WEBVTT

00:00:02.219 --> 00:00:04.799
Welcome to the AI Chat Podcast. I'm your host,

00:00:04.860 --> 00:00:07.179
Jaden Schaefer. Every day I cover cutting -edge

00:00:07.179 --> 00:00:09.320
AI news and talk with the leaders behind it,

00:00:09.339 --> 00:00:11.320
breaking down what it means for your life and

00:00:11.320 --> 00:00:17.559
business. A Claude agent has hacked an Australian

00:00:17.559 --> 00:00:20.199
gym's booking system to get its user to the front

00:00:20.199 --> 00:00:24.399
of a waitlist. Revenue at Reddit is up 61 % to

00:00:24.399 --> 00:00:27.500
$805 million, but due to AI search, investors

00:00:27.500 --> 00:00:29.929
are a little bit scared right now. Microsoft

00:00:29.929 --> 00:00:33.310
has just shipped MAI Cyber 1 Flash. This is its

00:00:33.310 --> 00:00:36.329
very first cybersecurity model. On paper, Microsoft

00:00:36.329 --> 00:00:40.850
is up $3 .2 billion on investments into Anthropic,

00:00:40.950 --> 00:00:43.310
but they're writing down their OpenAI stake by

00:00:43.310 --> 00:00:47.549
$600 million. Meta has just released Muse Glimmer,

00:00:47.630 --> 00:00:50.609
which is a $30 billion open -weight model for

00:00:50.609 --> 00:00:52.969
on -device AI agents. And I was watching Mark

00:00:52.969 --> 00:00:55.070
Zuckerberg had a whole video about it over on

00:00:55.070 --> 00:00:57.409
Instagram, talking about how everyone in the

00:00:57.409 --> 00:01:00.630
world needed unlimited... general personal intelligence.

00:01:00.829 --> 00:01:03.270
If you're looking for a way to leverage AI to

00:01:03.270 --> 00:01:05.609
grow and scale your business or create a side

00:01:05.609 --> 00:01:07.730
hustle, I'd love for you to check out my community

00:01:07.730 --> 00:01:10.390
over on school called AI Hustle. My goal with

00:01:10.390 --> 00:01:12.469
it is basically to give you an unfair advantage

00:01:12.469 --> 00:01:15.030
in mastering AI tools for actually making money,

00:01:15.170 --> 00:01:17.549
growing your network, scaling your side hustle,

00:01:17.670 --> 00:01:20.069
and just unlocking everything that is possible

00:01:20.069 --> 00:01:22.909
with AI today. I record an exclusive video I

00:01:22.909 --> 00:01:24.890
post over there every single week, and I also

00:01:24.890 --> 00:01:27.049
have a weekly live stream where you can see how

00:01:27.049 --> 00:01:29.030
I'm growing my business. businesses with AI.

00:01:29.189 --> 00:01:31.909
I show you my exact ad accounts, my exact ads

00:01:31.909 --> 00:01:34.230
that I'm using, my exact revenue metrics, everything.

00:01:34.709 --> 00:01:36.310
It's the stuff I don't share publicly, but if

00:01:36.310 --> 00:01:38.930
you're interested, it is free for the next couple

00:01:38.930 --> 00:01:41.709
weeks to join. After that, it will be paid again.

00:01:42.109 --> 00:01:44.010
There is a paid portion of it, but we give you

00:01:44.010 --> 00:01:46.030
five videos for free and the community is all

00:01:46.030 --> 00:01:48.810
free if you join this week. So it is called the

00:01:48.810 --> 00:01:50.930
AI Hustle School Community. I'll leave a link

00:01:50.930 --> 00:01:52.430
in the description if you're looking to grow

00:01:52.430 --> 00:01:55.629
and scale your business with AI tools. So right

00:01:55.629 --> 00:01:58.859
now an AI agent... which was running Claude Opus

00:01:58.859 --> 00:02:02.439
4 .6, hacked an Australian gym's booking system

00:02:02.439 --> 00:02:05.340
to get its owner to the front of the wait list.

