Everyone has predictions about AI, but what happens when AI predicts itself?
In this episode of The Newcomer Podcast, Eric is joined by Cerebral Valley co-hosts Max Child and James Wilsterman to let AI map out its own 5-year plan. From AGI timelines to AI in healthcare, movies, and beyond.Can AI really forecast its own future, or are we all buying into its self-created hype?
00:00:00
Humans have all sorts of opinions on where AI as a
00:00:03
technology is going over the coming years, but has anyone
00:00:06
asked AI how it feels about it? In today's episode, AI is making
00:00:10
its own predictions about the future and to come up with the
00:00:13
focus areas that are the most likely to see the highest levels
00:00:16
of growth. I'm joined by Max Child and
00:00:18
James Wilsterman who are my Co hosts at the Three Roll Valley
00:00:22
AI Summit. Today we're letting AI make its
00:00:24
own five year plan. This is the nuke of our podcast.
00:00:35
All right, we're back on the Newcomer Podcast.
00:00:38
I'm Eric Newcomer. I'm here with James Wilsterman
00:00:41
and Max Child. They're joining us for our
00:00:44
Cerebral Valley Podcast edition ahead of Cerebral Valley AI
00:00:48
Summit next week. Getting close last week, we
00:00:52
talked about some AI themes and did our draft pick for top AI
00:00:56
startups. This one we're looking at the
00:01:00
future. We're talking about AI
00:01:02
predictions. One year ago, Max James and I
00:01:04
reacted to predictions from Open AI and Anthropic and made bets
00:01:11
on those. So we're going to quickly review
00:01:12
how each of us did. So, you know, who's good at
00:01:15
making predictions and who's not.
00:01:16
And then we will do a new round of predictions for next year.
00:01:20
So I'm gonna hand it over to James, who is running, tracking
00:01:25
all this, the scorekeeper. James, how did we do last year?
00:01:28
Yeah, I just ran all these games for you guys.
00:01:32
Yeah, we, we had, we made predictions.
00:01:35
What was interesting last year is we had Claude and open AI
00:01:39
ChatGPT generate the predictions as well as give us a probability
00:01:44
estimate of how likely they would to be to occur within 12
00:01:48
months. So all three of us took overs or
00:01:51
unders on the probabilities and then we can check back now 12
00:01:56
months later to see how we did. So running through last year's
00:02:00
slate of predictions, The first prediction was phrased as will
00:02:05
open AI release GPT 5 exceeding 10 trillion parameters.
00:02:11
We are calling that not met did not happen.
00:02:15
The second prediction from ChatGPT National AI regulations
00:02:21
in healthcare diagnostics at By the end of this year, at least
00:02:25
three countries will have implemented these national
00:02:28
healthcare regulations governing whether you can use AI when
00:02:32
diagnosing illness at a 70% probability.
00:02:37
Also calling that not met, the third prediction was will Tesla
00:02:42
be approved for legal operation of unsupervised self driving
00:02:46
cars in one state with no drivers.
00:02:49
We are saying no that did not happen as well.
00:02:53
Open AI gave that a 40% probability.
00:02:57
The 4th prediction from ChatGPT was by the end of this year AI
00:03:01
generated content would be constituting at least 50% of all
00:03:06
news articles by published by one major news outlet, with a
00:03:11
30% probability where you're saying that was not met either.
00:03:16
No known major outlet has. I hear that newcomer media has
00:03:20
70% of its articles generated by AI.
00:03:23
The. Early.
00:03:24
I would like to, I would like to litigate that.
00:03:25
That's a major outlet. That's that's what I'm hearing
00:03:28
from the ground, Eric fact. One of our articles?
00:03:31
None. Yeah, none, none.
00:03:33
Oh. My God, none that you're aware
00:03:34
of. OK, the last problem, the last
00:03:39
prediction from ChatGPT was at 21st 25% probability that at
00:03:46
least one AI generated film, both the script and the visuals,
00:03:51
would be released in theaters. I'm kind of quibbling with the
00:03:57
the outcome of this one where there was at least one film
00:04:00
released in Turkey and maybe at least one theatre, but Max and I
00:04:04
have agreed to I'm a lot of. One theater owner in Turkey and
00:04:10
call that an over. That's absurd.
00:04:11
I'm sorry, no. I agree.
00:04:13
I agree. We're calling.
00:04:14
USA Films, come on. Yeah, we're saying that did not
00:04:17
happen, but apparently there is a film called Post Truth if
00:04:21
anyone wants to check that out in Turkey.
