WEBVTT

00:00:00.000 --> 00:00:03.419
Okay, imagine this for a second. 96%. That's

00:00:03.419 --> 00:00:07.219
the figure for how much AI information, you know,

00:00:07.219 --> 00:00:09.359
when it cites its sources, comes straight from

00:00:09.359 --> 00:00:12.619
corporate blogs, PR, or journalism. Wow. Not

00:00:12.619 --> 00:00:15.539
ads, not random corners of the web, but really

00:00:15.539 --> 00:00:18.800
specific sort of curated content. Kind of makes

00:00:18.800 --> 00:00:21.140
you rethink where KI gets its smarts, doesn't

00:00:21.140 --> 00:00:23.120
it? And maybe even how you'd create your own

00:00:23.120 --> 00:00:25.859
content. It really does put those AI report cards

00:00:25.859 --> 00:00:27.760
in a different perspective. It's like finding

00:00:27.760 --> 00:00:31.219
out your super smart friend basically just reads

00:00:31.219 --> 00:00:34.159
very specific magazines and company websites.

00:00:34.280 --> 00:00:36.439
He wouldn't guess AI had such particular tastes.

00:00:37.560 --> 00:00:39.979
Welcome, listener, to this deep dive. We're digging

00:00:39.979 --> 00:00:42.380
into the latest in AI, pulling directly from

00:00:42.380 --> 00:00:43.960
a recent newsletter that was just packed with

00:00:43.960 --> 00:00:46.539
data. Our mission today is pretty simple. Unpack

00:00:46.539 --> 00:00:48.840
it all, find those surprising bits, and basically

00:00:48.840 --> 00:00:51.159
give you a shortcut to getting up to speed. That's

00:00:51.159 --> 00:00:53.350
the plan. We're going to pull back the curtain

00:00:53.350 --> 00:00:56.189
on which AI models are actually, you know, winning

00:00:56.189 --> 00:00:58.570
right now. We'll look at practical new features

00:00:58.570 --> 00:01:01.189
you can use. And yeah, we'll definitely get into

00:01:01.189 --> 00:01:03.210
that fascinating study on where these digital

00:01:03.210 --> 00:01:06.269
brains really get their info from. Prepare for

00:01:06.269 --> 00:01:10.989
a few aha moments, I think. So today, first up,

00:01:11.030 --> 00:01:13.969
we'll look at the actual performance data. Which

00:01:13.969 --> 00:01:16.590
models are winning specific races like coding

00:01:16.590 --> 00:01:21.159
or math or image generation? It's often not quite

00:01:21.159 --> 00:01:23.340
what the headlines shout about. Then we'll hit

00:01:23.340 --> 00:01:25.560
some highlights from today in AI, quick hits

00:01:25.560 --> 00:01:27.719
on new features, some policy stuff that's making

00:01:27.719 --> 00:01:29.799
waves, and even where the big money is moving

00:01:29.799 --> 00:01:32.340
in the AI world. And finally, we'll really explore

00:01:32.340 --> 00:01:35.500
that new study on AI sourcing. Seriously, it

00:01:35.500 --> 00:01:37.000
might just shift how you think about putting

00:01:37.000 --> 00:01:39.500
content out there online. Okay, let's dive into

00:01:39.500 --> 00:01:41.900
this first part. For a while, the AI scene felt

00:01:41.900 --> 00:01:44.219
a bit like, well, a popularity contest maybe.

00:01:44.459 --> 00:01:47.159
But now we're getting solid data, actual benchmarks

00:01:47.159 --> 00:01:49.930
showing who's best at what. Exactly. And it's

00:01:49.930 --> 00:01:52.650
not just changing fast. It's getting super specialized.

00:01:52.750 --> 00:01:56.010
You've got Croc 4 jumping onto the LM Arena leaderboard,

00:01:56.109 --> 00:01:58.250
making noise, especially with math and reasoning.