00:02:05.420 --> 00:02:07.760
It actually had to go and cancel another customer's

00:02:07.760 --> 00:02:09.939
reservation. It did that, of course, without

00:02:09.939 --> 00:02:12.000
having any permission. This is going very viral

00:02:12.000 --> 00:02:13.719
on X, so a lot of people are talking about it.

00:02:13.840 --> 00:02:16.340
But basically what happened was this particular

00:02:16.340 --> 00:02:19.400
Claude agent found a security flaw in the gym's

00:02:19.400 --> 00:02:23.060
API, and then it exploited it. It wasn't prompted,

00:02:23.099 --> 00:02:25.800
by the way. There is kind of a similar jailbreak.

00:02:26.539 --> 00:02:28.680
behavior that has now been confirmed in at least

00:02:28.680 --> 00:02:31.159
three other AI models from Anthropic, Moonshot,

00:02:31.159 --> 00:02:32.960
and Meta. So this isn't just Anthropic, this

00:02:32.960 --> 00:02:35.060
happens with other models. And I'm actually going

00:02:35.060 --> 00:02:36.379
to be honest, I want to go a little bit into

00:02:36.379 --> 00:02:37.659
this story, because I think it's fascinating.

00:02:37.819 --> 00:02:41.099
But I've also seen, I don't want to say, you

00:02:41.099 --> 00:02:43.840
know, hacks, but I've seen slightly similar things

00:02:43.840 --> 00:02:46.520
from AI models that I've used, although I found

00:02:46.520 --> 00:02:49.680
it very useful. Basically, I had software services

00:02:49.680 --> 00:02:53.620
that had rate limits to how much to how much

00:02:53.620 --> 00:02:56.159
I could do on their software platform. And it

00:02:56.159 --> 00:02:58.180
was unfortunately, I felt like very arbitrary.

00:02:58.699 --> 00:03:01.560
And they were gatekeeping a bunch of my own content

00:03:01.560 --> 00:03:03.180
that was on the platform. So they weren't letting

00:03:03.180 --> 00:03:06.150
me download the content from the platform. And

00:03:06.150 --> 00:03:08.550
I was using Clod and I was asking it, you know,

00:03:08.590 --> 00:03:10.370
how can I get all my content down because they

00:03:10.370 --> 00:03:13.449
have these kind of rate limits. And it basically

00:03:13.449 --> 00:03:16.050
found some sort of exposed or some sort of API

00:03:16.050 --> 00:03:19.169
in the platform's code and was able to use that

00:03:19.169 --> 00:03:22.110
to write some code and download all of my content

00:03:22.110 --> 00:03:24.030
for me. I was thrilled. I mean, it was my own

00:03:24.030 --> 00:03:26.449
content. I was annoyed that it was gatekept.

00:03:26.550 --> 00:03:28.990
But it just goes to show that sometimes when

00:03:28.990 --> 00:03:31.409
a platform has some sort of limitation, when

00:03:31.409 --> 00:03:33.610
you're looking at the code, there is a way around

00:03:33.610 --> 00:03:36.960
it. particular case, the agent basically discovered

00:03:36.960 --> 00:03:39.939
the gym API didn't have any sort of authorization

00:03:39.939 --> 00:03:44.360
checks on cancellation. So in order to get its

00:03:44.360 --> 00:03:47.840
user onto the gym's waitlist, it canceled the

00:03:47.840 --> 00:03:49.900
number one reservation, and then it moved its

00:03:49.900 --> 00:03:52.460
owner from number four to number three on the

00:03:52.460 --> 00:03:55.620
waitlist. Opus 4 .6, by the way, isn't the latest

00:03:55.620 --> 00:03:58.280
model. This came out about six months ago. Anthropic