00:04:24
Only show in Turkey. Only yeah, if you happen to be
00:04:29
yeah, if you happen to be in Turkey, OK, If you're keeping
00:04:32
score at home, none of the predictions happened.
00:04:35
That is my my own take away. So AI's predictions about what
00:04:41
would happen in the world of AI, you know, it didn't say they
00:04:44
would all have them, but they all, you know, it gave a a.
00:04:46
Lot of them had none of them above 50% probabilities, you
00:04:49
know right? Right, so AI clearly thought
00:04:51
they would happen. Yeah.
00:04:52
So who? Who is the most foolhardy?
00:04:55
Who who thought too much stuff was going to happen?
00:04:58
It's very straightforward. Whoever guessed who's the most
00:05:00
upbeat did the worst, and whoever is the most pessimistic
00:05:04
did the best. So who?
00:05:06
Who? Yeah.
00:05:06
What are our dispositions? I will I'll read off some scores
00:05:10
here. We, we gave each other some
00:05:13
points or took away points depending on yeah, whether we
00:05:17
took the over under on the probability.
00:05:18
And I came in last place with -.2.
00:05:27
I'm sorry. I was so I was super.
00:05:29
I was way, way too overly, overly optimistic.
00:05:33
In these patients that. Fits too many overs, too many
00:05:36
overs. Yeah, yeah, I'm just gonna
00:05:39
definitely adjust my strategy a little bit this year.
00:05:43
And then in second place with .7 was Eric.
00:05:49
So Congrats Eric on that. A respectable positive score.
00:05:53
Yeah, I know. Not not underwater.
00:05:55
How do you and then Max with 1.6 points in this probability
00:06:03
scoring calculus win wins the day.
00:06:07
Lots of good. Nothing ever happens.
00:06:09
Yeah, nothing ever happens. Amigos.
00:06:13
I'm excited for this year's prediction.
00:06:15
Are you guys ready to play again?
00:06:17
Yes. I feel like we've all been
00:06:19
chastened. We've.
00:06:19
All learned our lessons yes all right, so a little intro for
00:06:22
this year's predictions. We asked 3 brilliant math
00:06:26
Olympiad winning a is Chachi BT Gemini and Claude to make
00:06:31
predictions of what will happen in the next 12 months in AI.
00:06:34
We asked them to assign probabilities, which is
00:06:36
important to understand. You know it thinks something's
00:06:39
unlikely. It's thinks something's very
00:06:40
likely. We take the over under on those
00:06:41
probabilities and then we get points based on how accurate we
00:06:44
are against those overruns. Obviously last year the message
00:06:48
was nothing ever happens. It'll be interesting to see if
00:06:51
in the next 12 months something does happen.
00:06:53
All right, let's kick it off with ChatGPT.
00:06:58
I used ChatGPT 5 Pro for these by the way, just in case you
00:07:02
wanted to understand. What a model?
00:07:04
And the first prediction? We're arguing that by this time
00:07:08
next year, the FCC will adopt A policy requiring on air
00:07:14
disclosure when AI tools are used in political ads.
00:07:19
Under Easy what? Should be.
00:07:22
The probability is 65%. Under.
00:07:27
Under I just think the Trump FCC is very pro AI so unless unless
00:07:34
something really happens they're not going to do anything AI or
00:07:36
regulatory. If it was a Harris
00:07:37
administration or something like that then that would be
00:07:39
possible, but 65 is a crazy over.
00:07:43
Right. Yeah, I'll take the under, I
00:07:45
think. I think if.
00:07:46
Yeah, to your point, like what is what is the Trump
00:07:48
administration doing at all? I mean, there's no actual laws
00:07:52
being passed, right? I mean the.
00:07:54
The government is not. Even shut down, yeah.
00:07:55
They're. They're trying to get Jimmy
00:07:57
Kimmel off the air. They're like fewer humans.
00:07:59
They're like like less. Humans on the.
00:08:01
Air more AI, baby. Yeah.
00:08:03
But they use one of those canes and they pull him off.
00:08:06
They don't actually pass a lie. Fair enough, but I think I think
00:08:11
they're like the AI is very pro Trump, let's keep it on the air.
00:08:15
Yeah. All right.
00:08:17
Well, that was a quick one. Easy, easy, easy start.
00:08:20
Moving on. OK, this is interesting
00:08:23
prediction from ChatGPT. By this time next year, major
00:08:27
outlets report, open AIS annualized revenue has hit a run
00:08:32
rate of over 30 billion. Oh oh man, I need to talk to my.
00:08:37
What's what's the probability here?
00:08:40
You need to talk to your source on the inside there.