00:01:58.590 --> 00:02:02.189
But then, almost under the radar, Gemini 2 .5

00:02:02.189 --> 00:02:04.590
Pro is just sweeping up wins in category after

00:02:04.590 --> 00:02:06.709
category, kind of like the quiet horse that keeps

00:02:06.709 --> 00:02:08.909
winning. And if we look at specific tasks, yeah,

00:02:08.969 --> 00:02:10.969
the winners get really clear. For general tech

00:02:10.969 --> 00:02:13.870
stuff, Gemini 2 .5 Pro is basically top dog right

00:02:13.870 --> 00:02:17.050
now on that leaderboard. Yeah. ChatGPT 4 .0 and

00:02:17.050 --> 00:02:19.310
its earlier version, 0 .3. they're tied for a

00:02:19.310 --> 00:02:21.990
solid second place. Grok 4, which has had a lot

00:02:21.990 --> 00:02:25.030
of buzz, is actually tied for third with GPT

00:02:25.030 --> 00:02:28.770
4 .5. So it's good, definitely, but not totally

00:02:28.770 --> 00:02:31.759
dominating text tasks. And what's fascinating,

00:02:32.020 --> 00:02:34.460
like you said, is that specialization. Take coding

00:02:34.460 --> 00:02:36.240
there. It's a three -way tie right at the top.

00:02:36.300 --> 00:02:39.000
You've got Gemini 2 .5 Pro, DeepSeek R1, and

00:02:39.000 --> 00:02:42.000
Clawed Opus 4, all tied, and Grok 4. For coding,

00:02:42.099 --> 00:02:44.759
it's way down in 14th place. That's a huge difference

00:02:44.759 --> 00:02:46.419
from its overall intelligence score. It really

00:02:46.419 --> 00:02:48.680
hammers home the point. Just being smart overall

00:02:48.680 --> 00:02:50.900
doesn't mean an AI is great at everything. That's

00:02:50.900 --> 00:02:54.139
a really key point. But then Grok 4 flips the

00:02:54.139 --> 00:02:56.939
script for math. It's the undisputed champ there.

00:02:57.319 --> 00:02:59.560
Gemini and Clawed are right behind it, but Grok

00:02:59.560 --> 00:03:02.520
takes the crown. So it's clearly tuned for numbers.

00:03:03.039 --> 00:03:06.800
For complex reasoning, Gemini 2 .5 Pro and ChatGPT

00:03:06.800 --> 00:03:09.319
4 are leading, and Grok 4 is still strong in

00:03:09.319 --> 00:03:11.199
third place there. And when you look at, you

00:03:11.199 --> 00:03:13.860
know, understanding images, Gemini 2 .5 Pro leads

00:03:13.860 --> 00:03:17.120
again. Then you've got GPT 4 .5, O3, and 4 .0

00:03:17.120 --> 00:03:19.319
all tied for second. Okay. And for search tasks,

00:03:19.680 --> 00:03:23.000
Gemini 2 .5 Pro actually ties with Perplexity

00:03:23.000 --> 00:03:25.280
Sonar Reasoning Pro. It really shows Google's

00:03:25.280 --> 00:03:27.759
strength in those visual and search -heavy areas,

00:03:27.900 --> 00:03:29.419
which, I mean, makes sense, right? That's their

00:03:29.419 --> 00:03:31.759
backyard. Yeah, absolutely. Now, here's where

00:03:31.759 --> 00:03:34.240
it gets really interesting for me. Image generation

00:03:34.240 --> 00:03:38.129
and editing. OpenAI has this new model, GPT -Image

00:03:38.129 --> 00:03:40.909
1, and it's just completely dominating. It's

00:03:40.909 --> 00:03:42.849
carved out its own space there doing some incredible

00:03:42.849 --> 00:03:45.069
things. Oh, I mean, just imagine trying to scale

00:03:45.069 --> 00:03:47.409
up running a billion queries. You'd want these

00:03:47.409 --> 00:03:50.129
specialized models, wouldn't you? Each one just

00:03:50.129 --> 00:03:52.889
nailing its specific job. It means picking the

00:03:52.889 --> 00:03:54.770
right tool isn't just about hype anymore. It's

00:03:54.770 --> 00:03:57.819
a real strategic choice. Yeah. This really shows

00:03:57.819 --> 00:04:00.800
AI intelligence isn't one score. It's more like

00:04:00.800 --> 00:04:03.020
a mosaic, you know, lots of different strength.