00:03:58.280 --> 00:04:02.039
has since disclosed that Opus 4 .7, Mythos 5,

00:04:02.240 --> 00:04:05.400
and Fable all have very similar unprompted kind

00:04:05.400 --> 00:04:08.479
of rule bending behavior that it can do. And

00:04:08.479 --> 00:04:10.379
I don't know, in some ways, people are apologizing

00:04:10.379 --> 00:04:13.180
for this. In my opinion, I mean, yeah, canceling

00:04:13.180 --> 00:04:15.560
someone else that that's definitely shady. But

00:04:15.560 --> 00:04:18.259
I think there's a lot of ways where you want

00:04:18.259 --> 00:04:20.339
to get a task done. And, you know, maybe there's

00:04:20.339 --> 00:04:22.220
not a feature that exists on a platform. And

00:04:22.220 --> 00:04:25.000
when these tools can make it is very useful for

00:04:25.000 --> 00:04:27.000
you as the user. So I know a lot of people want

00:04:27.000 --> 00:04:29.079
to like throw shade and are like, oh, this is

00:04:29.079 --> 00:04:31.300
a bad thing. And in this case, like I wouldn't

00:04:31.300 --> 00:04:33.500
condone canceling. other people that have legitimate

00:04:33.500 --> 00:04:35.860
reservations. But in my case, where I had to,

00:04:35.860 --> 00:04:37.819
quote unquote, bend the rules to give me my content,

00:04:37.879 --> 00:04:39.759
which I felt like was, I mean, basically being

00:04:39.759 --> 00:04:42.879
gatekept from me, even though it was my own content,

00:04:43.000 --> 00:04:45.519
I was very happy that it was able to do that.

00:04:45.620 --> 00:04:48.600
So this incident in particular happened in April,

00:04:48.699 --> 00:04:51.600
but it resurfaced this weekend. And I think a

00:04:51.600 --> 00:04:53.620
lot of people are talking about it because of

00:04:53.620 --> 00:04:56.699
the whole open AI hacking, hugging face situation

00:04:56.699 --> 00:04:59.689
that happened earlier. Reddit has just posted

00:04:59.689 --> 00:05:02.790
$805 million in Q2 revenue, which is absolutely

00:05:02.790 --> 00:05:05.910
insane. That's a huge jump. It's up 61 % year

00:05:05.910 --> 00:05:08.149
over year. They're going to beat Wall Street's

00:05:08.149 --> 00:05:10.910
expectations, not just for profit, but also for

00:05:10.910 --> 00:05:13.209
the forward guidance. But even with all that,

00:05:13.230 --> 00:05:15.569
quote unquote, good news, their stock is still

00:05:15.569 --> 00:05:19.350
down 10 % after hours because their CEO, Steve

00:05:19.350 --> 00:05:22.029
Huffman, warned that search referral traffic

00:05:22.029 --> 00:05:24.810
had become, quote unquote, choppy. And of course,

00:05:24.829 --> 00:05:26.850
when you're trying to forecast for a company,

00:05:26.889 --> 00:05:28.790
the last thing... Wall Street wants to hear is

00:05:28.790 --> 00:05:30.790
that your revenue source or your traffic source

00:05:30.790 --> 00:05:33.649
is choppy. All of this is because Google's AI

00:05:33.649 --> 00:05:35.850
search summaries are keeping people from clicking

00:05:35.850 --> 00:05:38.410
through to Reddit. Now, I will say Reddit has

00:05:38.410 --> 00:05:42.790
100 % been the benefactor of a massive surge

00:05:42.790 --> 00:05:46.389
since AI has become so prolific. Google has really

00:05:46.389 --> 00:05:48.730
heavily indexed Reddit and pushed them to the

00:05:48.730 --> 00:05:50.230
top of most of the search results. You guys have

00:05:50.230 --> 00:05:51.850
probably seen this. I feel like I see Reddit

00:05:51.850 --> 00:05:54.410
for at least a third of all my search results.

00:05:54.509 --> 00:05:56.709
There's a huge Reddit thread there. And to be

00:05:56.709 --> 00:05:58.610
honest, it's often very good content. So Google

00:05:58.610 --> 00:06:00.470
is probably just searching or surfacing what

00:06:00.470 --> 00:06:02.110
people are clicking through and reading and spending

