What's up, Sam? Hey, you like the you like the fit?
Where did you get that?
Our boys at Jamby sent it over the no small boy stuff Christmas edition.
I feel like I can rule the world and know I could be what I want to.
I put my all in it like days off on a roll.
You know, it's pretty funny.
I actually use the phrase no small boy stuff like kind of a lot.
Yeah, I remember the guy who tweeted it.
I think his name was Bengali 87.
And this was back in 2022.
He said, best business entrepreneurship podcast out there.
Big money, that is no small boy stuff.
I love that. And that's basically the phrase that we use for this podcast a lot.
No small boy stuff.
But frankly, I kind of use it a lot in my life.
Like, I don't know, man, that small boy kind of stuff.
Like, it's sort of like in succession where they say you're not a very serious person.
It's kind of like that.
I also use the phrase, but I never say it because saying it to me feels so cringe.
But I think it like a thousand times for every one time that I say it.
And every one time I say it, it feels so awkward to me.
It's like saying it's like saying, just do it in the Nike slogan way or something.
You don't really want to say that.
Hey, guys. Yeah, this is the fourth quarter we just like you, baby.
Just do it. And then they'll be like, what, why are you saying slogans at us?
But I do think it a lot.
It's actually like thinking a lot meaningfully affected the trajectory of my life is to use this phrase.
And because there's so many situations where there's like a little small boy response, I'll behave like a little small boy in this situation or.
Yeah, that phrase and what Amjad said recently about what will make the better story that has had a fairly meaningful change just in, you know, it's only been a few weeks.
But like, I think about that actually a lot.
He also says something else where he was talking about he was basically so comfortable with this like 10 year plus odyssey that he's been on building this.
And we're like, wow, you've been doing this for so long.
And wow, you did this for years before you had really any recognition or any funding and you just kept going.
And he was just like, yeah, I persist.
And he was just like, yeah, he's like, I think that's what I do.
He's like, I know I didn't really think about it that consciously, but like, I'm pretty comfortable pushing the boulder for a long time up the mountain.
And I guess I realized, like, I guess that's my like competitive advantage.
Like I made it for the long haul and I'll just persist.
It's like, oh, and we were both like small, like a quick intake of breath.
What do you want to start with today?
All right. I got a good story for you.
So there's this great Nassim Taleb quote or tweet where he tell lab who wrote Black Swan and antifragile is kind of this like contrarian thinker.
Was he like a successful hedge fund investor?
But he was successful because he had an interesting life philosophy and then he became like a thinker.
Is that his story? I believe so.
I believe so. I believe he's like a successful trader and part of his success, unless I'm mixing him up with somebody else.
Part of his success was that he he noticed that humans are we would rather win frequently in small amounts and then lose a bunch when we're wrong.
It's like gambling.
It's like playing craps, right?
You know, one roll of dice, you made a little money, two rolls of dice, you win a little bit of money, but eventually you roll a seven and it wipes out the entire board.
All of the chips go away.
But he's like humans are more comfortable with that versus he was willing to bleed a little every day and look stupid every day for years.
But then when his big, you know, sort of like the big short, when his big bet pays off and his contrarian bet pays off, he makes all the money in one day.
Got it. And I think that's his story.
If it's not his story, his book is talking about the guy who does that.
So I can't recall if it's him or if he's the author or he's the author and the hero of the story.
OK, so he teleped tweeted this thing.
He goes, I conjecture that if you gave an investor the next day's news 24 hours in advance, he would go bust in less than a year.
And this is basically the back to the future premise, right?
So I don't remember the movie Back to the Future, but what's his name?
Biff or whatever he gets, he finds like the sports betting book that tells all the winners for the next decade.
Exactly. He goes back in time.
And then he's just becomes a gazillionaire because he knows the scores.
OK, so like, let's take.
Yes, if you knew the exact score, you'd have to be pretty dumb to not win.
What these guys did, what Naseem Taleb is saying is I could give you the news.
So not the price change, but I could give you the news.
And I bet you would trade incorrectly.
Dude, I think about this all the time, by the way, all the time.
I think if I know what I know today, but I was 10 or 50 years ago, how would I capitalize on that?
I think about that all the time.
Right. And, you know, usually the easy answers for that are, I just buy Bitcoin, I just buy Google, right?
Like, yeah, it would actually wouldn't be that hard if you if you could convince yourself, hey, do this one thing and just shut up and trust me.
Like, don't don't touch it for 15 years.
Now, what these guys did was a little bit of a different experiment.
So what they did was they took 118, as they called them adults trained in finance, and they did the crystal ball test and the crystal ball test was as follows.
They said we're going to give you money.
They gave them 50 dollars each.
So I said, you have 50 bucks and you get to place trades and you're going to trade.
