[Paul Ford]

So here's what I've noticed. Now this stuff gets better every month. Every couple of months there's something new and it does more. But. But I use it to do all kinds of stuff. I use it for research. I use it for all kinds of things. I use it for coding. I use it for a million different applications. It's really, really good at the first mile. It's really great at the quick summary, at the research project, at generating a bunch of images, and it's really good at that because it seems human. And so it's really nice to have this humanity simulator that's like, I will go out and I will research that. That microphone for you and make sure it's the right one for podcasting. Yeah, okay. But the last mile is very human. You've got four people working on this podcast right now. We can't all see them, but they're making this thing come together and they're thinking about. And then there's more people at the podcast company. They're thinking about marketing and distribution, and there's people out there and they have to. They have to get this thing on schedule and. And, you know, like, how are the microphones working? And what are we going to do about quality? And who's the next guest? Those are all last mile tasks that lead to this thing getting done and into people's houses. AI can help with the sort of, like, coordination parts, but it's not good at that assembly because it's not human. It doesn't know what the final product is. It's really bad a lot of times at debugging code. Sometimes it's great at it, but it doesn't know what the code is really supposed to be because it can only do those weird queries. And so you're kind of in this. It keeps surprising you because it seems so human, and then you get kind of used to it, and then you sort of get bored because it does the same thing over and over. Now, if the same thing happens to be. I really need to have a summary of what the news is about this given subject every day across five newspapers. It's great. Really, really works well, and it can even kind of draw some conclusions. It can be useful. But that last bit where you're doing the thinking or you're making the decision, it can come up with ideas, it can be weird and seem kind of creative, but that last mile is still there. So if your labor is with that first mile, if you are a professional news summarizer, that's not great. Just not great right now. Because whether for better or for worse, no matter how much Craft you have, everybody is saying, I could do without that. I can just do that with a bot. But if you are closer to the last mile, you're much harder to replace, even if you're getting coffee for somebody. So, like, if I was, like, if I was a professional podcast summarizer, not great. I can eat the transcripts and summarize. I would try to get as close to your house as I could. Even if I was, like, getting coffee and doing basic stuff, because that's where the work and labor is going to be in the future, I can learn how the last mile works. That would be the advice I would give. Look for where things actually go into the world and stand as close to that as possible. You don't want to be one of like 30 fund analysts at a big bank.

[Steve Burns]

Right. I hear what you're saying. And even in my casual use of AI, I can see that last 5% is where it kind of ceases to be amazing. There is something missing. There's no ghost in the machine at the end. Right. And I get that, but isn't that just a question of yet?

[Paul Ford]

Well, look, every time there is a really big new release, it gets further and further. Like, the deep research stuff is extraordinary. But I'll tell you what, Steve, it's like, you know, I'm looking at this stuff, it becomes more like programming to me every couple of months in that. Okay, it does reasoning now, but all reasoning is, as far as I can tell, is it goes, you give it a prompt and then it goes, and it kind of makes a whole nother blobby prompt. Instead of going straight to the answer, it tells itself to ask more questions and it introduces ambiguity. And then it sends that back to itself and it resolves some of the ambiguity. And it does that a couple of times. And as a result, you get something that looks like it was reasoning, but it's actually just a circular set of queries where it keeps telling itself to broaden the topic a little bit.

[Steve Burns]

Well, but I mean. Okay, but what if that's not thinking?

[Paul Ford]

You don't do that.

[Steve Burns]

No, I know, I know. I get that it's not thinking the way that we are thinking, but we think it's thinking the way that we are thinking. And then it just becomes about how we're using it and how much we trust this as actual thought. And we can misuse, from what you're saying, we can misuse this tool.

[Paul Ford]

Yeah, and it's coming at that, I'm going to be frank, like, it's coming At a really rough time. Because this is not an environment where people are having open, frank discussions in the Senate about how to regulate. It is an absolute risk. Of course.

[Steve Burns]

Who has their eyes on this? Who, who, who is dealing with this? Who's putting these measures and safeguards in place? Is it you? Is it you, Paul?

