Is AI our salvation or is it going to kill us all? Tom and Vivek roam widely on others’ takes about artificial intelligence, adding their insight and experience to the mix. Along the way they consider Descartes, Ray Kurzweil, Salt Bae, Marc Andreessen among others. If you are looking for a down to earth conversation that tempers the extremes at either end of the debate, this is the one you’ve been waiting for. Hosts Tom Chavez and Vivek Vaidya delve into the multifaceted world of generative AI, examining its effects on truth, deepfakes, business integration, and the workforce. They explore provocative questions such as the potential for AI to befriend or betray humanity, touching on concepts like the singularity and reflecting on notable thinkers' warnings. The conversation also navigates the portrayal of AI in popular culture, challenging the often dystopian images and asking why AI can't be seen as a cooperative force. Tune in for a balanced discussion that embraces both the exciting possibilities and pressing concerns of artificial intelligence. www.theclosedsession.com @ClosedSeshPod @tommychavez @vsvaidya superset.com
Is AI our salvation or is it going to kill us all? Tom and Vivek roam widely on others’ takes about artificial intelligence, adding their insight and experience to the mix. Along the way they consider Descartes, Ray Kurzweil, Salt Bae, Marc Andreessen among others. If you are looking for a down to earth conversation that tempers the extremes at either end of the debate, this is the one you’ve been waiting for.
Hosts Tom Chavez and Vivek Vaidya delve into the multifaceted world of generative AI, examining its effects on truth, deepfakes, business integration, and the workforce. They explore provocative questions such as the potential for AI to befriend or betray humanity, touching on concepts like the singularity and reflecting on notable thinkers' warnings. The conversation also navigates the portrayal of AI in popular culture, challenging the often dystopian images and asking why AI can't be seen as a cooperative force. Tune in for a balanced discussion that embraces both the exciting possibilities and pressing concerns of artificial intelligence.
Speaker 1:
Welcome to the Closed Session, How to Get Paid in Silicon Valley, with your host, Tom Chavez and Vivek Vaidya.
Tom Chavez:
Welcome everybody to The Closed Session. My name is Tom Chavez.
Vivek Vaidya:
And I'm Vivek Vaidya.
Tom Chavez:
What do you want to do this time?
Vivek Vaidya:
I think let's talk about generative AI, again.
Tom Chavez:
Again. But this is going to be like, this is, we get to swing at [inaudible 00:01:11], chop it up a little bit.
Vivek Vaidya:
Yeah. No guests on this one. It's just the two of us. Yeah. Roaming widely. Let's roam widely, Tom.
Tom Chavez:
We've earned it.
Vivek Vaidya:
Yeah.
Tom Chavez:
We've earned it. Daddy finally gets the big piece of chicken. Let's go.
Vivek Vaidya:
Yeah.
Tom Chavez:
Where do you want to start?
Vivek Vaidya:
What do you think about all this talk about AI will make truth irrelevant? Are you even looking at me right now, Tom? Do you know that? You don't.
Tom Chavez:
Well see, this is the problem with a guy like me. Okay, let's do this. Because once upon a time I couldn't decide if I was going to study computer science or analytic philosophy, and I decided, damn it, we're just going to smudge them together. And so I studied computer science and philosophy, which at the time was considered crazy. Why would anybody want to do that?
By the way, if there are any computer science and philosophy college people listening to this, good for you. It's a very ripe time. Here you are, asking epistemological questions about what is truth?
Vivek Vaidya:
Yeah. What is truth?
Tom Chavez:
And it's computer science, AI problem.
Vivek Vaidya:
Yeah.
Tom Chavez:
Okay.
Vivek Vaidya:
There we go.
Tom Chavez:
Coolness. Well look, and so I won't go into a long acts of Jesus here, I promise. But it turns out Descartes was wrong, and this rationalist view we have of a truth out there, this positivist view of something that's false or true has been mostly debunked. The American pragmatists, God bless them, Emerson, Paris, all of these, and physicists, top ranked physicists now, will show up and tell you, "Is the cat alive? Is the cat dead? Is it in the box? Is it not in the box?" So these prior notions we had about live cats or dead cats in boxes at the realm of physics, and in other places, they just don't hold water. And I know that's very unsettling for people, 'cause we all want there to be truth.
So in the context of AI and deepfakes and so on, there are things that are truth here and more relevant and more practical and more useful and more believable than other things. You're a born again [inaudible 00:03:19] too.
Vivek Vaidya:
Yeah. No. But I think it's good to go back and look at similar things, not identical, but similar things in the past. People talk about deepfakes a lot. And in our last podcast, I was talking to Rama, and he gave the perfect example of the original deepfake is a counterfeit note. A counterfeit dollar bill is the original deepfake.
Tom Chavez:
Right.
Vivek Vaidya:
And we figured out how to get around that.
Tom Chavez:
Absolutely. So I was taking the grandiose view of truth, but just wanted to establish at a first principles level, go easy. There are things that are more believable, more true. Is this note, is this dollar bill I'm looking at, now it turns out that we've established, and this is exactly the point that you're driving to. There's a standard and a threshold by which we determine this bill is true or not. Right?