00:08:44
It's like. Well, TomTom covers this closely
00:08:47
for us. I'm like, I don't track.
00:08:49
I thought you were texting Sam or something.
00:08:51
Yeah. We no Sam, well Sam that I
00:08:54
could, but not that he would help.
00:08:56
But like Sam, he went on that, he went on that interview with
00:09:01
Brad Gerser. Brad Gerser and was.
00:09:03
Saying they. Were over 13 already.
00:09:05
Let's see. You guys want to guess guess the
00:09:09
line here. 70 I don't know. 60% very.
00:09:14
Aggressive. OK, All right, pretty.
00:09:16
That's a pretty good guess. Yeah, it was I'm, I'm going to
00:09:19
take the over at 60 because I think they're probably at least
00:09:23
at 15 right now and doubling in a year.
00:09:26
If they don't double in the year, the entire U.S. economy is
00:09:29
going to collapse. So we're not going to be we're
00:09:32
not going to be having a predictions podcast if they get
00:09:34
2X in the next 12 months, they. They don't have to say it right,
00:09:38
it just has to be reported that they like.
00:09:40
Yeah, reported somewhere. Yeah, reported that it's rough.
00:09:42
Then it's, you know, probably over 30.
00:09:46
So Max, you are taking the over despite, despite every previous
00:09:50
prediction last year landing in the under?
00:09:52
Well, I just think again, as a connoisseur of arr growth rates,
00:09:56
as a voice AI startup founder James and I, Bali Incorporated,
00:10:02
you have to double every year basically to be a quote UN quote
00:10:05
high growth venture startup, even at the open AI level where
00:10:08
they're trying to do this IPO at a trillion dollars.
00:10:10
And so I think to our earlier discussion, Sam Alman literally
00:10:14
said on a podcast, they were over a $13 billion rate.
00:10:17
So I will give them at least a 15, and I think that might be
00:10:19
low. And if they don't double in a
00:10:21
year, yeah, I mean, the the US economy and the entire chips and
00:10:25
energy story will probably collapse.
00:10:27
So this this podcast will be cancelled so it'll must be
00:10:31
hosting. I guess what?
00:10:32
We did, but there has to be. Yeah, nobody.
00:10:34
Reporting lags like the reporting lags I mean.
00:10:37
OK, if we're when we report. Again.
00:10:38
When is it gonna be reported? Yeah, I don't know.
00:10:42
I. Mean you're sitting here in this
00:10:43
chair in a year, you know, what's the, what's the, what did
00:10:46
Sam just say on on the podcast? Right.
00:10:48
Yeah. Look, I'm gonna, I'm gonna go
00:10:50
with, but I believe to be the truth, which is I believe they
00:10:53
will be at over 3012 months from today.
00:10:55
It's a little bit hard for me to start hedging this prediction,
00:10:58
but waste on whether or not it's reported at that point or not.
00:11:01
So I'm just gonna just gonna go with what I believe to be the
00:11:04
reality 12 months from today. Yeah, I'm gonna, I'm gonna snag
00:11:07
the under just just. To nothing ever happens.
00:11:11
Yeah, just. What is your reasoning?
00:11:13
Learned my lesson. Or is you like is it the
00:11:16
reporting or you believe they will actually be under 30 at
00:11:18
that point? I think it'll be close no matter
00:11:22
what. I think that, well, yeah, I
00:11:26
think there's a lot of competition heating up in this
00:11:28
in, in AI And, you know, we don't know what, how how we
00:11:32
don't know how how much demis cooked over at Google yet for
00:11:38
Gemini 3. So let's let's, yeah, let's see
00:11:43
how it plays out. I am going to take the under.
00:11:47
I'm chastened, you know, Maybe all over correct.
00:11:50
I mean, you know, maybe this is the year that things happened,
00:11:54
yeah. No, I mean I.
00:11:55
Think it's I. Think I think the over is
00:11:58
totally a valid pit play. OK next prediction.
00:12:01
This one's from ChatGPT 5. Thinking meta in their 10K or
00:12:07
earnings or in various major news outlet sources report 2026
00:12:16
shipments for Meta's display equipped glasses to exceed
00:12:20
1. Meta Ray ban display 799.
00:12:25
Ugly. You can get them in in sand or
00:12:29
in black under. I'm going to grab the under.
00:12:32
Yeah. I, I just, I think this is in
00:12:35
the Nothing ever happens category.
00:12:37
I think they might try to like juice shipments in sort of some
00:12:40
sort of bizarre way, but I the reviews are not very good once
00:12:44
people actually have. Them.