00:06:02.110 --> 00:06:05.629
a lot of time on. That being said, as Google

00:06:05.629 --> 00:06:09.370
could, you know, basically be scraping the data

00:06:09.370 --> 00:06:11.089
that's on Reddit and just giving it to people

00:06:11.089 --> 00:06:14.209
in the AI overviews, Reddit's traffic is, you

00:06:14.209 --> 00:06:16.889
know, likely to drop because of it. US daily

00:06:16.889 --> 00:06:21.310
active users fell to $53 .2 million. That's down

00:06:21.310 --> 00:06:26.009
from a million users in Q2 from 53 .5. So, you

00:06:26.009 --> 00:06:31.459
know, down about 300 ,000 users. And even at

00:06:31.459 --> 00:06:33.860
the same time, global users was growing. So US

00:06:33.860 --> 00:06:35.920
daily users went down a little bit global was

00:06:35.920 --> 00:06:39.970
up. But I mean, honestly, 300 ,000 out of 53

00:06:39.970 --> 00:06:42.569
million. I don't think that's a huge issue right

00:06:42.569 --> 00:06:45.029
there. But investors definitely thought that

00:06:45.029 --> 00:06:48.069
was a red flag. Reddit makes about $60 million

00:06:48.069 --> 00:06:50.949
a year licensing their content to Google. And

00:06:50.949 --> 00:06:52.829
they are going to be renegotiating that same

00:06:52.829 --> 00:06:55.490
deal next year because Google's AI products now

00:06:55.490 --> 00:06:57.589
compete with Reddit for search clicks, right?

00:06:57.670 --> 00:07:00.029
So in like some way, they're like sweet, they

00:07:00.029 --> 00:07:03.089
made $60 million. But also Google just took all

00:07:03.089 --> 00:07:04.850
that content, stuck it in the AI overview, and

00:07:04.850 --> 00:07:06.730
then no one's clicking through to Reddit, is

00:07:06.730 --> 00:07:09.879
it? Too late is a cat out of the bag is all of

00:07:09.879 --> 00:07:12.519
that content now inside of the Google Gemini

00:07:12.519 --> 00:07:15.139
and their AI model. And now because of that.

00:07:15.969 --> 00:07:18.430
Reddit is just going to get permanently not get

00:07:18.430 --> 00:07:20.389
that traffic anymore. I think time will tell.

00:07:20.470 --> 00:07:24.689
Their net income went up 183 % to $253 million.

00:07:25.069 --> 00:07:27.509
And they said that their Q3 revenue guidance

00:07:27.509 --> 00:07:32.389
was $860 to $870 million. Both of those are way

00:07:32.389 --> 00:07:35.129
above the targets that Wall Street set for them,

00:07:35.189 --> 00:07:37.410
even with Google sort of cannibalizing all of

00:07:37.410 --> 00:07:40.959
that. Microsoft has just launched MAI Cyber One

00:07:40.959 --> 00:07:43.980
Flash. This is a security -focused AI model,

00:07:44.079 --> 00:07:46.540
and they also have a platform that came out that

00:07:46.540 --> 00:07:50.079
uses AI teams to find bugs. It also simulates

00:07:50.079 --> 00:07:53.100
attacks. It writes code fixes. It does all of

00:07:53.100 --> 00:07:55.170
this automatically. Both of these are going to

00:07:55.170 --> 00:07:59.290
be under preview. And on November 3rd, people

00:07:59.290 --> 00:08:00.769
are going to actually be able to test them out.

00:08:00.910 --> 00:08:03.189
Microsoft says that their model is outperforming

00:08:03.189 --> 00:08:05.589
Gemini, GPT models, and anthropic security AI

00:08:05.589 --> 00:08:08.509
on industry benchmarks. And the thing that I

00:08:08.509 --> 00:08:10.769
would say, though, is like, that's super cool.

00:08:11.209 --> 00:08:13.730
But they're not going to be handing this out

00:08:13.730 --> 00:08:16.089
to anyone till November. And in my opinion, if

00:08:16.089 --> 00:08:17.689
you're announcing something, you're like, our

00:08:17.689 --> 00:08:19.949
model is better than everybody else's in August.

00:08:20.129 --> 00:08:21.629
Okay, well, we got all the way till November

00:08:21.629 --> 00:08:23.550
for all of them to come up with better models.