[Paul Ford]

No, no, no. I don't have the power. Europe does a lot. But it also slows them down. Their progress in the, in the LLM space is slower than the US we are people. It's very interesting because, like, in general, OpenAI is relatively anti regulation. It used to be pro regulation, the Anthropic, which makes Claudia far more pro regulation. Kind of. They're the number two. But the CEO of Anthropic is still like a true believer, but he's way more likely to be like, hey, we gotta actually have some conversations about this stuff. It's a lot trickier than anyone expected. So there are different voices, but there is definitely a sense of like, no adults in charge in our country right now. Right. And this is yet another thing that's showed up that needs adult supervision. And who knows? There's a lot of books about it, there's a lot of discussion about it, a lot of stories, a lot of things coming up. But I don't know when the rubber hits the road. And there is another reason for concern. This is the fastest growing product I think probably ever. Like hundreds of millions of people in very little time. It grew faster than social media. It's just amazing what it does. And everybody loves it. And students are using it and not doing anything. They can't believe they have to actually write a term paper. They won't do it. There's sort of like, it is a pretty big cultural slap across the face. All of those are real concerns, but those are about people not being able to get it together.

[Steve Burns]

Yes.

[Paul Ford]

Over time, we often get it together. That is different than a robot taking over. People not getting it together is a different problem that we do tend to solve over the horizon of 50 to 200 years.

[Steve Burns]

Okay. So our learning curve for AI is nowhere. Is not even gonna come close to matching its exponential singularity style flip out. That it's about to do. Okay, I'm getting scared.

[Paul Ford]

You just nailed the real issue. Yeah. It's not that it's going to take over. It's that we can't adapt. We just don't have it.

[Steve Burns]

No. And I worry about how we're gonna use this. I'm hearing what you're saying, honestly, about how this is not exactly thinking. This is not. This is not sentience. Right. This is a different thing. But that's not the way. You know, no one's. No one who's using almost no one who's using this technology is thinking about it that way. They're saying, what can it do for me? How can it. How can I drive through chat, GTP McDonald's and arrive on the other side with valuable information, if not a worldview and advice? And it's not just going to be. Be facts. It's going to be, how can I. How can you resolve the problems of my spirit? How can you resolve my anxiety? How can you. We're going to look at this like an oracle eventually.

[Paul Ford]

It's getting very dangerous that way because there are people who, you know, if they have a sort of emotional breakdown or disorder and they ask it to, like, you know, can you please explain the connection between the constellation Orion and why my mother barks like a dog? Like, it will do it. It'll just, like it'll tell them the story they want to hear about how they are chosen by God or things like that. Yes. Because it's a mirror, Right? It's a mirror of all human consciousness as it was expressed through text and video transcripts, slurped into a smushy database, and then it gets pulled back out and turned into something that looks like the smartest wizard ever. We are not ready. But we are here and we have to learn. So you're having this conversation with me. I've spent the last. I stopped my career, we stopped our product, and we spent four or five months, my partner and I and our head of engineering, just working every day for hours and with. With Claude and chat cbt, until we understood how we were going to react to this change. It changed everything about our industry. And that's coming for everybody. And they don't have the time and they don't have the background, and I really feel for them. And as you saw, it's hard to explain. There's so many moving pieces and there are very few people who are kind of on the side of the regular knucklehead out there. Like, God bless. I don't know what it's like. Imagine what it's like when somebody tries to have this conversation on Joe Rogan Hogan.

[Julia Louis-Dreyfus]

Well, hi, everybody. It's Julia Louis Dreyfus from the Wiser Than Me podcast. And I'm not going to talk about food waste this time. I'm going to talk about food resources, all that uneaten Food rotting in the landfill. It could be enriching our soil or feeding our chickens because is it's still food. And the easiest and frankly way coolest way to put all its nutrients to work is with the mill food recycler. It looks like an art house garbage can. You can just toss your scraps in it like a garbage can. But it is definitely not a garbage can. I mean, it's true I'm pretty obsessed with this thing. I even invested in this thing. But I'm not alone. Any mill owner just might corner you at a party and rhapsodize about how it's completely odorless and it's fully automated and how you can keep filling it for weeks. But the clincher is that you can depend on it for years. Mill is a serious machine. Think about a dishwasher, not a toaster. It's built by hand in North America and it's engineered by the guy who did your iPhone. But you have to kind of live with mill to understand all the love. That's why they offer a risk free trial. Go to mill.com wiser for an exclusive offer.

[Podcast Advertisers/Guests (e.g., Julia Louis-Dreyfus, Chelsea Clinton, Sarah Amos)]

Is it just me or are things actually really scary right now? In the world of public health, every day brings another confusing headline or yet again a far fetched claim. Vaccines are somehow up for debate and parents are scrolling TikTok for medical advice. I'm Chelsea Clinton, an advocate, author, investor, teacher and mom navigating this insane time right alongside you. I hope you'll join me on my new podcast, that Can't Be True, a show that sorts fact from fiction, especially on issues impacting our health. From limitada Media and the Clinton foundation, that can't be True is out October 2nd.