But we're going to before you make a trade, we're going to show you the front page of the Wall Street Journal, the actual front page of the Wall Street Journal from 15 random days in the last like, I think, 20 years or something like that.
OK, so 15 random days, we're going to show you the front page of the Wall Street Journal.
And that's a Wednesday edition.
And you're going to place the trade that would execute on the Tuesday.
So the day before that news.
So you had the news and 24 hours in advance.
They blacked out the stock prices, so they wouldn't just show you, oh, Johnson Johnson is up 20 percent.
Right. But they would say like there would be the headline about Johnson and Johnson.
They would just redact the actual stock price.
Johnson and Johnson beats earnings.
Beats earnings. Exactly.
Record job record unemployment.
Record jobs, posting Fed indicates, blah, blah, blah, things like that.
And by the way, anybody can go play this game online.
There's like a link to it.
We'll put in the show notes, but you can go actually do it yourself.
I did it too. There's a couple other caveats to this when you go do it, which is it's not just a buy and hold because what I was going to do, I went and did the thing.
I was like, oh, cool.
This is news from 15 years ago.
I'll just put all my money in by and I won't trade.
I won't do anything else for the next 15 years.
I know there was a bull market, so I don't need to be smart.
But the way that this test was designed was the trade executes and you either go up or down that day.
So it's kind of like the trade closes that day.
You know what I mean?
You get one day gain based on the one day news.
So they do it and the results are not good, as you might expect.
Otherwise, I wouldn't really talk about this.
So the results are not good.
Half the players lost money, even having been given the news.
One out of every six players lost everything.
And the way they lost everything was they let you trade on leverage.
So you can trade up to like 20x leverage if you want to in this thing, like an options trader could, or you could just trade, or you could skip.
You don't even have to trade any given day.
You could just say pass.
I don't feel confident.
Like an example trade would be they would see something that says like Fed has got a cut rate today.
I guess you would assume that the index is going to go up like a couple percent.
So you would bet on the index or bet on what?
It's basically the S &P 500 or it's the 30 year treasury.
So you bet on one of those two things.
You buy the SP index for that day.
Or you could short a bond.
You go long or short, or you can go long or short the bond.
So that's let me just show you an example.
This is the Wall Street Journal that they, the article that they show you.
So it's Obama does something.
And then you see this business and finance section and talks about Rupert Murdoch.
Chesapeake Energy says that their CEO is going to step down.
Auto sales are up. Homeland Security says blah, blah, blah.
So there's all this news.
And so you then go here and you place a trade.
So you say, all right, I'm going to go in this little game here.
You can see my screen where it gave me a million dollars.
So I'm going to trade.
And it says today's movement.
I bet a million dollars.
So I use my full stack with no leverage and the day was up 0 .62%.
So I got an extra $6 ,200, right?
Then it gives me the next one.
And it says, oh, there's a deadly plane crash.
Iran is doing some shit.
Okay, cool, blah, blah, blah.
Kraft is in talks to acquire this Brazilian company.
And they're just blacking out any of the stock price news, right?
So you read this and you can decide what you want to do.
And you do that over and over and over again.
So 15 days in history.
And they tried to do it as 15.
They did a third of the days where fed quarterly meeting days, a third was jobs reports, and a third was complete randomness.
And they're like, we're not trying to trick you.
We're not cherry -picking misleading days.
These are just actually random front pages, okay?
This is an awesome experiment.
And so back to the results now.
So, like I said, half the people lose the money.
One out of six lose everything because they got over -leveraged.
The average person was only able to gain 3 .2%.
So even being given the news.
During that era, the market was up on average, I think, 15 % a year.
Over the last 15 years, I don't know when this was done.
Again, these are one -day trades, right?
So you're not just buying and holding.
Oh, it was for 15 days.
The experiment was a 15 -day experiment.
Exactly. And so, okay, now why?
There's two ways you could lose in investing.
One is you bet wrong, meaning you pick the wrong direction.
You think it's going up, it actually goes down.
So basically, even given the news, they were basically only able to bet the direction correctly, 51 % of the time.
So it's the same as if you just flipped a coin, you would have been right the same amount of times as you were being given the actual front page of the Wall Street Journal, okay?
So information doesn't lead to actual insight, especially news.
The second thing is that, why did they do poorly?
They bet sized very poorly.
So when you had an event, even when you were correct, people didn't size up their bets enough.
And when they're incorrect, they sized up their bets too much for the level of conviction that they had.
And this doesn't go in line with what people think.
So they surveyed people separately, and basically, 70 % of people thought that even if they got the news, basically, they thought that even four -week old stale news would be predictive.
And 70 % of people thought that, but in this case, it just showed that one day fresh news doesn't even really help you.
Okay, so then they went and they did an extra experiment.
They go, okay, maybe those 118 financial trained adults, maybe they're just not the best of the best.