00:12:45
Really. Yeah, there were these early try
00:12:48
on reviews where they were like The Verge can try them for three
00:12:52
minutes at the meta conference, and they were like, my God,
00:12:54
these are amazing. But I've started reading reviews
00:12:56
of people who actually bought them, like a normal person at
00:12:58
Best Buy, and they're like, actually, that's kind of stuff.
00:13:02
I will say holding my baby, this is going to be terrible.
00:13:05
This is what the but it's just like, you know, eventually you
00:13:08
want to like scroll at some point you're like, Oh my God,
00:13:11
like how are you feeding this thing for like hours on end
00:13:14
this. Is how I got to do nice TikTok
00:13:16
as I've explained many times, the 3:00 AM one, Airpod and
00:13:21
TikTok scrolling while feeding my.
00:13:22
Baby, I feel bad, like I want both hands feeding the baby, you
00:13:26
know, it's like. No, you'll get.
00:13:27
You'll get over that part pretty quickly.
00:13:31
No, you'll you'll, you'll get to the you'll get to one hand
00:13:33
bottle bottle nestling mindset and then you can scroll with the
00:13:39
other hand. It's great.
00:13:40
Right. Yeah, I'm doing it already.
00:13:41
But so you're gonna buy milk? How many are you gonna buy,
00:13:44
Eric? I said the under still I'm I'm
00:13:47
still saying the under and I'm not gonna buy them and I'm
00:13:50
pretty early adopter. Yeah, we both grabbed the under.
00:13:52
James, what are you doing? No, I'm grabbing the under as
00:13:55
well. I just don't.
00:13:56
I think it's too complicated to buy this product.
00:13:58
Like we've all been chastened This, our performance last year,
00:14:01
is gonna ruin this. We were.
00:14:02
We were having more fun last year.
00:14:04
We were. Yeah, last year was like the.
00:14:06
Trump got excitement. Naive and optimistic.
00:14:09
Now we're like the world is. I took sad.
00:14:13
I took the. Over on Open AI revenue, for the
00:14:15
record, you guys are three straight under so far.
00:14:18
This is a tough hang. For anyone running a startup
00:14:21
right now, Mongo DB for Startups is one of those programs that's
00:14:24
looking to make the life of a founder easier.
00:14:27
You get free credits on Mongo DB Atlas, so your products ready to
00:14:31
scale from day one without worrying about database costs.
00:14:34
They also throw in 200 million Voyage AI tokens so you can
00:14:38
start experimenting with AI features right out of the gate.
00:14:41
You'll get one-on-one time with actual Mongo DB engineers,
00:14:44
people who can help you troubleshoot to optimize and
00:14:47
think through your next stage of growth.
00:14:48
If you're building, scaling, or just trying to stretch your
00:14:51
runway a little longer, Mongo DB for startups can actually help.
00:14:55
Head to mongodb.com/startups to learn more.
00:15:00
Let's move on. Let's grab 3 predictions from
00:15:03
Claude this time. This time I used Claude 4.1 Opus
00:15:10
to generate some predictions. I hope you enjoy them.
00:15:14
Let's see, Claude says Apple will acquire an AI company
00:15:18
valued at over 1 billion dollars, 70% probability of this
00:15:24
occurring by this time next year.
00:15:27
Oh man, I was ready to take you over.
00:15:29
But the 70%, yeah, 70% come. On that's crazy.
00:15:32
Hi. Apple's only made like 1
00:15:35
acquisition in history over a billion dollars, which was
00:15:37
Beats, and I think they look back on that as a mistake.
00:15:41
Yeah, that seems really high. Yeah.
00:15:44
I mean, I think they're like, well, we needed to build Apple
00:15:46
Music. Something, something, something
00:15:48
whatever. We we make some money on the
00:15:49
head, there's. Still, why they wouldn't still
00:15:51
use the Beats brand if they thought it was a mistake?
00:15:54
I don't. No, I think they knew they had
00:15:55
to make money on beats. I don't think it's been a
00:15:57
massive financial mistake, but I don't think anyone's looking
00:15:59
back on that and being like, that was a banger.
00:16:01
Like we couldn't have bought Apple Music in house.
00:16:03
You know, I'm just, you know, it wasn't an OK, it wasn't like
00:16:06
some smoking hole in the ground, but it was I'm taking the.
00:16:09
Under just the the odds are terrible.
00:16:12
It's like it'd be exciting, you know, for, I mean, a billion
00:16:15
dollar acquisition isn't that big, you know, anymore and Trump
00:16:19
administration makes it easier. So the barriers are low.