[Steve Burns]

You said some comforting things today and I appreciate those. And you said some things that scare you. The absolute shite from my body. And when you just said it will simply confirm your bias for you, it can it. That is the moment. Because I understand everything you're saying about its sentience and about how that is different. You know, I, or at least I feel like I understand that enough. But that has nothing to do with how we're going to use it. How we respond to this technology, emotionally, I think is what I'm most afraid of. And if we view it as an oracle with a superhuman, intergalactic, supernatural, nearly divine level of information and trust it as such, and we're using it in our current polarized shit show to confirm our biases, Dear God.

[Paul Ford]

Well, I don't know. And it's also who's in charge. Right. So I don't know if you saw this, but some time ago, they tweaked Grok, which is the. Yeah. X.com AI engine. And they over tweaked it. They changed the core prompt and they got it all confused. I'm being metaphorical. It doesn't think or have a personality. It's just following instructions that people put in. And it just sort of responded to everything with like, hey, there's a lot of real valid questions about the white genocide. And just sort of like, it got really into like, weird white supremacist stuff and then we're right about it. Yeah. And it was bad stuff. It was bad stuff. So what you realize is, like, we're counting on a lot. We're counting on these big orgs to kind of like have good, healthy filters. I think we know that Elon Musk is really going in a lot of different directions. And so I wouldn't use that product to build something because it's too sketchy when they can do stuff like that. But, yeah, like, no one. I hate to say it because the changes are coming now, but it's decades, maybe longer, before we can fully metabolize this change. And what everybody's saying is like, no, actually the robot will become intelligent before that. You don't have to worry about it. Which always reminds me of people, literally a lot of fundamentalist Christians don't say, say like, don't worry about global warming because Christ is coming back. And it's sort of like, like, no, I. I know. Look, I'm. I'm here with you, but we're. We're in this together. You know, like, it's. Thank you, Paul. I want to. We're in this together. I do want to say that, like, I and my partner, we've built a tool that will help people have their software easily. Not for zero, not for a credit card, but like, you know, like, and that we can do that. I can give more people their thing than before. And I'm, I'm. There are not for profits and orgs that I'm really excited because they'll be able to do more. And so, like, there are ways to use this for tremendous good. We're using it for climate change adaptation because it can analyze. It can tell you how to protect your warehouse from climate change if you feed it good, clean data, write you a nice report, and build you software to manage the adaptation so that your warehouse employees are less at risk of heat and flooding. And, like, that's my Wife is working on that with us right now. So, like there are absolutely positive applications that are much, much more. And like, you're gonna have to choose. You have to choose where you're gonna go. Are you gonna let the robot tell you that certain races are inferior or are you gonna use it to help the warehouse workers get ready in case it floods?

[Steve Burns]

See and go ahead. Sorry.

[Paul Ford]

No, that's on us. That's on us. Unfortunately, and us is not doing so good right now.

[Steve Burns]

Yeah, I mean, I believe you. I trust you. I know you. I know you to be an altruistic, kind hearted Paul. You know it's my brand. Yes, yes. And I'm sure that a board is doing that, right?

[Paul Ford]

Sure.

[Steve Burns]

But when you say it's a choice, not everyone's going to choose to use it that way. I guess my question to you is how much do we need to know? Me, right? Remember? Do you remember the Diamond Rio MP3 player story that I referenced?

[Paul Ford]

No. You smashed that thing.

[Steve Burns]

I did. Because it betrayed me, Paul. It didn't work and it lied and it didn't work. It said it would work and it didn't work. And I became so frustrated with it because I couldn't understand how to make it work. But I wanted that technology in my life.

[Paul Ford]

You know, you also had some undiagnosed issues at that time. So it's a little more complicated than you just spiraling into rage and smashing an MP3 player. But you did do that. That is true. I remember it quite well.

[Steve Burns]

But my point is someone like who is sort of really desperate almost to incorporate cool new technology into their lives, how much does an ordinary person need to understand about large language models or AI in order to use it safely?