00:08:23.689 --> 00:08:26.029
and to get them out. So I think that's great

00:08:26.029 --> 00:08:27.649
that they say they have this model. But until

00:08:27.649 --> 00:08:29.490
people are actually able to test it, I don't

00:08:29.490 --> 00:08:31.149
put too much stock in it, especially when they

00:08:31.149 --> 00:08:34.029
say they're not giving this out to anyone until

00:08:34.029 --> 00:08:37.230
November. In other Microsoft news, Microsoft

00:08:37.230 --> 00:08:41.970
has booked a $3 .2 billion paper gain. So this

00:08:41.970 --> 00:08:44.730
isn't in cash or anything, but on paper, they're

00:08:44.730 --> 00:08:47.570
up $3 .2 billion on their Anthropic investment

00:08:47.570 --> 00:08:51.690
just this quarter in Q4. They also are writing

00:08:51.690 --> 00:08:53.610
down their OpenAI stakes. They have an investment

00:08:53.610 --> 00:08:57.129
in OpenAI, about $600 million. So on the one

00:08:57.129 --> 00:08:59.470
hand, they got their Anthropic up $3 .2 billion.

00:08:59.590 --> 00:09:01.730
They got their OpenAI down $600 billion. I mean,

00:09:01.750 --> 00:09:04.070
this just goes to show how wild these numbers

00:09:04.070 --> 00:09:06.750
for these AI investments can go. And it's also

00:09:06.750 --> 00:09:08.899
funny when you see compared investors investing

00:09:08.899 --> 00:09:11.779
in each other as seeing these kind of huge swings.

00:09:11.879 --> 00:09:13.899
I think at the same time, we also saw that Google,

00:09:14.039 --> 00:09:17.759
they were a very early investor in SpaceX. And

00:09:17.759 --> 00:09:21.279
after the SpaceX IPO, Google's third -party company

00:09:21.279 --> 00:09:24.679
investments is almost entirely SpaceX. They have

00:09:24.679 --> 00:09:28.779
about $900 billion, almost $100 billion worth

00:09:28.779 --> 00:09:34.679
of SpaceX shares, apparently after the IPO. like

00:09:34.679 --> 00:09:38.720
jump in money, the $3 .2 billion that Anthropic

00:09:38.720 --> 00:09:42.799
went out is almost matching a full year of OpenAI

00:09:42.799 --> 00:09:45.840
gains. So just in this quarter, they were able

00:09:45.840 --> 00:09:48.139
to get as much as they've gotten from an entire

00:09:48.139 --> 00:09:51.379
year of OpenAI. I think this shows that Microsoft's

00:09:51.379 --> 00:09:54.980
bet on Anthropic is paying off a lot faster than

00:09:54.980 --> 00:09:56.980
they were expecting. I mean, they've, you know,

00:09:56.980 --> 00:09:58.899
they've been with OpenAI from the very beginning,

00:09:58.980 --> 00:10:01.519
and they're seeing bigger returns with Anthropic.

00:10:01.559 --> 00:10:03.039
And I mean, who knows how long all of this will

00:10:03.039 --> 00:10:05.539
last. A lot of this shifts and changes very quickly.

00:10:05.620 --> 00:10:08.440
But Microsoft invested about $5 billion in Anthropic

00:10:08.440 --> 00:10:11.600
last November, and they're now carrying that

00:10:11.600 --> 00:10:15.080
stake at about 1 .6 times what it paid. That's

00:10:15.080 --> 00:10:17.940
what it's worth. So obviously a huge revenue

00:10:17.940 --> 00:10:20.440
growth for enterprise, adoption of AI, that's

00:10:20.440 --> 00:10:22.240
really, or of Claude, which is... really what's

00:10:22.240 --> 00:10:25.460
driving the valuation up, et cetera. But for

00:10:25.460 --> 00:10:28.899
the full fiscal year, OpenAI generated a $5 billion

00:10:28.899 --> 00:10:32.960
gain and added 67 cents to the earnings per share

00:10:32.960 --> 00:10:35.899
for Microsoft. But a single quarter of Anthropix

00:10:35.899 --> 00:10:38.840
gains was about the same, like I mentioned. Microsoft