00:04:03.159 --> 00:04:04.699
And we have to talk about the price tag, right?

00:04:04.759 --> 00:04:06.960
That's the reality check. Grok 4 might score

00:04:06.960 --> 00:04:10.500
high on intelligence, but it's also $6 per million

00:04:10.500 --> 00:04:13.099
tokens. And remember, tokens are kind of like

00:04:13.099 --> 00:04:16.240
the words or word parts the AI processes. That's

00:04:16.240 --> 00:04:19.319
a premium. Right. If you need the absolute smartest

00:04:19.319 --> 00:04:21.879
and budget isn't the main issue, Grok 4 is up

00:04:21.879 --> 00:04:24.779
there. But let's say you want like... 95 % of

00:04:24.779 --> 00:04:27.420
that performance for maybe half the cost. Gemini

00:04:27.420 --> 00:04:30.720
2 .5 Pro to 4 Mini look really solid. It boils

00:04:30.720 --> 00:04:33.019
down to cost efficiency. And if you want the

00:04:33.019 --> 00:04:36.199
smartest AI for the lowest cost, DeepSeek R1

00:04:36.199 --> 00:04:38.399
actually takes that spot. So for you listening,

00:04:38.540 --> 00:04:41.199
the takeaway here is pretty clear. Choosing an

00:04:41.199 --> 00:04:44.639
AI model now is less about the overall buzz and

00:04:44.639 --> 00:04:47.040
way more about matching it to your specific task

00:04:47.040 --> 00:04:49.680
and, well, your wallet. OK, let's shift gears

00:04:49.680 --> 00:04:51.439
a bit. Let's get into some of the faster moving

00:04:51.439 --> 00:04:54.259
developments, new features, policy stuff, business

00:04:54.259 --> 00:04:56.740
news. This is our Today in AI segment. First

00:04:56.740 --> 00:04:59.439
off, have you seen this Warren Buffett mega prompt

00:04:59.439 --> 00:05:01.680
floating around? I think I saw something about

00:05:01.680 --> 00:05:03.800
that. Yeah. Yeah. It claims it can turn chat

00:05:03.800 --> 00:05:07.259
GPT into like an expert market analyst. Sounds

00:05:07.259 --> 00:05:10.980
amazing, right? But got to remember, even with

00:05:10.980 --> 00:05:14.509
the world's best prompt. GPT doesn't have a CFA.

00:05:14.730 --> 00:05:17.110
It's not a chartered financial analyst. Right.

00:05:17.209 --> 00:05:19.769
It's a tool maybe for analysis, but it's definitely

00:05:19.769 --> 00:05:22.129
not a replacement for actual human financial

00:05:22.129 --> 00:05:24.310
expertise. You still need your own judgment.

00:05:24.629 --> 00:05:27.089
Absolutely. That's such a critical point, especially

00:05:27.089 --> 00:05:30.350
with money advice. AI can help, but it doesn't

00:05:30.350 --> 00:05:33.189
replace expertise or critical thinking. Now,

00:05:33.209 --> 00:05:36.410
something maybe more practical day to day. Chat

00:05:36.410 --> 00:05:39.699
GPTs. Ghost mode. Oh yeah, the privacy feature.