[Paul Ford]

Unfortunately, right now, a lot. And so let's reduce it to one simple principle, which is don't trust. And what I mean is only use it for things where you can verify, where you can figure out what's really going on, where you can at least do some spot checks. It will continue to make things up and hallucinate because it doesn't know what's real or not. It'll make up websites and URLs and stories. It's better than it used to be. It has limits. You have to learn what those limits are. This is hard. I don't think anyone, no one expected this level of adoption this quickly. The product and sort of by the rules of the computer industry, it doesn't really work. Like stuff breaks and stuff doesn't make sense and so on. If Windows used To crash. That was a big deal, and they had to fix it. And here it's like, they just release stuff and it does whatever the hell it wants and it tells you things. And they're like, yeah, why isn't it so? You are living in a frontier. You are not protected. Caveat emptor. And just don't trust. Don't trust these things. Don't trust them with your health. They actually don't deliver. They seem to deliver really good medical advice. Don't trust them with your nutrition. Don't trust them to guide you on relationships and so on. Let them give you ideas that you then go and explore. And I know how much everyone would rather lie on the sofa with a cognitive remote control and let this thing just spoon stuff into your brain. I've done it too. I've done it too. But it doesn't really work. And you will be one of the weird, confused chatgpt brains out there, as opposed to someone who can make good, clear, informed decisions. You know, like, don't ask it about your birth control. Don't ask it about what to wear. Like. Or do. But then go fricking find the source. Follow the source. And it's so powerful because it seems human. And it also, you. You are the human, though. And when you give that up, when you say, okay, it must be right, because it showed up in a computer screen and it had words. You're really giving up something really fundamental and you shouldn't. That will make you a lesser person. So don't do it.

[Steve Burns]

Okay. I mean. I mean, I feel better about the technology itself. I think I feel worse about our capacity to assign all of the attendant dysfunctions of consciousness that we have to it.

[Paul Ford]

Steve, we can't get a break, man. Like, we, you know, it's just as we, like, okay. Social media, boy, that's a lot. Whoa. It's really messing up the globe. Hey, hold on a minute. Slow down there, Trump. Trump. Oh, my God. Okay, well, okay, now we gotta back up. Nope. Trump again. Like, just Facebook, Facebook, Facebook X, X, X. Like, you can't get done. Okay, but now we're starting to figure it out. We're starting to figure it out. Nope. AI like, you just can't get cognitively. We need a break. We need, like, 10 years off, and there's no chance we're more likely to get war than a break. And so we gotta figure out how to be resilient in the face of this much change. We have to figure out how to listen. To listen and learn and Understand what? And then you gotta kinda ride the wave. And I tell you, I like a lot of change. I made it my whole career. I like weirdness. I like the strange things that humans get up to. I have a pretty cynical view of the world, but I'm also very affectionate towards this. And boy, this is like getting your head hit with a hammer on a regular basis. It really is. It's a lot for me. So I feel for everybody on this. I wish I could find an easy way out for you to bring this level of change in and just kind of live your life.

[Steve Burns]

I'll just ask one more question to assuage anxiety and then maybe we'll do something fun real quick and I'll let you go. But you mentioned earlier safeguards. Can you talk a little bit more about the safeguards that are in place?

[Paul Ford]

There are no safeguards in place.

[Steve Burns]

You said safeguards. I wrote it down.

[Paul Ford]

There are safeguards. They are really trying. But the reality is it's what you build up from. These are big platforms and you kind of build on top of them. They're more like foundations. And so. And there's also the reality that people have made LLMs that are open sourced and are less sort of on guardrails than the big ones and people are using them. You know, that's where kind of more of the deep fake porn and sort of those things are happening. And then, you know, there's the deep sea in China, which is very carefully. It doesn't want you to talk about Tiananmen Square, but it'll let you talk about lots of other stuff. And so everybody's trying to wrestle this to the ground. And you'll notice like Deepseek doesn't mess around on that topic because they would get in big trouble. So there are ways to put things on rails. You can also make them do bad things by tweaking them and messing with the prompts. Like it is just sort of like it's a spongy technology because it's a muddy database filled with compressed human weirdness. And trying to make it beh is like trying to make every book in the library behave. It just doesn't work. And so that's what you're trying to regulate. We don't have inspired governance in our world right now, in my opinion. And you gotta. Your hope here is that the mass of humanity actually doesn't want an infinite porn spewing monster machine that fills their brain with the worst parts of their id, but instead would like to summarize some stuff, Do a little research, learn some things and make some code and build some stuff. There's plenty of reasons for pessimism and there are some reasons for optimism. We didn't do so great with social media, in my opinion.

[Steve Burns]

That was going to be what I.

[Paul Ford]

Was going to say. No, I think we should all have the bad tum tum feel. I think we should. Should. It is real like it is. And it, it's even if the people running this stuff have the absolute best intentions aligned with your exact values in a perfect way, there's still so much going on and it's really hard to, to keep it under control. But again, like this is the thing, like the fantasy is that the robot comes alive like Ultron and it's just like, I'm angry now. But the reality is that humans are reflecting and consuming and then acting by the hundreds of million and that leads to weirdness in our world. And that's where the real change comes from. And it because all of a sudden your brother in law thinks that wizards from the sky are telling him that he can, you know, he should vote for a lizard and just sort of like. And he's telling you this at Thanksgiving because he got it from ChatGPT. And so like that is change. That's the change we can't handle. It's not that there's suddenly this new robot that does funny tricks, it's that the people around us change and we don't have tools for talking to them or negotiating it.