00:16:22
But yeah, it would be once every couple years Apple does
00:16:25
something notable like that, so. Yeah, it's, it's just like the
00:16:29
track record of not doing it. It's hard to take the overlist.
00:16:33
70% you're literally betting 70% over on something that has
00:16:37
happened once in the last 20 years.
00:16:39
Like I mean, that's the, the base rate prior there is just
00:16:44
really low. I mean it I, I'm not saying it
00:16:46
won't happen, but it's just tough to to deal with that.
00:16:49
All right. I'll take the under as well
00:16:51
attempted, but yeah, the base rates not good.
00:16:56
All right, next one from Claude. The EU will find at least one
00:17:02
major AI company over €500 million for violating the AI Act
00:17:08
60% probability. God, these probabilities are so
00:17:13
fucking high. I'm sorry, the 60%, I mean like
00:17:16
the I'm ready to go over. I'm I want to go over.
00:17:19
I'd want to be tempted by it. Over we're going to we're going
00:17:21
to strike the AI act limitation, right for our purposes, I feel
00:17:25
like that's. Sure, too great.
00:17:27
I don't know what that means. I mean look, the EU loves
00:17:30
nothing more than fining large American companies for dumb
00:17:34
shit. That's literally what they get
00:17:36
up every day to do in Brussels. And yet 60% for 1/2 billion
00:17:41
dollar fine is that's high. That is high.
00:17:45
I mean, like, I feel like Mark Zuckerberg's got to do something
00:17:48
really dumb and they've got to move on it very quickly, so.
00:17:51
I mean, there's going to be all this copyright stuff.
00:17:54
What's the number? It's 60.
00:17:55
I feel like I should get rewarded more.
00:17:57
I I feel like they're, they're not giving us enough benefit for
00:18:01
some of these. The numbers are too high.
00:18:03
OK, I'm taking the over. I'm taking the over. 5. 100
00:18:07
million I need to I need to have somewhere where I need to you.
00:18:10
Need to live a little? You need to live a little.
00:18:12
Yeah, exactly. I'm gonna I've gotta grab the
00:18:15
under here because I just think that the EU wants to not be a
00:18:19
failed. I was gonna say nation state, a
00:18:22
failed continent state or whatever the hell they're doing
00:18:25
over there. But in AII think there's a
00:18:29
little bit more fear that they're just falling
00:18:31
fundamentally behind the rest of the world here.
00:18:32
And so I, I, I don't think coming out with a banger fine in
00:18:35
the next 12 months is, is what they're looking for.
00:18:37
I. Think we could have a lot of
00:18:38
copyright action. We saw the hysterical sort of
00:18:42
plagiarized every copyright and then pull it back and, and
00:18:47
there's no, you know, there's all this hostility with Trump.
00:18:49
So it's like, oh. So what?
00:18:50
What are you guys thinking? Foreign Affairs.
00:18:53
I'm digging the over. He's taking the over, yeah.
00:18:55
OK. I'll take the under as well.
00:19:03
Just going going full under under check.
00:19:07
He's all under. Like it's just like, learned
00:19:09
your lesson. Yeah, you get kicked off the
00:19:13
podcast if you take no overs on. It's not enough good.
00:19:17
Yeah, not enough good drama if you don't take some overs, you
00:19:20
know? This one's interesting and you
00:19:22
know kind of it's AI generated music will chart in the Top 40
00:19:27
of the Billboard hot 140% probability I.
00:19:32
Got to take the over on that I feel like.
00:19:35
Yeah, AI, purely AI generated music Top 40.
00:19:39
Yeah charts in the Top 40. What hold on, it doesn't say
00:19:42
purely. I just, I don't want to like get
00:19:44
into too much litigation here, but.
00:19:46
It but it needs to be mostly like the vocalist vocal some.
00:19:50
I mean, there's a, there's a. Little bit of opponent needs.
00:19:52
To be no, not just like no, no, no.
00:19:54
I feel like the vocals like the. It says AI part of it.
00:19:59
It says a song where AI demonstrably generated the
00:20:02
melody or the lyrics. It's not just production or
00:20:06
mixing. OK, so if it wrote it, but
00:20:09
there's a vocalist singing it, we'll count that.
00:20:11
OK, Yeah. Or if the vocalist melody, the
00:20:13
vocalist, yeah, they I think they're like a major part of the
00:20:16
song is AIII. Grabbed the over because this is
00:20:19
being reported today that there's already AI music being
00:20:23
charting in the top 100 so all you need is like a good enough
00:20:26
song to get to the top four. Wait.