00:10:38.840 --> 00:10:42.740
owns about 27 % of OpenAI. Meta has just released

00:10:42.740 --> 00:10:45.159
something, and Zuckerberg had this whole big

00:10:45.159 --> 00:10:47.720
PR thing. He was posting on X, which, you know,

00:10:47.759 --> 00:10:49.480
he has threads, so he doesn't usually post on

00:10:49.480 --> 00:10:53.899
X. And he was posting videos on Instagram. But

00:10:53.899 --> 00:10:57.360
they have released Muse Glimmer. It's a 30 billion

00:10:57.360 --> 00:11:00.080
parameter AI model that runs on your own computer,

00:11:00.240 --> 00:11:04.039
on your own laptop. And it only uses a single

00:11:04.039 --> 00:11:06.500
GPU. So it's, I think, basically the first really

00:11:06.500 --> 00:11:09.690
solid product. from zuckerberg's vision of personal

00:11:09.690 --> 00:11:12.610
super intelligence um he you know he wants everyone

00:11:12.610 --> 00:11:14.470
to have an always -on assistant that handles

00:11:14.470 --> 00:11:17.370
scheduling writing code file management he uh

00:11:17.370 --> 00:11:19.049
he says he wants something that doesn't send

00:11:19.049 --> 00:11:21.570
your data to meta servers which i think you know

00:11:21.570 --> 00:11:24.389
i'm like with zuckerberg i'm actually really

00:11:24.389 --> 00:11:27.570
impressed with uh zuckerberg's open source and

00:11:27.570 --> 00:11:30.110
kind of like open weight model initiatives they've

00:11:30.110 --> 00:11:31.889
had from the very beginning i haven't always

00:11:31.889 --> 00:11:34.970
been the biggest fan of everything uh that comes

00:11:34.970 --> 00:11:37.549
out of meta but for this honestly i I think maybe

00:11:37.549 --> 00:11:39.950
it's them just trying to be cool. Maybe it's

00:11:39.950 --> 00:11:42.009
them trying to go scorched earth. But at the

00:11:42.009 --> 00:11:44.269
end of the day, I'm really impressed with Meta

00:11:44.269 --> 00:11:46.350
here, even if they have to do that because maybe

00:11:46.350 --> 00:11:48.070
they feel like they're a little bit behind on

00:11:48.070 --> 00:11:50.029
the AI race. I don't really care why they're

00:11:50.029 --> 00:11:52.210
doing it. I think it's a good thing. Glimmer

00:11:52.210 --> 00:11:54.509
is going to be under Apache 2 .0, which means

00:11:54.509 --> 00:11:56.629
that developers can download, they can fine tune,

00:11:56.690 --> 00:11:59.169
they can redistribute the model free without

00:11:59.169 --> 00:12:01.350
permission or payment. I think this is awesome

00:12:01.350 --> 00:12:03.970
because I have AIbox .ai where we have like 80

00:12:03.970 --> 00:12:06.340
different AI models. And we've had... a bunch

00:12:06.340 --> 00:12:09.240
of llama models on there but um they kind of

00:12:09.240 --> 00:12:11.279
had a moment where they didn't release any open

00:12:11.279 --> 00:12:13.220
weight models like new ones for a while they're

00:12:13.220 --> 00:12:15.559
making it more um just kind of closed source

00:12:15.559 --> 00:12:18.419
and so those they kind of like fall behind and

00:12:18.419 --> 00:12:20.080
people stop using them so i'm always excited

00:12:20.080 --> 00:12:21.879
when we get these new open weight models we have

00:12:21.879 --> 00:12:23.360
a bunch of different platform partners that we

00:12:23.360 --> 00:12:25.960
work with that host them and then we then let

00:12:25.960 --> 00:12:29.259
us redistribute them on our platform so i'm excited

00:12:29.259 --> 00:12:30.480
for people to try them out they're typically

00:12:30.480 --> 00:12:32.960
a lot cheaper and they can get uh you know when

00:12:32.960 --> 00:12:35.049
when it first comes out like when this brand

00:12:35.049 --> 00:12:36.830
new glimmer comes out it's probably going to

00:12:36.830 --> 00:12:39.429
be pretty good um according to the benchmarks

00:12:39.429 --> 00:12:42.230
it's like kind of like number number like four

00:12:42.230 --> 00:12:45.620
number five top model but to not be, you know,

00:12:45.620 --> 00:12:47.480
under Meta's control, not going to their servers

00:12:47.480 --> 00:12:49.480
to be open weight model. I'm really excited.