[Steve Burns]

That's what I'm afraid of. Okay, Paul, thank you. Before you go, I thought it would be fun if we live prompted chat.

[Paul Ford]

GT oh, let's do it. Watching you use a computer brings me back to the entirety of my twenties.

[Steve Burns]

Hopefully I will not smash this with a hammer. But I have opened my laptop and I have logged on to chat gtp.com and it is immediately very polite and it is saying, what can I help you with? Well, let's see. I'm going to ask. Ask it. You, you tell me how I would prompt this. I want to ask it a very good question to ask you in in a podcast. So how should I prompt it?

[Paul Ford]

Oh, I am creating a consumer level broad audience podcast and I have the technology expert Paul Ford visiting to talk about AI and the future future. I'd like to ask five accessible questions that will make me look really smart to both Paul and my general audience, period.

[Steve Burns]

And I'm typing all that in.

[Paul Ford]

Paul Ford here.

[Steve Burns]

All right, let's see what Chat GTP has To say, I've entered that in. It's chewing on it. It's telling me that it loves this idea. Idea. It's very nice. It's very kind.

[Paul Ford]

It is. It actually got too sycophantic some time ago and they had to wind it back because somebody pointed out, somebody asked it to please help it market their poop on a stick idea. And it was like, this is the greatest idea ever. Oh, my God, you have done it. It's the most counterintuitive product ever.

[Steve Burns]

Right, okay. Honestly, right here, I see a problem. I see a problem with the poop on a stick thing. Right, right there.

[Paul Ford]

Well, they wound that back. They wounded. They're like, whoops, sorry, we went too far. But.

[Steve Burns]

But it immediately disingenuously started kissing my ass.

[Paul Ford]

So if, if I know, but that's good. That's good product right there. That means you're going to come back.

[Steve Burns]

It's bad product. If I assume that this is an oracle, it's bad product.

[Paul Ford]

The, the worst thing about these things is that they keep simulating humans. If they could just be a little more body and like a little more like, you know, click the icon, I think it would really help people. But they talk like people.

[Steve Burns]

I think you're right. All right, here are five questions that should impress both Paul and your listeners. Says chat.

[Paul Ford]

Okay, let's see.

[Steve Burns]

I'm going to use an NPR voice for this as I read them. Paul, do you think we're. It's already good. Paul, do you think we're mistaking fluency for understanding when it comes to AI just because it can generate language so smooth, smoothly, does that fool us into thinking it knows something or even that it's thinking? I basically ask you these questions, but, but, but just not as well.

[Paul Ford]

No, it actually really sounded pretty smart when it did it.

[Steve Burns]

Yeah.

[Paul Ford]

Yeah. Okay.

[Steve Burns]

But I feel like we covered that one. Is AI accelerate. I'm sorry? Is AI accelerating our culture's existing problems with. With certainty making us more confident in answers that are actually just well, packaged guesses? That is my exact question to you, actually. Thank you, Chat. That is kind of what I've been trying to get to.

[Paul Ford]

Yeah.

[Steve Burns]

So the.

[Paul Ford]

No, I think we. The answer is yes.

[Steve Burns]

Yeah. All right. So it's, it's done a pretty good job, I would say. Tell us what it did to generate these questions.

[Paul Ford]

I mean, we talked about it earlier. Like, what I'm trying to do is make sure I get all the words in there so that when it parses and tokenizes and breaks Apart. It breaks it apart into little pieces. It's got words that will sort of align with particular outcomes. So broad audience, right? Like not. I wanted to. I emphasize that. You emphasize the things and you might even say them a couple different times. You're kind of loading it up. You're programming it with the kind of response you want. I want to look smart. Yeah, I want. So it's. And then it goes in its little brain, it goes, oh, smart. Like, what are things that are about AI connected to smart? What are things that like. And it's sort of like traversing that database and pulling stuff out. You could also ask it to, you know, do it in the most sort of Cletus style stupid way possible. And it'll do that too.

[Steve Burns]

I mean, go for it.

[Paul Ford]

That was obviously literally all you have to say now is can you, can you ask the same questions, but only in one syllable words?