Vivek Vaidya:
Yeah.
Tom Chavez:
And in the context of computer science and AI, good engineers figure out thresholds and protocols.
Vivek Vaidya:
Right.
And I think ... Because the thing with deepfakes, and generative AI in particular, is that the spectrum is so broad. You might want to argue that the counterfeit note problem is a very localized one. Right?
Tom Chavez:
Yeah.
Vivek Vaidya:
But with generative AI, the spectrum is so broad. There's text, there's image, there's video, there's audio, there's a whole bunch of things. But I think that we will figure out a solution.
Tom Chavez:
By the way, here's another good example of things that seemed non-figureoutable at the beginning, and then we figured them out. Know your customer in financial markets. Right?
Vivek Vaidya:
Yeah, exactly.
Tom Chavez:
Before 9/11, okay, that's impossible. Is this person behind the screen a terrorist or not a terrorist? How do I really know? It seemed impossible at the time.
And I guess the other relevant point to make here is that, do you completely banish counterfeits? Do you completely banish spurious bank account? Of course not, right?
Vivek Vaidya:
You can't.
Tom Chavez:
You can't. But you can, as we're saying, come up with really effective engineering protocols for reducing the frequency and impact when it happens. So you and I sound like shiny optimists on that one.
Vivek Vaidya:
Yeah, no. I think that in the long arc, or in the grand scheme of things, there will be good solutions to this. And as with any technology, it will be misused. We just have to be open-eyed about that.
Tom Chavez:
That's right.
Unintended consequences abound.
Vivek Vaidya:
Yeah.
Tom Chavez:
Always.
Vivek Vaidya:
Yeah.
Tom Chavez:
How about an easy one now? Because a lot of businesses are saying, "Well, should I add AI to my product, website, company, blah-biddy, blah-biddy." Everybody wants to sprinkle some AI on. It reminds me of Salt Bae. Wants to sprinkle some AI on it, right?
Vivek Vaidya:
Yeah.
Tom Chavez:
What do you think of that?
Vivek Vaidya:
Well first, I don't think it makes sense unilaterally. You say, oh, let me just sprinkle some AI on it and it'll just become shiny.
Tom Chavez:
But if you sprinkle some AI on it, you can add 200 million dollars to your evaluation right there.
Vivek Vaidya:
True. True.
Tom Chavez:
Because all investors want to hear-
Vivek Vaidya:
So if that's your goal, sure. Do that.
Tom Chavez:
By the way. Exactly. If there are entrepreneurs cobbling together pitches listening to this, sprinkle some AI on it. You're going to add a lot of money to your valuation right there.
Vivek Vaidya:
But every now and then you might come across an astute investor who might just ask you.
Tom Chavez:
That's where they screw you.
Vivek Vaidya:
Right?
Tom Chavez:
Wait, back it up. You said-
Vivek Vaidya:
An astute investors.
Tom Chavez:
An astute-
Vivek Vaidya:
There are ... Come on.
Tom Chavez:
No. You're right. You're right. I'm just being a little bit of an asshole. But they're out there.
Vivek Vaidya:
Yeah, yeah. But actually, your customers might ask you too.
Tom Chavez:
Oh, see that's-
Vivek Vaidya:
Then that's when it gets really tough.
Tom Chavez:
Well, yes. But let's note, I mean ... Boy, okay, quick sidebar. You and I were at a privacy conference with many esteemed legal minds espousing and carrying on about AI. And then soon after that, I remember I was at a marketing conference. And the only thing I think that's worse about hearing CMOs speak so authoritatively about AI, is hearing lawyers talk authoritatively about AI.
I'm sorry I had to get that out there. But everyone's talking a big game.
Vivek Vaidya:
Yeah.
Tom Chavez:
It's a little depleting.
Vivek Vaidya:
Yeah. I think, and I will come back to one thing I say a lot: what's the problem you're trying to solve? AI is a means to an end.
Tom Chavez:
There you go, again.
Vivek Vaidya:
You have to focus on the problem you're trying to solve. So if AI helps you solve that problem, then by all means put it in.
Tom Chavez:
Yeah.
Vivek Vaidya:
But I think in the wake of ChatGPT, everybody is kind of moving towards, oh, let me just call the API and get an answer, and then I'll build a feature. That's stuff is going to be obsolete.
Tom Chavez:
That's right.
Vivek Vaidya:
Everybody's going to be able to do that. Right? ChatGPT ... OpenAI.
Tom Chavez:
Can we also just say on a sidebar, the people who are doing that naively should put the numbers to what it's going to cost them.
Vivek Vaidya:
Yeah, exactly. Exactly. And so I think that the blindly going into just saying, "Oh, yeah, this works with AI." Is maybe gets you a little more of what you want in the short short term, but when things start to stabilize, when the hype cycle starts to subsidize or subside, people will get found out.
Tom Chavez:
Yeah.