So they went and tried to find the best of the best.
The best of the best actually did better.
So they went and found five people that were hedge fund guy, the head of trading at a top five bank, seasoned macro traders.
So they're used to trading on this type of news.
They're considered the best in the world at this.
And they actually did better.
So what they did was all of them finished with gains.
So all five finished with gains.
On average, they were up 130%.
And they also didn't bet on one out of every three news things.
One of the big ways that they were better was they just didn't bet all the time.
Whereas the casual was too active.
Okay. What else did the pro do differently?
They were only right 6 % more.
So if the, I think if the test group was right 51 % of the time, if it's going up or down, these guys were only right 57 % of the time.
It wasn't like they correctly interpreted it, but when they were right, they bet size properly and they never risk too much of their bankroll to where they couldn't recover.
And so isn't that amazing that, you know, only being, you know, 6 to 10 % better at your predictive ability, but would yield a much bigger result, right?
3 % average gain for the test group, 130 % average gain for the pros.
So it's these small edges that can make a huge difference when you apply leverage properly, which was the bet sizing.
And what's the takeaway that Nassim said, which is what?
Well, he was saying it in a polarizing way.
He goes, I conjecture if he gave an investor the next day's news, he would go bust in less than a year.
And this kind of, this basically showed that one out of six would go bust because they would get overzealous around this perceived edge that doesn't exist.
And on the whole, most people would just do worse than if they didn't have the news.
It's no better than random, right?
And that's actually one of his books, fooled by randomness.
And at the end, they use this quote by Ray Dalio in there.
This is a great quote.
It goes, he who lives by the crystal ball will die eating shattered glass.
Dude, that's insane.
It's weird that multiple smart people come to that conclusion that I never would have come to.
Like I would have thought, like I guess everyone have thought that if you know the future, you know, the news, you absolutely are going to outperform.
Exactly, exactly. The counterintuitive wise conclusion.
Let me tell you one other related one.
Said there was one other story here that was kind of interesting.
There was a real world version of this where a hacking group got access, they hacked the press release system.
So they had access to the next day's press releases that companies put out when they have like major announcements, earnings, results, et cetera.
They got access to all of the press releases that were coming out the next day.
And they were using it as like their own form of like, you know, home brewed insider information, right?
They were able to get insider information and so they could place a bet in the market overnight or the next morning.
And that's like a 12 hour leading indicator, maybe.
Because like if you're if you're going to fire your CEO, you submit the press release, maybe at five or six PM on a went on a Thursday and then nine am on Friday, you announce it or the wire goes live.
Something I don't know the exact timing.
I would imagine that it's a tighter window that because there's too much leakage.
But even a 12 second advantage would be a huge advantage.
If you knew 12 seconds ahead of time, what the news is about to be, you could just push the button, right?
That's all you got to do.
That's like an interesting like, you know, like HubSpot, for example, whenever they have that I'm a shareholder of whenever they have like an earnings, I like I know when it's going to go live that day.
So someone is like writing that and they've submitted it to, I forget what the PR come with the things called.
It's the popular news wire, whatever.
Yeah, like whatever the popular thing is, that is kind of I did, I never even realized that actually.
That's why there's rules, right?
When I was at Amazon, you couldn't trade the stock, there's a window, there's like a frozen window.
So x days before the announcement, you can't make any trades.
Oh, I know that, but I'm saying the employees of the PR.
Oh, right, right, right.
They probably have the same, right?
Because I didn't even think about that as a leakage.
I tell you, when I accidentally did that trade and then I had to go to the.
What were they like, you're an idiot?
So I'm like, I've learned about this afterwards, right?
I'm a startup kid. I don't know anything about this.
We get acquired. I make a trade and then I'm like, oh shit.
What was the trade?
So we're at a subsidiary of Amazon, right?
Like you bought Amazon before you sold to Amazon.
I bought Amazon stock, I bought more Amazon stock or something, or I sold some Amazon, I don't remember what it was.
And I was like, oh shit, I just, I just inside our trade.
Did I just get my hands dirty with a little big boy business?
And so I'm like, oh shit, what do I do?
Like you need to go speak to the general council.
And I was like, what?
So I get a meeting with the general council, urgent, urgent, possible, possible big money move made.
And I send the email, they get me a meeting stat.
I go in and he's like, so what happened?
And I was like, I went and made a trade, you know, I'm in the window and, you know, I'm an executive.
So Hank, take me away.
Yeah, back me up. I'm a bad boy.
I'm in a bad boy. Take me away.
And he's like, so how much did you trade?
And I was like, I was like, yeah, it was like 150 grand.
And he was like, it's okay.
You just, uh, my lunch is outside.
Can you bring it in before you leave?
He was like, this is for actual execs at the actual company who make actual trades.