[Steve Burns]

Great. This is where I was hoping we were going to go, Paul. Okay, so now I am asking. Ask the same questions using only one syllable. This is great. First of all, it's kissing my butt again, which I don't like. I don't like all of this unnecessary unsolicited affirmation which if I'm treating this as an oracle, will I will interpret as being on the right path or that I have. That I truly have a great idea. Right.

[Paul Ford]

Well, you know what you wanted to do, you wanted to like. I would prefer if it would sort of show me kind of. Here's what I did. I think this will be useful to you for these three reasons. Here's the response, right? Like, I'd love a little dashboard as opposed to this endless stream of conversations.

[Steve Burns]

Yeah, yeah, I'm already seeing a problem with that. All right, you want it to be.

[Paul Ford]

More comprehensive, computery, Go ahead.

[Steve Burns]

I think I do. I think I want it to be more of a tool and less than a friend anyway, so what it's saying is. Ooh, it wrote that. What a fun and brain bendy challenge. Let's rework those same big ideas into clean one syllable word questions. Here we go. When AI talks smooth, do we think it knows more than it does?

[Paul Ford]

That? Not bad.

[Steve Burns]

We want AI to match our core. Good. But do we all share the same one decent.

[Paul Ford]

Okay.

[Steve Burns]

If flaws make us, what do we lose when we build tools that can't slip up? Oh, I see what it's asking. All right, well, that's. That, that, that was really it just doing sort of like a, a, like a grammar puzzle or something. Let's let's ask. Let's use it wrong. Let's, let's, let's, let's do something dumb. Let's, let's have it make a TV show or do. I don't know. Let's do, let's use AI. Wrong, Paul.

[Paul Ford]

Well, there is no wrong. How can you query a database incorrectly? It's a robot.

[Steve Burns]

Okay, why don't we do like Hamlet or Macbeth as a squirrel or something?

[Paul Ford]

That's fine. Or some really hard like something really complicated like anxiety management or you could do that. Yeah.

[Steve Burns]

Okay, so you, you come up with. So react to this.

[Paul Ford]

I'll give you a prompt. Okay?

[Steve Burns]

Yeah.

[Paul Ford]

Okay. Steve, I got a prompt for you. Yeah. You are a world leading therapist who specializes in anxiety. You are also a large excited squirrel. Explain to someone suffering from chronic anxiety all the things that you do as a squirrel to manage your own anxiety and help them as much as you can.

[Steve Burns]

I'm typing.

[Paul Ford]

Thank God.

[Steve Burns]

We can edit this, huh?

[Paul Ford]

Yeah.

[Steve Burns]

Okay.

[Paul Ford]

Don't hit return. We have to add one thing. Okay. Be extra squirrely.

[Steve Burns]

Be extra squirrely. Hitting enter now. Okay. Hello, friend. I'm Dr. Nutters, licensed therapist, board certified in anxiety management and yes, a rather large and kindly squeezed squirrel. Thank you for noticing my tail. I groom it daily. I see you're feeling anxious. That's okay. So do I all the time. Let me share my best squirrel approved tips for handling it. Acorn emoji and a number one. I gather nuts before I panic. When I start to feel that tight chest, fast heart, tail, quiver feeling, I pause and ask, ask what's real right now? What do I need for this moment? Like how I stash acorns in 47 different spots just in case. I make little lists of what I can do now. Small nut sized tasks. Breathe slow, stretch my paws, sip water. Number two, I scamper to a high branch to get perspective.

[Paul Ford]

I mean, look, it's magical, right?

[Steve Burns]

Well, this is delightful. Why don't we just use AI for this all of the time?

[Paul Ford]

Here's what it is. Here's what's going on. Okay. It's a wonderful translation engine. Like literally language to language, but idea to idea. You give it an idea and you say do it as this. You give it a really big pile of text and you say, turn this into bullet points or turn this into bullet points, but pretend to be a squirrel or whatever you can catch your own interest and teach yourself really things that you have a lot of trouble focusing on. Like you just engaged with anxiety content like that because it's in the voice of a silly squirrel. And you went straight to it. And at the same time as I'm listening to it, it's like, right. Get perspective. Oh, you know, squirrels are kind of anxious. They stab, they, they. They hide all those acorns, right? Like, like suddenly you're going from one way of thinking about the world to another way of thinking about the world. And you would never have done this. You would never have sa. I'll be anxiety squirrel. So it's a way to explore and understand different ways of looking at and perceiving things by translating from one format to another. And you can translate from a language, from a computer language to another computer language, from a perspective to a perspective. One of the things I really like to do is like, how would X see this? How would a certain philosophers understand the movie Empire Strikes Back? And it won't nail it down. It's not as good as a college professor doing the same exercise. But it'll just give you a way to see and understand the language and kind of what's going on over there. I understand Empire Strikes Back. I've probably seen it 10 times in my life. I don't understand Hegelianism, but I can ask how Hegel might see it and what he would think about the Death Star. And suddenly I have at least a little perspective, even if it's not. Like, I wouldn't bet my bet the farm on it, but I have a little perspective on what that philosophy is, is because it's contrasted with something I do understand. And so that's a power tool. Again, we focus on so many aspects of this thing because it seems human and because it can do bad things and. Or could do bad things, but it really is a kind of cognitive power tool if used well.