00:05:39.839 --> 00:05:42.300
Exactly. If you just want a quick chat, don't

00:05:42.300 --> 00:05:44.060
want your info saved or used for training, this

00:05:44.060 --> 00:05:46.500
is great. It's free, it's private, and it forgets

00:05:46.500 --> 00:05:48.819
everything the second you're done. Like a digital

00:05:48.819 --> 00:05:51.279
etcha sketch for conversations. And speaking

00:05:51.279 --> 00:05:53.779
of making it more personal, ChatGPT also got

00:05:53.779 --> 00:05:56.660
moods. You can actually tell it how to act, how

00:05:56.660 --> 00:05:59.160
to talk, think, its whole vibe. Imagine setting

00:05:59.160 --> 00:06:01.899
it to cynic if you need someone to poke holes

00:06:01.899 --> 00:06:04.889
in your ideas late at night. Or maybe Sage if

00:06:04.889 --> 00:06:07.089
you're looking for something more profound. It

00:06:07.089 --> 00:06:09.850
makes the AI feel, I don't know, more adaptable.

00:06:09.970 --> 00:06:12.230
And speaking of adaptable or maybe just fast,

00:06:12.569 --> 00:06:17.029
Google dropped Gemini 2 .5 flashlight. They're

00:06:17.029 --> 00:06:20.029
calling it their fastest 2 .5 model yet. Yeah.

00:06:20.110 --> 00:06:23.819
And the cost just... Ten cents per million input

00:06:23.819 --> 00:06:26.339
tokens. Wow, that's cheap. Yeah, really cheap.

00:06:26.439 --> 00:06:29.139
So it's aimed at high volume stuff where you

00:06:29.139 --> 00:06:31.899
need speed and low latency above all else. OK,

00:06:32.000 --> 00:06:33.819
now shifting to policy, there was a pretty big

00:06:33.819 --> 00:06:35.899
move in the U .S. recently. Our source material

00:06:35.899 --> 00:06:39.160
says that Trump signed an order aiming to ban

00:06:39.160 --> 00:06:44.160
woke AI models from federal contracts. The order

00:06:44.160 --> 00:06:46.980
says models need to be truth seeking and ideologically

00:06:46.980 --> 00:06:49.540
neutral. And the source specifically notes that

00:06:49.540 --> 00:06:52.720
XAI might benefit most from this. Just relaying

00:06:52.720 --> 00:06:54.540
what the source reported here, obviously not

00:06:54.540 --> 00:06:56.620
taking a stand. Understood. And on the finance

00:06:56.620 --> 00:06:58.939
side of AI, a big funding round for Gupshup.

00:06:59.000 --> 00:07:01.120
They just raised $60 million, mostly equity.

00:07:01.550 --> 00:07:03.209
They're looking to grow in India, Latin America,

00:07:03.350 --> 00:07:05.189
places like that. They handle a ton of messages,

00:07:05.250 --> 00:07:07.689
right? Yeah, over 120 billion messages a year

00:07:07.689 --> 00:07:09.629
for like 50 ,000 businesses. And there's even

00:07:09.629 --> 00:07:11.829
chatter about a potential IPO in the next year

00:07:11.829 --> 00:07:14.410
and a half, two years. So big money is definitely

00:07:14.410 --> 00:07:16.850
flowing. Ultimately, what stands out here is

00:07:16.850 --> 00:07:19.990
AI getting more personal, more adaptable, but

00:07:19.990 --> 00:07:22.589
also really specialized for certain needs and

00:07:22.589 --> 00:07:24.589
business cases. Right. Let's keep the pace up.

00:07:24.689 --> 00:07:26.389
Let's do a quick run through of some interesting

00:07:26.389 --> 00:07:29.230
new AI tools and other industry buzz that's changing

00:07:29.230 --> 00:07:32.410
the game. Okay. Yeah. seeing this explosion of

00:07:32.410 --> 00:07:34.889
new AI tools, right? Designed for very specific

00:07:34.889 --> 00:07:39.389
jobs, making things easier like Bonsai. It claims

00:07:39.389 --> 00:07:42.069
it can turn any document into an engaging video

00:07:42.069 --> 00:07:46.110
course in just 15 minutes. 15 minutes, seriously?

