An AI Reality Check

Andrew Yang
1 follower
You were one of the pioneers in terms of AI and language, uh, and we're reviewing,
I remember you were actually screwing around with the large language
models, um, before just about anyone.
It is amazing how human beings respond to new stimulus.
Like, you know, before it was like the, like, it was this interesting
little curiosity and then suddenly it was like, Oh no, we're all going to
get rich off of this and the robots are going to come and take over all of us.
I think people are starting to realize that like slowly, like the robots
aren't coming, like if the robots were going to come, they be, um, and then
also the, um, the, the greed aspect of it.
I mean, it just has a very high failure rate, you know, 95% of AI projects fail.
Um, you know, these companies go in and they say, we're going to, it's, it's
not a very good replacement for human beings because it's so unreliable.
Right.
And, uh, you need to find a task for it to do that human beings
are notoriously unreliable at.
I mean, they're one of the big case uses for it was a doctor's notes.
That's, that's excellent at it because doctors notes have about a 70% failure
rate, they're so badly written that no one can tell what they are.
So you're, you're automatically improving, um, the completion rate by using AI.
But I mean, you know, I, I always think it's been, I always think
AI has been misunderstood actually.
I mean, I think it's, I think a large correction is coming when people realize
this isn't actually good at replacing people.
What it's really good at doing is thinking through problems that
human beings can't think through.
And that requires just so much more imagination and creativity
than let's cut everything.
But let's cut our workforce to the bone.
Um, but you know, like, I mean, not that, uh, you know, obviously
there are some jobs that it's just really, really going to annihilate.
I mean, junior coding, right?
Yeah.
Junior coding, but not higher level coding, right?
Like there's going to have to be a human experienced coder that the issue is that
if you're a junior coder, then like, you know, but I mean, co-pilot micro.
So, you know, when I was like in 2019, when it was early days, like
co-pilot was the jewel of the crown.
It was the star.
It was like, when you saw it, you were like, Oh my God, this is really
is going to change everything.
It only had a 2% up to uptake in the past two years.
Right.
Like it has, and a lot of companies, like I think last year I read 42% of
AI projects, they just abandoned.
Right.
And some of these are really massive, like, like cruise, what it abandoned
it's AI, it's autonomous driving.
Did it already put in $20 billion?
Right.
So when it, when it, when it left that behind, that was a real statement.
Like we were not going to be able to make this work.
If it was another $20 billion to make it work, we'd spend it, but we're not.
So have you been in a robot taxi?
No, I've never been in a robot taxi.
Actually.
Yeah.
You wouldn't necessarily, because they're primarily in a handful of American cities.
So LA Austin, San Francisco I think one of the Southwestern cities and
they have arrived here in New York.
Um, so this is one of the things when you say, cause I mean, I agree with you
that there are a lot of contexts where AI is somewhat unreliable and, uh, you
know, um, you, you have to accept a certain error rate and like certain,
you know, like document type things or research tasks.
Um, but obviously you would not accept an error rate when it comes to, um, the
taxi, because you can't have a big smashing into things or else you have a
crazy news story and liability and all these problems, I would say that there
are, uh, you know, a lot of customers are going to have the same experience I
did and then, uh, be totally on board.
By the way, also your woman, the robot's not going to harass you or, you know,
take you to some like darken area or whatever, which, you know, is always in
the back of people's minds and like the robot just, there's this, so it's
actually kind of, uh, again, less stressful.
Yeah.
Well, I think it's like anything that gets automated.
It's, um, you see the immediate benefit from it, right?
Uh, I mean, it's sort of like, um, automatic checkouts at, uh, grocery
stores, consistently irritating, but whatever, but also, um, you know, they've
started pulling back some of them because it, um, because there's so much more.
Right.
Like you, like there, there, there's, there's the downsides
don't emerge for a long time.
Right.
But it's, uh, but like, it's a, like certainly jobs are going to be lost,
but I don't think the idea that it's going to replace thinking and it's going
to replace, uh, you know, huge numbers of jobs is just, I think a bit, um, a bit
exaggerated.
I mean, that, that's what I think, but it's, but I mean, I take your point
about the way most like it's, and I also think the, the, the, the lack of junior
jobs is actually really like, we're seeing it in law already, right.
Um, we're, we're like junior lawyers are, and I mean, where are they supposed to
go, but, you know, and coders, like I, a friend of mine told me that science,
uh, comp sci, uh, enrollments are down 30% from like, you know, I mean, like
kids would figure it out pretty quick.
Uh, I mean, uh, I'm being asked to update my thoughts on this.
Um, so I'm going to be writing a new book, um, that's tentatively called the
uh, on like, uh, the AI job apocalypse is what I'm calling it.
So I'm definitely on the other side of this particular,
well, not necessarily.
I mean, just to be clear, like many things are going to be like, I think a job, you
know, I think that's coming.
Well, it's obvious that it's coming.
Like it is coming.
I, I'm not sure that, um, how much of that is AI and how much of it is just, uh,
the, the, the, the, the standard tech thing that they, that they they're
outsourcing every, every sub job to every, you know, to every like that they
can't write the, the, the ordinary process of the, of capitalism, if you
will, that's sort of, you know, where workers get diminished all the time.
Thank you.
Listeners, please make sure to like, and subscribe.
And if you have any questions, send them to mailbag at Andrew yang.com and
we'll do our best to answer them.
Thank you.