I was like, okay, gotcha, gotcha, gotcha.
Let me go sit down.
That's actually hilarious.
He just like totally dismissed you.
But what about, let me finish the story about the hackers.
So let's put you, let's test your criminal mastermind, which I love.
I love doing this, by the way.
How would I, how would I cheat if I was going to, by the way, you know, that it's always the women always think to themselves.
How would I get away from this bad person trying to hurt me?
And the men always think in this, in the, they align with the criminal, I'd be the bad person yet.
Like, how would I get away with this crime?
That's like, I realized that after watching a lot of true crime.
It's so go ahead. I like this experiment.
So you're the hackers.
You get this, but here's the problem.
You they're like, Sam, we got it.
We hacked him. You know, Dave over here in the corner did it.
He got into this. He got root access and they print out all of the press releases coming out, but they put it on your desk.
They're like, Hey, we got like an hour.
We got to make a trade.
And now there's 60 ,000 press releases on your desk.
What do you do? I guess pick like a random five and hole and act on those as soon as.
Yeah, I mean, that's a very challenging situation.
I guess the challenge is the first five.
And if it's good news, buy the stock.
If it's bad news, somehow short about it, I mean, I'll do that.
So I guess I would only find like the five good news ones.
Like I think I would still just end up holding the index fund vendor.
I'm just doing exactly what I always do.
80, 20 stocks and bonds, baby.
So what they ended up doing was they were like, all right, you sort of need to do a search function to figure out what news affects the price the most in a positive or negative direction.
And, um, I think what they figured out was that.
It was merger announcements that would be the highest kind of like volatility for the company that was getting acquired because it almost always gets acquired at like a 50 % premium to the way the stock was trading.
And so I think what they realized was we need to be able to quickly discard 98 % of the news and information because it's noise.
Right. Which goes back to the same experiment, right?
Most of the news information is noise.
The secret is figuring out what is actually signal and most of us can't do that.
We overestimate our ability to figure out signal versus noise.
And so they, um, they figured out the signal.
It was these merger things and they, even they were only right in their predictive ability about 70 something percent of the time, it was enough to make hundreds of millions of dollars very quickly before they got caught for doing this.
And then they all went to jail.
But isn't that cool also, I think that's the ending.
When we sold the hub, when we sold the hub spot, I think the share price, I think it was $350 and then like the week they announced it, it went to like $460 or something like this.
Anyone can go back and look at it.
It was February of 21 and damn Sam, the needle mover over here.
Well, so that that stock price went up.
Like, I guess it's a market cap of like one or $2 billion.
And I remember going to Kip the CMO.
I go, you're welcome.
He's like, Oh yeah, it was this acquisition that got mentioned one time in our earnings call.
It just barely, it wasn't the fact that we had just announced that we grew by 45 % and have been compounding growth of like this, this, this.
And I was like, Yeah, causation is difficult to prove.
Yeah, I agree. Yeah, I'm like, you don't understand.
Can I tell you? Alright, so we two or three years ago, we talked about AI girlfriends.
I sort of understood it because I like have actually developed like pretty good friendships, mostly via text messages.
I think a lot of people who have group messages here feel the same way.
I didn't entirely understand it.
But in the last two or three months, I've been using chat GPT in a way that now I'm like, yeah, this would go away, I would be very upset.
And I understand why people were very upset when their when their AI girlfriends replica got, when they did like a software update.
Yeah. And so basically, I've been using chat GPT as like my thought partner slash assistant slash therapist.
And you actually said something recently, that made it a lot better.
So I sat down and I'll explain how I've used it.
But I sat down and I said, Hey, can you ask me all the questions that a therapist or life coach or an executive coach would ask, and we could spend a few hours, but just me downloading, giving you a download of my life.
And I did that. And since then, it's been magical.
And I've been using it for all types of purposes, I use it all day.
And I wanted to maybe explain to you how I'm using it.
Maybe you could explain to me if you're doing the same, which I think you are and how you're using it.
Right. By the way, I'll just give you a quick one.
My prompt that I used yesterday for this, I said, I was explaining the situation.
I go, Ask me a few questions, one at a time.
Then when you feel you have enough info, then try to give me a suggestion, because otherwise it just tries to like, you know, man, you know, like mansplaining or what is it called when like, guys hear like your girlfriend is explaining stuff to you and you're trying to fix the problem right away.
She's like, No, I'm not trying to get the fix right now.
I just want you to hear me and understand me.
And you're like, What?
I thought you just want the answer as fast as possible shoved into your throat.
And like, that's what chat GPD does by default.
It's it's yeah, and there's a bunch of other downsides that I want to explain to all this and how I'm working around it.
But first, I'm using it for a variety of things.
So I'm using it for personal finance stuff.
And I'll give an example for each in a second.