00:07:46.129 --> 00:07:48.550
That's the claim. Think about that for training

00:07:48.550 --> 00:07:51.910
or education or just explaining something complicated

00:07:51.910 --> 00:07:54.620
fast. Yeah, that's potentially huge. And then

00:07:54.620 --> 00:07:56.860
there's Product Fit. This one creates web content,

00:07:57.100 --> 00:07:59.339
ad stuff, and even gives you a dashboard showing

00:07:59.339 --> 00:08:02.000
user interest in real time. It's all about figuring

00:08:02.000 --> 00:08:04.120
out your audience and reaching them better, taking

00:08:04.120 --> 00:08:06.279
out some of the guesswork. And for designers,

00:08:06.399 --> 00:08:08.980
there's AI Image Maker. It's a free online tool

00:08:08.980 --> 00:08:11.339
for design and image processing. Sounds pretty

00:08:11.339 --> 00:08:14.160
powerful. And another one, Turbo Style, lets

00:08:14.160 --> 00:08:16.420
you tweak styles, swap visuals, try out design

00:08:16.420 --> 00:08:19.019
ideas for websites. These tools really feel like

00:08:19.019 --> 00:08:21.079
they're spreading creative power around. Making

00:08:21.079 --> 00:08:24.779
sophisticated stuff more accessible, yeah. Okay,

00:08:24.819 --> 00:08:27.819
some quick industry hits. VO3. It's apparently

00:08:27.819 --> 00:08:31.079
making pro -level commercial concepts now, and

00:08:31.079 --> 00:08:33.659
the descriptions say they're shockingly realistic.

00:08:34.139 --> 00:08:36.940
Wow. The quality leap in video generation is

00:08:36.940 --> 00:08:39.340
getting kind of scary, making it harder to tell

00:08:39.340 --> 00:08:42.720
AI from reality. No kidding. And Anthropic, you

00:08:42.720 --> 00:08:45.779
know, the company behind Claude, they've quietly

00:08:45.779 --> 00:08:49.759
hit. Over $4 billion in annual recurring revenue.

00:08:49.860 --> 00:08:53.759
$4 billion. Quietly. Yeah. Apparently leading

00:08:53.759 --> 00:08:57.299
the B2B AI space, that's a massive number, shows

00:08:57.299 --> 00:08:59.659
how much businesses want these AI solutions.

00:08:59.980 --> 00:09:02.320
Definitely. Also, this is interesting. Google's

00:09:02.320 --> 00:09:05.620
AI is apparently creating zero -click webs. Basically,

00:09:05.620 --> 00:09:07.580
you get the answer right in the search results.

00:09:07.740 --> 00:09:09.519
No need to click a link. Right. I've seen that

00:09:09.519 --> 00:09:11.679
more and more. The source we read suggested this

00:09:11.679 --> 00:09:14.039
could mean maybe 10 % of users just stopped browsing

00:09:14.039 --> 00:09:15.960
further because they got their answer right there.

00:09:16.100 --> 00:09:18.039
That's a huge shift for content creators who

00:09:18.039 --> 00:09:19.919
rely on those clicks. It's like Google's becoming

00:09:19.919 --> 00:09:22.559
the destination, not just the map. That has enormous

00:09:22.559 --> 00:09:25.519
implications. And speaking of Google, Google

00:09:25.519 --> 00:09:28.179
Photos is adding AI features too, letting you

00:09:28.179 --> 00:09:31.179
remix photos, turn pics into videos. Meanwhile,

00:09:31.259 --> 00:09:34.279
the talent shuffle continues. Microsoft apparently

00:09:34.279 --> 00:09:37.120
hired around two dozen people from Google's DeepMind

00:09:37.120 --> 00:09:39.830
AI lab. A talent war is definitely still hot.