Okay, so a related topic is: AI is going to replace us all. Or is it going to replace what kinds of jobs? We'll get to the question of whether it's going to kill us all a little bit later, but right now, here on earth, go easy. Go easy.
Vivek Vaidya:
We're still alive. We're still alive, baby.
Tom Chavez:
We're still alive.
So we'll get to the doomsday stuff in a minute. But in the shorter term, is AI going to replace all of our jobs? And let's start with something about which you know a lot which is coding.
Vivek Vaidya:
Yeah.
Tom Chavez:
Have you used Copilot?
Vivek Vaidya:
I have a little bit, and it's good. It's good. Actually, again, I'm going to copy Rama's lines here. He had a very interesting way of looking at this, and I've been copying it. So his point was that with generative AI, and in particular tools like ChatGPT and Copilot, etcetera, the unit cost of producing output has gone down.
Tom Chavez:
That's right.
Vivek Vaidya:
So you have two choices, now. You can either keep your output constant, or you might want to increase your output.
Tom Chavez:
That's right.
Vivek Vaidya:
So if you want to increase your output, you keep the same number of people and you give them these tools, Copilot, the ChatGPT, etcetera, and your total output goes up. Or you can choose to keep your output constant, the same as it was before, and reduce the workforce. Give them tools like ChatGPT and Copilot, and keep the same amount of output. What you decide to do depends on the function. So in the case of coding, you might want to keep the same number of engineers, and you might get better product, more product, faster, because you have these tools.
Tom Chavez:
Right.
Vivek Vaidya:
In the case of marketing, you might want to reduce your workforce because you can't deal with the extra content that your marketers will generate. You might not have a large enough sales team to process all the leads that might come in because you're not putting out more content, for example.
Tom Chavez:
That's right.
Vivek Vaidya:
So these are things you can think about and then make decisions based off of.
Tom Chavez:
Okay. That's good. That's good, old-fashioned marginal costs, labor economics at work right there.
Vivek Vaidya:
Yeah.
Tom Chavez:
Going in a little deeper now, it does depend on what kind of function we're talking about here, right?
Vivek Vaidya:
Yes.
Tom Chavez:
So let's agree, there are jobs ... If you are doing a job that can be completely replaced by a robot, and an example that immediately comes to mind for me is marketing copy. You got to look at some of the stuff that ChatGPT is generating for marketing copywriters and compare it to what we have traditionally paid marketing copywriters a lot of money to do. Oh my gosh. If I were in that patch, I'd be a little worried.
Vivek Vaidya:
Yeah. But let me actually pose a question to you as the CEO. So say you decide, okay, you know what, marketing people in marketing is obsolete. I'm just going to use output generated by, copy written by, ChatGPT. The first mistake that's made, the first hallucination, the first wrong fact, the first wrong tone. What are you going to do then?
Tom Chavez:
Right. Which gets to, so here's my little framework, my suggestive framework for this. You're either doing something that: you're working for a robot, you're working alongside a robot, or you're telling a robot what to do. Okay. So in your example, and exactly to the point I think you're getting at is, no, no, no. It's not going to be a hands-free do what the robot says.
Vivek Vaidya:
Yet.
Tom Chavez:
Yes. And yeah, put a thumbtack in that. Let's come back to it. But you're co-piloting with an AI that's enriching, accelerating, extending what you're able to do.
Vivek Vaidya:
Correct.
Tom Chavez:
And that's a very, it's an optimistic, but it's also a very practical vision of where we're headed.
Vivek Vaidya:
Practical. Exactly.
Tom Chavez:
Boy, we were in unison on practical.
Vivek Vaidya:
Yeah.
Tom Chavez:
Say it one more time. 1, 2, 3. Practical.
Vivek Vaidya:
Practical.
Tom Chavez:
Bang. But the time before that wasn't ... It's 25 years.
Vivek Vaidya:
You know what we didn't do? We didn't say, jinx, you owe me a soda.
Tom Chavez:
That's right. My kids do that, but jinx, no backsies. Oh my god. Focus.
So now, if you're telling a robot what to do, if you're programming, designing these systems, I mean, that's an interesting place to be. And by the way, especially speaking of Copilot, and that's for people who have traditionally thought, look, I'd like to be in computer science, but I don't know if I'm a very good coder. We are approaching a time where organized thinking and superior systems design becomes a first class kind of capability as the prequel to the code. Yes, you've got to do your hundred pushups and do some coding at some point, but we are approaching a time wherein the creation of these software systems becomes a lot more conversational.
Vivek Vaidya:
Yes. I think-
Tom Chavez:
You can see that coming.
Vivek Vaidya:
I agree with you. And I think that there will be ... The curriculum in computer science departments I think will shift to allow for this conversational kind of modality. But also, you need to be able to understand the code that is generated by tools like Copilot.
Tom Chavez:
That's right.
Vivek Vaidya:
And you need you to be able to assemble it into systems in a way that you can understand if I assemble it like so, it's inefficient. If I assemble it like so this other way, it's more efficient. Right?
Tom Chavez:
That's right.
Vivek Vaidya:
So that will become a skill.