[Steve Burns]

Interesting. Can we, like, ask Yoda.

[Paul Ford]

Can we.

[Steve Burns]

Can we ask Yoda to explain. Explain the dangers sexual health?

[Paul Ford]

No. Ask Yoda to explain sexual health? No. Do it, Steve.

[Steve Burns]

No.

[Paul Ford]

Yes. No.

[Steve Burns]

Let's ask Yoda. I don't know. What's the question we should ask Yoda? Paul? Let's ask Yoda something.

[Paul Ford]

Oh, ask. No, let's ask Yoda something that Yoda would never typically be asked, like about automotive repair or recipes. Ask Yoda about the best. How to make the best breed brownie. You are Yoda. Please tell me the best brownie recipe. Make sure it's very clear that you are Yoda.

[Steve Burns]

Oh, I'm gonna have to bust out my Yoda.

[Paul Ford]

You're gonna do the voice too?

[Steve Burns]

Yeah, I'm gonna Have to try. I mean.

[Paul Ford]

Yeah, yeah. Don't. No, there's no trying, Steve.

[Steve Burns]

There is no. Okay, here's what chat. Here's what chat came back with. I'm not gonna be able to do it, so I'm not going to start doing it. Mmm, brownies.

[Paul Ford]

Yes.

[Steve Burns]

Powerful they are. Yoda. I am La Force. I know and bake. I do. Rich and chewy these brownies shall be. Share this wisdom. I will follow the recipe. You might must tempted by the dark side of over baking. Do not be. Oh, this is great.

[Paul Ford]

All right.

[Steve Burns]

Yoda's wise. I'm making these tonight, Paul, and I'm sending these to you.

[Paul Ford]

Fine.

[Steve Burns]

Yoda's wise and fudgy brownies. Gather you will.

[Paul Ford]

Yep.

[Steve Burns]

1 cup unsalted butter, melted like the ice on Degobah in spring. Two and a quarter cup sugar sweetness balance this brings. Four large eggs from free creatures they should come. One tablespoon vanilla extract. Aroma of the Force. This is one cup.

[Paul Ford]

You cannot stop doing the voice you just like you. It's like a virus.

[Steve Burns]

Salt. A pinch balance it brings. Method of the Jedi. Baker, heat your oven. You must 350 degrees Fahrenheit. Well. In a great bowl, stir melted butter and sugar until 1. They are smooth like the surface of a still swamp beaten eggs. You get it? It's great.

[Paul Ford]

It's great.

[Steve Burns]

This is so fun. Now, I understand that this is not what this is for for, but it's very.

[Paul Ford]

No, it is what it's for. It is what it's for. There's no. It doesn't know. It's it's just a thing. Like, it is absolutely what it's for. To me, I look at that and I'm like, you could get a grumpy 10 year old to make brownies with you and he. They would be annoyed. They'd be like, oh, okay, but. And then you do it in the Yoda voice and now there's a conversation and everybody's being silly and you're in the kitchen and you're doing stuff together like. Like there are applications here that are really new in the universe. And it's sort of the drama of this thing is so intense. The fact that you just said this isn't what it's for. Like, as a computer person makes me so sad because it is for play, it's for weirdness and it's for exploring. And if we were a little more open to that and less worried about trying to use it to like get our master's degree, I think we'd be less scared of it, and it would feel less oppressive. And that's a great use for it. Like. And then you actually should cross reference that with a real brownie recipe. But then again, there's also common sense. Like, if it says, like, you know, pour a gallon of cyanide into this, like, don't, but, you know, a pinch of salt. But the reality is there's so many brownie recipes out there that it's probably pretty good. It's probably a pretty good brownie recipe because it's soaked up and compressed all the brownie recipes.

[Steve Burns]

Yeah, I'm making. I'm making Yoda brownies tonight. Well, Paul, thank you, man, for. First of all, it's great to see you always. And you scared the hell out of me. You really did. But you've also.

[Paul Ford]

Let me tell you, though, I'm not scared. I'm not. I'm strangely optimistic, even though I have the same anxieties that you do.