I'm using it for business questions.
I'm using it as like a sparring thought partner of like, I'm thinking about doing this, what's your opinion?
I'm using it as a therapist of like, you know, I'm struggling with this person at work or my personal life.
How should I handle this?
Or what should my life goals be?
And I'm also using it for helping me decide which tasks so I'll give an example.
So for net worth, I use Kubera Kubera is like a net worth tracker, you just log in with your bank accounts and all your other accounts and it tells you your net worth whatever.
Well, they actually have a feature where you can download the information specifically for chat GPT and you upload it and it doesn't have any identifying information.
It's not like it has passwords.
It just has a bunch of numbers.
And so you can I will upload this chat GPT.
And I'll say things like, you know, I like to be conservative, like, what would you rate this portfolio out of 10 of risk?
Or, you know, like, what's your opinion on it?
Like, what would Warren Buffett say?
You can ask it all types of questions like that.
Or you can also say like, you know, how much should I spend on a house?
Or what will my net worth be in 20 years?
Like things like that.
And it's been actually really amazing.
Another thing that I did was I took the main KPIs from my company, and I uploaded it to it.
And I'll be like, what are the needle moving things that I can do for this company?
And you could do your KPIs, which is typically like an Excel spreadsheet, like your company's churn, new users, things like that, you can also do your company financials.
And then another thing that I've been doing is I will actually take screenshots of my calendar, and I'll upload it and be like, what tasks should I be doing for the next week, the next month, the next quarter, to get to the goals that I've told you about, you know, my life goals, which, by the way,
you helped me create, you helped me create quarterly and annual goals.
How should I be spending my time today, tomorrow, next week, and next and it gives me an agenda that I literally print out.
And I work according to that.
It's like pretty wild.
And that's why I've been using it.
And then all day, I'll be like, how should I reply to this email?
What's your opinion?
It's kind of crazy.
So that's how I've been using it.
It's like you have neurolink, they just never did the surgery.
All right. You're basically putting AI, like, as the, you know, operator in your brain in many ways, but you're just like, you know, we just haven't reached that tech point where the chip is already implanted.
Well, the next step of that is here's what's gonna happen.
There's gonna be software, it probably exists.
I'm tinkering with a few of them that records your computer screen, your phone screen, the words that you say out loud, the things you type, and it's gonna, and it's going to give you feedback on how you spent your day, it's going to give you feedback on what to do, things like that.
So it's gonna, like, you know how they, there's a book.
I forget what the book is, but the premise is Google knows more than you because you are more honest in your Google searches than you are when you talk to your spouse or your friends or whatever.
The same thing happens where it's like, yeah, you know, I, I spend this much time working on this, this, and this, and I just be like, no, you did not spend that much time do it.
And also, you told me that you're trying to be nicer.
You wrote like eight really mean emails.
Do you know what I mean?
Like, that's how it's going to be in the next six months.
I think there's gonna be products like that that are actually nailing that.
Yeah, I think the CEO of Microsoft, I don't know if you heard this story, but I guess when Balmer stepped down and they needed a new CEO.
And at the time, Microsoft was kind of in a downward, downward to flat.
It was an uninspired stock and company at the time.
So they needed something.
And I don't know if you heard the story.
So the guy who became the CEO, Satya Nadella, actually wrote a memo, like a, wrote a kind of like a manifesto, an internal manifesto about like what, what Microsoft needs to do.
And he ends up getting the job.
And at the time it was like, he's like, I didn't, he's like, I never thought I'd be the CEO of Microsoft.
Like, you know, you join Bill Gates as a CEO or whatever, and then Balmer and you just assumed they're always going to bring in somebody, but they actually promoted him from within.
And he, he wrote this thing.
And one of the key principles that he wrote in this, this is a while back when he wrote like Oh, five or 10 or something.
This was in 2014. So, um, he wrote, uh, he bet on two things.
I don't remember the second one, but I remember the first one he called ambient intelligence and ambient intelligence is kind of what you're describing, which is basically like, um, how do you have, uh, you know, computer intelligence, artificial intelligence, but just like kind of on and be like kind
of in your, in your environment so that it can be helpful to you.
So it just knows what you need without you having to go fetch it without you having to go ask specifically, it can either anticipate it, it can be aware of all of your context so that you don't have to like first explain the whole situation and then be able to just ask your question.
It already knows your situation.
So you could just ask the question, that sort of thing.
And so isn't that cool that he, uh, you know, like so long before, uh, you know, opening AI wasn't even incorporated at that point or something like that.
This is a very long time ago.
So, um, to bet on that as like one of the two like ways that the tech puck is going, um, pretty baller, which is shockingly hard by the way.
It's hard to make these predictions and remove like the limiter part of your brain and just imagine like, yeah, but what would be, what would be amazing?