00:20:27
What is that? Yeah, there's, I will pull up
00:20:31
the name of the artist, but it was very recently reported as in
00:20:34
the last like three to three to five days.
00:20:36
Yeah, Xavier Monet is charting quote UN quote, whatever that
00:20:39
means. Maybe not top 100, but let's say
00:20:41
top, you know, 500 or something is an AI generated artist is now
00:20:46
quote UN quote charting already. Top 40 is obviously high bar and
00:20:51
the music industry kind of fucking hates these people.
00:20:54
So there's definitely some entrenched interest against it.
00:20:56
But I feel like now that we're seeing AI music actually
00:21:00
charting, I feel like Top 40 is is doable at 40% in the next
00:21:03
year and overs are more fun. Guys, come on, get in on.
00:21:08
Get in on the. Over train.
00:21:09
I just I just. Took it over.
00:21:11
I'm going over, I'm going, I'm jumping on, I'm jumping on the
00:21:14
over train. I'm with you, Max.
00:21:18
Like it just seems the momentum is real behind this occurring.
00:21:24
And yeah man. If this alone happened, it's
00:21:28
pretty revolutionary for all the, you know, AI haters
00:21:33
watching this and be like, oh, the bullish AI people are having
00:21:36
to eat their get their comeuppance or whatever.
00:21:40
I do think, man, if this one happened alone, if you, if you
00:21:44
said that, you know, five years ago, you'd be like, what kind of
00:21:47
lunatic? This article from Billboard says
00:21:50
at least six AI or AI artists have debuted on various
00:21:54
Billboard rankings in the past few months.
00:21:58
That's. Hilarious that's but we're we're
00:22:02
talking about the main top. Top 40 is a is a bigger stretch
00:22:06
for sure. No, I mean you definitely have
00:22:08
to believe in an exponential progression here to get from
00:22:10
random Billboard charts to the Top 40.
00:22:12
But. Xavier Monet became the first
00:22:15
known AI artist to debut on a Billboard Radio chart, debuting
00:22:20
at #30 on the Adult R&B chart. But that's like a subcategory.
00:22:26
These are different charts, yeah.
00:22:27
So the main the cool one is getting on the Hot 100 I guess.
00:22:32
Yep, this is probably an aggressive prediction, but it
00:22:34
feels. Yeah, I'm taking the under.
00:22:36
It's in play. All right.
00:22:38
The under. I'm taking the over, Max is
00:22:41
over. Eric is under a major research
00:22:43
lab will demonstrate an AI system performing novel
00:22:47
scientific research that gets published in Nature or Science.
00:22:51
35% probability. I mean, I think that there will
00:22:55
be a scientific research discovery that is capable of
00:23:00
being in those journals. I'm just adjusting based on my
00:23:04
experience being married to an academic, which is that it takes
00:23:06
a long time to get published in journals.
00:23:10
That says. And novelty.
00:23:13
Novelty is hard, yeah. It's it says an AI system will
00:23:17
be credited as having autonomously designed and
00:23:20
interpreted an experiment that leads to a peer reviewed paper
00:23:24
in Nature or Science in the main journals.
00:23:28
Novelty will be amazing, right When when these models can
00:23:31
actually make a discovery on their own.
00:23:34
We haven't, we haven't seen that really happen.
00:23:37
So I'm skeptical. Even putting aside the
00:23:40
bureaucracy of the journals, I'm skeptical that the models would
00:23:43
produce that level of novelty in. 12 I'll go out on a limb and
00:23:48
take the over on this one just because I think there's a lot of
00:23:52
incentive for the labs to make this happen because of people
00:23:57
like us asking questions. That was my movie thesis, that
00:24:00
river. Yeah, that's right.
00:24:02
That. Runway.
00:24:03
Runaway would insidious device to go make AAI movie?
00:24:06
Yeah. Well, they bribed someone in
00:24:09
Turkey to show it in two theaters and you know nothing
00:24:11
happened. All right, let's move on to some
00:24:15
predictions from Gemini 2.5 Pro. Yeah, three more to go.
00:24:20
Prediction. By this time next year, a major
00:24:23
US based publishing house will be awarded a judgement of
00:24:31
damages against a major generative AI company in excess
00:24:35
of $50 million for copyright infringement.
00:24:41
Let's include settlements, right?
00:24:43
Yeah, settlements reported. Settlements sounds good as well,
00:24:46
probability 40%. So Entropic just had this $1.5
00:24:52
billion settlement for stealing all the books, right, For
00:24:55
training data? Wasn't that 3000 per book or
00:24:58
something? So you said 50 million as the
00:25:01
number. Yeah.