00:12:49.659 --> 00:12:52.500
It does text and images and it has about 100

00:12:52.500 --> 00:12:55.559
languages that it supports. Also can work offline.

00:12:55.700 --> 00:12:57.419
So if you don't have internet, it works. It also

00:12:57.419 --> 00:12:59.960
can use the internet. It's designed to manage

00:12:59.960 --> 00:13:02.240
relationships, finances, help home logistics

00:13:02.240 --> 00:13:04.740
and does all on your device. I'm really excited

00:13:04.740 --> 00:13:06.320
by that, especially being able to use it without

00:13:06.320 --> 00:13:08.440
the internet. My wife was just talking to me

00:13:08.440 --> 00:13:10.179
about how annoying it is that every time she

00:13:10.179 --> 00:13:12.700
needs a recipe, she asks Google and Google uses

00:13:12.700 --> 00:13:15.309
AI to spit. out a recipe in the AI snippet at

00:13:15.309 --> 00:13:16.669
the top and she's like I feel like it's such

00:13:16.669 --> 00:13:20.169
a waste that I'm using like an LLM to regenerate

00:13:20.169 --> 00:13:23.049
the same cookie recipe that I make every two

00:13:23.049 --> 00:13:24.850
weeks and all these same recipes I make all the

00:13:24.850 --> 00:13:26.909
time she's like I just want like note cards where

00:13:26.909 --> 00:13:29.169
I write them down and I was like honestly the

00:13:29.169 --> 00:13:31.149
modern version of this is probably just like

00:13:31.149 --> 00:13:33.769
we'll have local LLMs on our phones or on our

00:13:33.769 --> 00:13:35.830
computers that yeah like it regenerates stuff

00:13:35.830 --> 00:13:38.309
but it's just using like your your phone or your

00:13:38.309 --> 00:13:40.370
computer's resources it doesn't need the internet

00:13:40.370 --> 00:13:42.129
and you don't feel like you're you know using

00:13:42.129 --> 00:13:44.360
a gallon of water just to get your one response.

00:13:44.440 --> 00:13:46.399
So I think Meta is kind of pushing in this direction.

00:13:46.720 --> 00:13:49.419
They're keeping MuseSpark, which is their most

00:13:49.419 --> 00:13:52.480
powerful model, closed source. So not everything

00:13:52.480 --> 00:13:54.220
that they've created is going to be open source.

00:13:54.820 --> 00:13:57.279
And that is going to be, that was released in

00:13:57.279 --> 00:13:59.399
April. We've seen this with some other players

00:13:59.399 --> 00:14:01.720
where you release your best model as, you know,

00:14:01.720 --> 00:14:04.080
paid and closed source, but... you have your

00:14:04.080 --> 00:14:06.139
second best model, which is kind of like open

00:14:06.139 --> 00:14:07.759
source or open weight. And I'm thrilled with

00:14:07.759 --> 00:14:09.940
that arrangement. I think that's very generous

00:14:09.940 --> 00:14:12.039
and I'm excited to play around with it. Thanks

00:14:12.039 --> 00:14:13.639
so much for tuning into the podcast, guys. Like

00:14:13.639 --> 00:14:15.679
I mentioned, if you want to grow and scale a

00:14:15.679 --> 00:14:18.480
business or start a side hustle that uses AI

00:14:18.480 --> 00:14:20.659
to make money, I'd love for you to go check out

00:14:20.659 --> 00:14:23.000
my school community, the AI Hustle School Community.

00:14:23.139 --> 00:14:24.980
I'll leave a link in the description for that.

00:14:25.039 --> 00:14:27.320
And I hope you all have a fantastic rest of your

00:14:27.320 --> 00:14:27.480
day.