You know, like what would be cool if it, then that's actually, that sounds easy.
It's really hard because you constantly think like, well, I can't do that, you know, like, because that's impossible or that would cost too much money.
Like there's all these limiters, but the way that I've been using this, like, if like, it doesn't work perfect yet though, by the way, this is like, there's a few issues with this and I am like super not technical.
The first thing is contextual or context windows, like, the more you talk to it, it doesn't always learn more.
You actually it runs out of memory in a weird way.
And, uh, and so I've been testing like a variety of different platforms Gemini versus, um, chat GPT, but I want to use chat GPT because I think it's gonna be around the longest and they're gonna innovate the fastest.
But it's not perfect at all.
But it's like, shocking how useful this is.
I finally, for a long time, I'm like, yeah, AI is great.
Like I can look just like Google a stat and it's gonna tell me, but now it's more like, this is my life.
Like I am using this more than anything.
And so like they had the new $200 a month thing come out and I don't even think I need the features, but I'm like, whatever, I'll take it.
And so I've like contemplated contemplating like, should I like invest a little bit of money into like building up these systems just for my personal operating system and like making my life great.
And keep in mind, I don't know anything about any of this shit.
I just know that it's just effective.
Like it just literally is helping me get my day done better.
And it's like a great bit of advice.
Like here's a really another like practical way.
I mean, I'll upload my measurements for my body and I'll be like, find me clothes that fit or like does this fit?
Does this pair of pants fit?
And you just like post a link like I just I've been using it constantly.
All right, let's take a quick break because I got to tell you about a friend of the pod who's got their own podcast.
If you know Steph Smith, she is a legend she's been on MFM many times.
And she's got her own podcast called the A16Z Podcast.
And it's all about technology.
If you think about it, technology has evolved like crazy.
I mean, I grew up in the 90s.
I had CDs phones, had cords, you couldn't use the internet if your mom was on the phone.
And now there's like 3d printers and there's rockets that can go up into space, AI, there's so much crazy stuff going on.
And you got to have a place that helps you stay ahead of the curve.
And that's what the A16Z podcast is trying to do.
It's a podcast from the VC firm, Andrieson Horowitz.
And it's trying to give you an inside look at the trends that are shaping our future.
They've had guests like Mark Cuban and Neil Stevenson on, and they talk about topics like deep fakes or the science behind GLP ones or autonomous drones.
No small boy stuff at all.
Steph is the host. She's awesome.
I think you'll enjoy the podcast.
So check it out. It is the A16Z podcast.
And I like this tagline to say, it's like eavesdropping on the future.
That's pretty cool.
That's a good tagline.
So check it out the A16Z podcast, wherever you get your podcasts.
How are you if you are using it to be like this, like sparring thought partner?
Yeah, yeah. Well, I think this is the key.
So, um, so what, what we're saying is basically the way that I think by default, people will use this is you ask a question, it gives an answer.
And actually a equally, if not more powerful way is to do the exact opposite.
You basically say, I have a, I'm trying to think about this.
Ask me questions and then you, and you get it to ask you the questions.
And then in that way, it's your sparring partner.
It is your thought partner in like kind of fleshing out or getting your own clarity around a situation.
And it's available 24 seven.
It doesn't judge. It's, it's, you know, super, super intelligent, but also has like, you know, empathy.
You can, you can go back and forth instantly.
It's always available.
Um, and there's no lag time, right?
It's better than a friend, right?
You know, you have a friend who you bitch to and you're like, I just need a vent.
And like, just give me like, what should I do here?
But you kind of feel guilty, like laying everything on them or making it all about you.
And like, they don't quite understand exactly what you're talking about.
This is just that person, but better.
It's one of the main reasons why coaches and therapists are great, because you're like, cool, we're going to have a completely one -way conversation here.
Like, I don't, I don't got to give you nothing.
I can come here and be a taker and that's the arrangement.
And like, you know, I gave you the money.
That's what that was for.
And now from there on out, I don't need to consider your feelings in this interaction.
That sounds like ruthless, but it's true.
It's why it's different than just talking to a friend.
Whereas friend you gotta be like, sorry, am I taking up too much your time?
I don't mean to put all this on you.
But you know, you're like, you're always trying to like kind of half apologize and then reciprocate.
And one of the cool things about a therapist or coach is like, that's not the social contract.
That's not what's expected in that situation.
AI is even better. It's like, Hey, sorry to bug you at 1am.
I just, I'd like to talk right now and have like instant responses with complete intelligence.
And I'll just keep saying no, tell me, you know, no, try again until I get something that's satisfactory to me.
It's like, you couldn't even treat a human like that.
Right. So it's pretty great to be able to do that.
It's become strange.
I call it dude. Sometimes I'm like, dude, what's your problem?