[Steve Burns]

I see that, and I appreciate that, and I feel that, and I needed to hear. Hear that, you know, and I feel better. Just having a little bit of fun with the technology with you as well has been helpful. I would say that I'm less afraid of what the technology is and a little more concerned about what we might do with it, you know, about where our instincts might lead us to go with it.

[Paul Ford]

I strongly endorse that viewpoint point. Don't take technology seriously. It's a tool for you to play with and explore. You are not the product.

[Steve Burns]

Right. All right, man. Thanks for stopping by, Paul. I love you dearly. This was. This was fun. We could do this again. Every now and then when I'm scared.

[Paul Ford]

I. I would love to come on and talk about what I'm learning, because by the time a couple months goes by, I some, it'll be very different. This is the strangest technology time I've ever lived through.

[Steve Burns]

Yeah. Well, thanks for being such a good buddy, and thanks for making me feel better about technology for 30 years. And thanks for the brownie recipe.

[Paul Ford]

Anytime, beloved. I'll see you soon. All right, man.

[Steve Burns]

Bye, Paul.

[Paul Ford]

Bye. Okay.

[Steve Burns]

I would say that I have a better understanding of the technology of AI for sure. I like that he called it compressed human weirdness. That makes sense to me. I was alarmed when he said that there are no guardrails. That froaked me out. And I guess that means that it's up to us.

[Paul Ford]

Right?

[Steve Burns]

It's just a tool. And it still at least depends on what we do with it, which is good news.

[Paul Ford]

And.

[Steve Burns]

And also.

[Paul Ford]

And also.

[Steve Burns]

Probably. Probably bad news.

[Paul Ford]

Yeah.

[Steve Burns]

Let's go. Let's go outside.

[Paul Ford]

All right.

[Steve Burns]

Well, I think I'm gonna choose to be optimistic about this one. I mean, Paul is. It's an interesting question, though. If you had an infinite answer machine, what would you ask it?

[Paul Ford]

Yeah. Yeah.

[Steve Burns]

Well, it's great seeing you always. Thanks for stopping by. You look great. Alive with Steve Burns is a Lemonade Media original. If you haven't subscribed to Lemon on a premium yet, now's the perfect, perfect time. You can listen to the show completely ad free, plus you'll unlock exclusive bonus content from me as I reflect on this episode. Just press subscribe on Apple podcasts. Head to lemonadapremium.com to subscribe on any other app or listen ad free on Amazon Music with your prime membership. That's lemonadapremium.com Alive is hosted by me, Steve Burns, and produced by Jeremy Slutskin. Our editor is Christopher Champion Morgan. Our associate producer is Akshay Tharabailu. Audio engineering by James Frank Barber. Lemonada's SVP of weekly programming is Steve Nelson. Executive producers are Jessica Cordover Kramer, Stephanie Whittles, Wax, and me. We'll see you next week. And you look great, by the way.

[Podcast Advertisers/Guests (e.g., Julia Louis-Dreyfus, Chelsea Clinton, Sarah Amos)]

Famous Amos. It's a name that is synonymous with chocolate chip cookies. He's also my dad.

[Steve Burns]

I'm in a supermarket. I'm in convenience stores. I'm in department stores.

[Podcast Advertisers/Guests (e.g., Julia Louis-Dreyfus, Chelsea Clinton, Sarah Amos)]

That's what makes Amos Wally famous. Amos. He opened the first ever chocolate chip cookie store 50 years ago. When he passed away last year, I set out to understand how he became one of the most famous black men in America.

[Paul Ford]

I remember dad on the COVID of Time magazine. The headline was the Hot New Rich.

[Podcast Advertisers/Guests (e.g., Julia Louis-Dreyfus, Chelsea Clinton, Sarah Amos)]

While also leaving his life and our family in chaos. What? What did you think when I first told you I was thinking of doing a podcast about our family?

[Steve Burns]

How much collateral damage is it going to cause?

[Podcast Advertisers/Guests (e.g., Julia Louis-Dreyfus, Chelsea Clinton, Sarah Amos)]

From Vanity Fair, I'm Sarah Amos, and this is Tough Cookie, the Wally Famous Amos Story. Available wherever you get your podcasts. Are you Team Batman or Spiderman? Is the ultimate dish pizza or tacos? SmashBoom Best will help settle those debates and so many more. Every episode, we take two cool things, smash them together, and we see which one is best. Debaters use facts, jokes, stories and more to argue for their side. And it's all judged by a teenager, because who is better at judging than a teen. It's fun. It's weirdly informative. It's smashboom Best get it wherever you get your podcasts.