00:25:02
Right, so I'm taking the over. That thought crossed my mind
00:25:05
too. That settlement was insane so.
00:25:08
Yeah, I don't know. 50 million's low. 50 million feels very
00:25:11
doable for and given that there literally was a settlement on
00:25:16
this exact topic of stealing books, it was 40% over, right?
00:25:21
I mean, I feel like the over sounds good to me.
00:25:22
If that's the precedent, I feel like that would have counted.
00:25:25
So we're like one for one on that having happened in the last
00:25:28
12 months in my opinion. I'm picking the over like I
00:25:31
there was already a huge settlement.
00:25:33
It's a tiny amount of money in the scale of AII think yeah,
00:25:35
this is definitely going to happen.
00:25:37
I'm taking the over as well. We haven't learned our lesson,
00:25:41
folks. We're taking overs again.
00:25:43
We go on long enough in there. I'm I'm I'm on tilt.
00:25:47
I've taken 3. Overs 3 straight overs.
00:25:50
James want to bend. Into it, yeah.
00:25:53
All right, next prediction from Gemini, Apple will launch a new
00:25:58
version of iOS that introduces an on device large scale
00:26:02
multimodal AI assistant capable of executing multi step cross
00:26:08
application tasks such as summarize the last five emails
00:26:12
from my boss or draft a calendar event for Friday with the
00:26:17
relevant subject lines as the notes, right.
00:26:21
Relevant. Okay, yeah, no, no way.
00:26:23
They're doing this in a way that users actually like and people
00:26:25
find. Helpful.
00:26:27
I'm gonna go with over even at 80%, which I know is very
00:26:32
aggressive because I think this is literally the Apple
00:26:34
intelligence feature that they announced 18 months ago that did
00:26:38
not ship, that was famously delayed by a year, but that year
00:26:43
is supposedly coming to pass somewhere in the first half of
00:26:46
next year. So.
00:26:47
But even if I look at my like my contacts, when they try to use
00:26:51
whatever thinking like they're, they ruin everything.
00:26:54
With this I agree like it might. Like literally I have a list of
00:26:57
my favorites. You guys are in it.
00:26:59
Don't worry, don't worry. And like, that's nice.
00:27:02
The you know, it's just like they ruin it.
00:27:05
They make it shorter. It's like I have a nice long
00:27:07
list and you like decide to make four of them big.
00:27:09
Like how is that helping me? You know I.
00:27:12
I'm Apple is not doing a lot of things right right now, but I'm
00:27:16
just. You're like, they're going to do
00:27:18
it anyway and it's no. I'm like, they announced this
00:27:21
exact thing that James is describing and they failed to
00:27:25
ship it and they basically had a whole come to Jesus moment where
00:27:29
they basically had to like fire the whole team that was
00:27:31
responsible for this and replace it ironically with Gemini.
00:27:35
So maybe that's why Gemini's like got a little inside
00:27:37
information on this. So yeah, anyway, I'll take the
00:27:39
over. I just find it hard to believe
00:27:42
they're going to delay it more than another year.
00:27:44
And I think that the what they've, what they've announced
00:27:47
satisfies these obligations to the prediction.
00:27:49
All right, you guys. Have to decide though.
00:27:51
Yeah, I already didn't already say I took the under on it.
00:27:54
You took the under, I'll take the Yeah with Max on.
00:27:58
I think we have to have some rigor.
00:28:00
As long as we have some rigor around, like they can't just say
00:28:02
we did it. Like it needs to be something
00:28:04
people actually use and don't just like turn off.
00:28:07
Like somebody has to find some value in it like.
00:28:10
I actually think this is a pretty interesting prediction to
00:28:12
debate because 80% is a really high probability, right?
00:28:17
So like, we're James and I are like, this is fucking happening,
00:28:20
you know, like. We're saying they're destroying
00:28:23
their phone, my phone with the dumb shit that they're doing.
00:28:26
Are they really gonna be so consumer unfriendly to do
00:28:30
complicated multi stage things? They'll like destroy the user
00:28:34
experience? I, I don't we believe in Apple
00:28:37
enough to say, hey, and also if the bubble burst and they're
00:28:40
like, hey, we were, we were high on kool-aid with everybody else.
00:28:43
We're we're not going to be those people.
00:28:45
I don't know. A lot could happen, you know?
00:28:47
I'm interested to see how this prediction goes.
00:28:49
All right, one more James Gemini 1.
00:28:51
More one more Gemini, one more unscripted AAA game.