That's wrong. Stop getting these like, like, like it's, it's, it's, it's strange.
Because if you think about it, when you're, you're texting your friends, like it's, because it's like in the same window or next to the same window on your computer, like you kind of forget that this is a machine and you can train it how to talk.
It's very strange, but it's actually quite effective.
Do you know how an LLM works?
Do you know what like deep learning is?
No. I went and watched some videos the other day, just to get like, cause I was like, how is this magic?
Magicing? What is going on here?
I was one by this guy, I think it's called like three Brown, one blue is like his, his username or something like that.
It's got millions of views and he explains, you know, a machine like what is deep learning, which is like the technique that worked with AI.
And the second thing was, you know, how large language models work.
What does it even mean?
What is large? What is the language model?
What is that? What does that even do?
But check this out.
So, okay. Like here's the example that, that, that it gave.
Okay. So this is me not even trying to explain to you what it is.
Cause my explanation is going to be pretty bad.
This is me just saying, I can't believe that this is what actually is happening.
I cannot fathom that this is the actual scenario.
Okay. So let's take this example.
I wrote this, I put this on a card cause like, I can't forget this.
I'll never forget what I learned.
All right. So imagine this number seven, right?
So let's say you're trying to train AI to be able to see that this is seven.
How do you do that?
You can hard code it, but well, every time you see the number seven, it's like a capture, right?
It's like written a little bit differently.
So it's like, you can't just say this is exactly a seven because you write your seven slightly different than me.
Maybe you put the little line through it.
Maybe you have a little angle to it, whatever.
Right? So you just want it to be able to recognize anybody's handwriting and figure out seven or not seven, right?
What number is it? So how does it work?
So imagine basically, um, a classroom.
Okay. So here's, um, a row of kids.
So there's 10 kids standing there and each of the 10 kids is, um, like holding one of these cards with a different number on it.
Right. But it actually, it doesn't have the whole number.
So, um, or actually they have the whole number, but for, at first it just says, all right, there's a whole index card.
We got to figure out, we don't even know if this is a seven or a dog or a car.
It could be anything.
Right. So it just zooms in and it says, let's look at this little section right here.
Like these 20 pixels.
Okay. These 20 pixels.
Um, you know, on this area, it's white.
So if you got color there, sit down kids.
Anybody who's got color over here, sit down.
Cause this picture is white over here.
Can't be, can't be you.
You're eliminated. And then over here, it's like, Hey, there's some blue ink.
Something is here. So if you've got blue ink in this little section, um, stay standing if you don't sit down.
Right. So that like eliminates a bunch of, you know, like kind of thought processes.
So then it passes it to the next layer, the next layer of 10 kids.
And it says, all right, uh, who here has got this flat line?
Okay. And so the seven stay standing.
The five stay standing.
You know, the threes are kind of like, Hey, we got some stuff up here up top the eights, but you know, the four, the number four doesn't have a little roof on top.
So it's like, I'm out, I'm out.
And they're like, okay, go sit down.
It's like paintball, right?
You're out. Go go to sit on the side.
And then say, now you're left with like, you know, some of the numbers and then it says, all right, we got a little, little stick over here.
Who's got a stick over there.
And it's like the threes are like, Oh, I'm out now.
That's not me. But the sevens of the fives like, Hey, we're still in, it might be us.
Right. Bingo. And so you just keep passing it from layer to layer, showing it like kind of more pixels on the screen.
And it's trying to get with some level of confidence at the end, right, it's going to be seven and maybe five at the end.
And the sevens like, yo, I'm 90 % sure it's me.
And the five is like, yeah, it's maybe 10 % that it's me.
It's just an ugly five.
And then that's how the AI knows that this is a seven, because it passes it from layer to layer to layer to layer, looking at the pixels on the screen and basically trying to figure out trying to guess, is it, is it one of you, I think with some probability, it's this.
Okay, that's just recognizing a number.
Okay, now imagine what you're doing.
You're giving it KPIs of your company, it has to understand what a KPI is, what a company is that you were looking for strategy, what strategy sounds like it's gotta say something that you as a successful business person who sold your companies for, you know, tens of millions of dollars that you will respect
the output of this.
Like, isn't that mind blowing that that's even a thing?
And so that now you take, how does that work?
So now you take, instead of the seven, take an example where it's like, the dog blanked, right?
So it's like, what's gonna come after it?
It basically sees a sentence, the dog or the dog, like what's a dog?
And what do they commonly do?
It doesn't even know that it has no idea what a dog is.
There's no meaning.
It just has it read the whole internet.
So what they did was they were like, Hey, go read the whole internet, which like, if you are I, we were like, yo, Sam, I gotta like, let's do this, man, we could do this.
We're gonna take so much Adroll, we'll stay up all night.