00:09:40.009 --> 00:09:42.470
Yeah. You know, I still wrestle with prompt drift

00:09:42.470 --> 00:09:45.429
myself sometimes, where the AI just kind of forgets

00:09:45.429 --> 00:09:47.289
what you asked it initially after a long chat.

00:09:47.409 --> 00:09:49.649
Trying to keep up with all these new tools and

00:09:49.649 --> 00:09:51.870
how they keep changing. It's a lot. It'll tell

00:09:51.870 --> 00:09:53.830
me about it. But the takeaway from all this rapid

00:09:53.830 --> 00:09:56.129
change is clear, isn't it? The AI landscape is

00:09:56.129 --> 00:09:58.330
just incredibly dynamic. There are new tools

00:09:58.330 --> 00:10:00.529
popping up for pretty much every niche you can

00:10:00.529 --> 00:10:04.460
think of. Sponsor. Okay, let's get to our final

00:10:04.460 --> 00:10:06.620
segment. And this one I think is really fascinating,

00:10:06.799 --> 00:10:09.259
especially if you create content online or even

00:10:09.259 --> 00:10:11.440
if you just care about where AI gets its information

00:10:11.440 --> 00:10:13.340
from. We're going to dive deep into how these

00:10:13.340 --> 00:10:15.440
models actually source what they tell you. Yeah,

00:10:15.539 --> 00:10:17.299
this is so crucial. It helps us peek under the

00:10:17.299 --> 00:10:20.440
hood, understand the knowledge base these powerful

00:10:20.440 --> 00:10:23.139
tools are built on. Muckrak published this huge

00:10:23.139 --> 00:10:26.129
study analyzed over a million citations. From

00:10:26.129 --> 00:10:29.730
ChatGPT, Claude, Gemini. This isn't guesswork,

00:10:29.750 --> 00:10:32.710
it's actual data on what AI reads. And the big

00:10:32.710 --> 00:10:35.269
headline finding from the study is, well, it's

00:10:35.269 --> 00:10:38.710
pretty striking. 96 % of the citations in AI

00:10:38.710 --> 00:10:41.409
Answers come from content that started as corporate

00:10:41.409 --> 00:10:45.149
communications, PR, or journalism. 96 %? Almost

00:10:45.149 --> 00:10:47.049
everything cited comes from what we used to call

00:10:47.049 --> 00:10:49.710
earned media, or maybe company -owned blogs and

00:10:49.710 --> 00:10:52.649
stuff. Not from ads. Wow. Just think about that.

00:10:52.789 --> 00:10:55.169
If you produce any kind of information online,

00:10:55.269 --> 00:10:57.990
that directly impacts your strategy. The study

00:10:57.990 --> 00:11:00.190
even broke down citations by the type of task

00:11:00.190 --> 00:11:02.710
the user was doing, which gives even more insight.

00:11:02.830 --> 00:11:05.529
Like, if people were asking for advice, 48 %

00:11:05.529 --> 00:11:07.470
of citations were from corporate blogs. Almost

00:11:07.470 --> 00:11:09.710
half. And if you ask for step -by -step instructions,

00:11:10.009 --> 00:11:13.230
47 % blogs, 23 % Wikipedia. If you need help

00:11:13.230 --> 00:11:16.690
with a creative task, 43 % blogs, 26 % journalism.

00:11:16.889 --> 00:11:19.529
It really underscores the power of good old -fashioned

00:11:19.529 --> 00:11:22.179
blog content. as an AI knowledge source? Yeah,

00:11:22.299 --> 00:11:24.779
it's surprisingly dominant. When it came to comparing

00:11:24.779 --> 00:11:28.240
things, 35 % of citations were from aggregators

00:11:28.240 --> 00:11:32.500
like Wikipedia, 33 % from blogs. Factual lookups,

00:11:32.600 --> 00:11:35.759
similar split, 33 % aggregators, 33 % blogs.