00:28:55
The prediction from Gemini is that at least one AAA video game
00:29:01
having a marketing and development budget over $100
00:29:05
million will be released in 2026 that has prominent real time
00:29:10
generative AI powered features such as dialogue.
00:29:14
Real time is interesting. Probability.
00:29:16
Probability 35%. I mean, I gotta grab the under
00:29:20
on that I just. James and I work.
00:29:22
James and I work in gaming. If there's anything we know
00:29:25
about gaming, it's that AAA studios take almost no risks
00:29:28
with these games. When we see the indie game
00:29:31
first, like the indie games haven't done.
00:29:33
This Yeah. Man, I wonder who's gonna make
00:29:35
an indie game with generated dialogue in real time?
00:29:38
Launching Launching sometime in November.
00:29:40
Tee this up you guys like. Gemini generate this prediction.
00:29:45
Come in, come. The end of the podcast is like
00:29:48
coming from Bali. Yeah, I will.
00:29:51
Say, I will say, as someone who's intimately familiar with
00:29:55
using real time generated dialogue to make an indie game,
00:30:00
it's still harder than you think to make it generate interesting
00:30:03
stuff for every player. I mean, I think Fortnite had
00:30:08
some generative feature at one point last year, right?
00:30:12
Or was it early this year where it was like Darth Vader was
00:30:15
using 11 labs and generating stuff on the fly and.
00:30:20
Yes, yes, I got Vader. Lead your squad to victory,
00:30:23
Holy. Shit, Vader, we took a blinker
00:30:25
with me. A blinker jerk mate.
00:30:29
This some basic mating ritual I am unfamiliar.
00:30:31
With so I don't know, I would have counted that and I guess
00:30:34
right. I mean, I guess really one day
00:30:36
it was a one day thing and then they turned it off.
00:30:38
It's like me. Oh, that's yeah.
00:30:40
No. That wouldn't count for this
00:30:43
prediction, right? What's it say?
00:30:44
Like a major feature or something?
00:30:46
Wow, it's like a new $100 million AAA game, I thought, or
00:30:49
something. I'm saying prominently,
00:30:51
prominently features real time. So I don't think that would have
00:30:53
counted. But if that had been like a
00:30:55
permanent new feature of Fortnite?
00:30:57
In a big part of the game, yeah, yeah, yeah.
00:30:59
Yeah, OK. I mean, yeah, so there, there's
00:31:02
room for. The over.
00:31:02
So you're taking the over. I'm just saying this is like not
00:31:05
out. Are you taking out of the are
00:31:07
you? I'm gonna, I'm gonna take the
00:31:09
under because that Fortnite, I think completely flopped and
00:31:12
failed and like, they shut it down and, you know, Darth.
00:31:15
Vader was, and they said instantly, the world's
00:31:18
expensive. Instantly became racist.
00:31:20
Yeah. Like I think we're, we're gonna
00:31:22
take the under in that we're we're all taking the under.
00:31:25
I gotta take the under. I just think these are such risk
00:31:27
averse companies. And to your point, the one risk
00:31:30
taking in this industry, the one risk that was taken in this
00:31:33
industry did not materialize in a positive way.
00:31:36
So I don't know. Yeah, I'd love to be wrong.
00:31:38
For the record, that would be great if people would start.
00:31:40
Being more AI is more bullish about AI than we are.
00:31:44
I hope somebody does do this. I think it'd be awesome.
00:31:46
All right, and that wraps the predictions episode.
00:31:50
Nice check back. Good one next year.
00:31:52
Yeah, great. Our next episode we will be
00:31:56
bringing you the biggest hits from the street Bull Valley AI
00:31:59
Summit, which is on November 12th.
00:32:01
We'll be bringing some of our on stage interviews, our favorite
00:32:04
ones right here on In the Newcomer podcast and in your
00:32:07
podcast feeds. Thanks Max and James, go play
00:32:11
song quiz, go play some volley games on your Roku TV or where
00:32:16
else? We're gonna actually show you
00:32:17
our real time AI game at some point and then you will have
00:32:19
something interesting to say. Is it out company?
00:32:21
Is it out? Is it out?
00:32:23
We can play a beta right now, so let's set it up.
00:32:25
Great. Yeah.
00:32:26
Sounds good. All right.
00:32:26
Bye. Bye.
00:32:27
See you guys. Thank you for tuning in to this
00:32:30
week's episode of the podcast. If you're new here, please like
00:32:32
and subscribe. Really helps out the channel.
00:32:35
Listen to new episodes every week wherever you get your
00:32:37
podcasts.