And we're gonna read 24 seven, all the texts on the internet, it would be like thousands of years before we could ever ingest what, you know, what they gave it in one training run, right?
So they said, go read all the internet.
Cool. Done. All right.
Now user puts in a sentence, the dog blank, guess what, guess what the next token is.
Guess what? The next little word is that comes after the dog, the dog.
It's like the dog barked, the dog jumped, the dog, you know, is hungry, right?
Whatever. It could be like one of many things.
So then it takes the next word, which might be like the dog barked.
And then it passes that phrase back through.
It's like, now you've got the phrase, the dog barked.
What comes after that?
And it's just loops that over and over again to generate the next word.
So that's when you see chat GPT writing.
It's literally taking like the next token.
It thinks it should say that it feeds it back through and then says, okay, well, if I said, if I said the dog barked, then I gotta say loudly.
Right. Okay. Loudly.
Period. If I said the dog barked loudly, what would I say next?
And then it would keep, and it keeps recursively doing that.
And that's what's actually, that's how it generates a training thing.
Right. And that's like, you know, this is only part of it half explained correctly, but let's assume for a second.
Then I'm not like completely misinterpreting this.
Let's assume for a second that this is only, you know, a percentage of what, what is actually going on, right?
There's still parameters and weights and all this other stuff that I haven't even talked about yet.
This is like God, right?
This is like, well, like how is this even a thing is so mind blowing to me.
It's mind blowing. It's absolutely mind blowing.
And I think that, um, you know, I think, yeah, you know, I don't hang around like 18 year olds.
I think they're using it for school.
So I think they get it.
I think I know a little bit about it because I hang out with smart people and I'm on the outskirts of like what these guys are doing.
So I kind of see it online.
I play with it for the average Joe for my mom and dad.
For a 35 year old who isn't like tech savvy, who just works as a mechanic.
I don't think that they're using it this way.
I don't think they're using it at all.
And it's going to change everything.
It's just like so crazy.
Like when the average Joe starts getting into this, I think young people like a 21 year old or something, I think it's like changing schools, by the way, it's like the grading system is like totally f -ed up, right?
Yeah. Like when I like think about this, I'm like, like, this is like there is no homework.
You can't do homework anymore.
You know what I mean?
It's like, someone DM me yesterday.
It's not just homework.
Someone DM me last night.
They were showing me, um, Sky Oliver, Oliver Han, he texted me this thing or did DM me this thing.
He said, uh, coding interviews, like so, okay, you school.
Yeah. Kids in school are using chassis.
We write essays and the teachers are like, fuck, how do we, how are we going?
It's a cat and mouse game to try to be like, Hey, how do I stop you from using AI to just like do your assignments?
Well, the same thing is true for coding interviews.
So coding interviews, which are used to hire programmers, there's this website, leap code wizard .io and basically it just helps you cheat on your coding interview.
So go, you got a coding test to get a job?
Just use this watch.
It'll write all. It's the same thing as a student will write the essay for you basically.
And uh, it's like, you know, doing 15 grand a month and recurring revenue.
I'm just helping people cheat on coding interviews.
This is insane. It's so difficult, right?
Uh, but it's kind of amazing.
How are you using this every day?
Um, like, let me just go to chat.
GPT, just tell you like my last piece chat GPT, your tool of choice, or do you like any of the other ones?
Yeah, it is my like default and then, you know, I play with everything else.
So usually if I'm like how factually correct does this need to be all perplexity.
So I go to perplexity if it's analysis, I'll use chat.
Have you used like the 01 stuff, like the deeper thinking stuff?
Only for 24 or 48 hours.
Yeah, it's brand new, but yeah, it's wild.
It takes a long time.
But it's wild. Well, yeah, that's the point of it.
It's basically if you told the computer, hey, you don't have to just quickly like again, shove an answer down my throat instantaneously where you're just predicting the next token and uh, good enough to go, right?
The 70 % chance it's this word.
Let's just put it in.
They found they could get, you could do more interesting tasks if you just said, Hey, take your time before you answer.
Give it more time to think and then it'll come up with a better answer.
It's temperamental, which is amazing.
Um, so I use that, but like check this out.
So there was this, um, press release recently for, uh, we were talking about IVF.
Remember, well, it's kind of this amazing thing.
I don't know if you saw it.
It's called Fertiloh.
Did you see what happened with this thing called Fertiloh?
So basically it was like the first live birth using eggs that matured outside the body.
So like if you've done IVF, it's like a pretty expensive and pretty like harsh thing on the body.
Like the woman has to get like injections, which are hormone injections.
Try to get your, they're trying to get your eggs to essentially, um, mature, be produced and mature inside your body.
And so what Fertiloh did was they were like, cool.
Instead of doing that, like long, expensive sort of hard on your body process, we can take an immature egg