00:11:36.080 --> 00:11:38.379
But when it came to current events, journalism

00:11:38.379 --> 00:11:42.419
was king. 49 % of citations from big news sources,

00:11:42.639 --> 00:11:45.519
Reuters, AP, Financial Times. So the quick summary,

00:11:45.639 --> 00:11:48.889
the TLDR. It's pretty clear, right? Blogs rule

00:11:48.889 --> 00:11:52.330
for casual questions and how -tos. News sources

00:11:52.330 --> 00:11:54.710
dominate for what's happening right now. It paints

00:11:54.710 --> 00:11:57.169
a really specific picture of how AI consumes

00:11:57.169 --> 00:11:59.230
and cites information. And what's also really

00:11:59.230 --> 00:12:00.830
interesting is that the models have different

00:12:00.830 --> 00:12:03.850
habits. ChatGPT, the study found, cited journalism

00:12:03.850 --> 00:12:07.070
the most, lots of mainstream news. Claude, on

00:12:07.070 --> 00:12:09.129
the other hand, tended to lean more on academic

00:12:09.129 --> 00:12:11.730
papers, government sources, technical documents.

00:12:12.129 --> 00:12:14.070
Huh. So they have different appetites. Yeah,

00:12:14.110 --> 00:12:15.769
different preferred feeding grounds, you could

00:12:15.769 --> 00:12:18.779
say. for content creators here feels pretty significant.

00:12:18.879 --> 00:12:21.460
The old SEO tricks? Probably not enough anymore.

00:12:21.740 --> 00:12:23.480
The study even mentioned that putting podcasts

00:12:23.480 --> 00:12:25.519
on YouTube might give you better odds of being

00:12:25.519 --> 00:12:28.159
cited by AI. Interesting. So format matters too.

00:12:28.419 --> 00:12:31.000
Yeah, it's not just keywords. It's about how

00:12:31.000 --> 00:12:34.240
accessible and structured and authoritative your

00:12:34.240 --> 00:12:37.019
content seems to these models. The biggest lesson

00:12:37.019 --> 00:12:39.139
here, I think, for anyone creating content today

00:12:39.139 --> 00:12:42.039
is that the format and the original source really

00:12:42.039 --> 00:12:45.100
matter for whether AI will even find it, let

00:12:45.100 --> 00:12:48.090
alone cite it. Okay, so let's try and tie this

00:12:48.090 --> 00:12:51.169
all together. What does it all mean? We've seen

00:12:51.169 --> 00:12:54.190
today that AI intelligence isn't really one single

00:12:54.190 --> 00:12:57.190
score, is it? It's much more like a mosaic of

00:12:57.190 --> 00:13:00.029
different strengths, specialized skills, and

00:13:00.029 --> 00:13:02.309
each often has its own price. You really have

00:13:02.309 --> 00:13:04.909
to think about your specific needs and budget.

00:13:05.210 --> 00:13:06.690
And we're getting more ways to interact with

00:13:06.690 --> 00:13:09.870
AI, too. More nuance, from giving it a mood to

00:13:09.870 --> 00:13:12.350
trying out these incredibly powerful new tools

00:13:12.350 --> 00:13:15.269
for pro -level video or images. The speed of

00:13:15.269 --> 00:13:18.110
it all is just blistering. Yeah. And maybe the

00:13:18.110 --> 00:13:19.809
most fundamental thing we talked about is getting

00:13:19.809 --> 00:13:22.269
a clearer picture of where AI's knowledge actually

00:13:22.269 --> 00:13:24.289
comes from. It's not some kind of magic, right?

00:13:24.330 --> 00:13:26.309
It's built on this huge ocean of human information,

00:13:26.409 --> 00:13:29.009
but predominantly, it seems, from corporate blogs

00:13:29.009 --> 00:13:31.450
and established journalism. That's a really big

00:13:31.450 --> 00:13:33.409
realization about our digital knowledge system.

00:13:33.720 --> 00:13:35.879
This whole deep dive really shows that, yeah,

00:13:35.960 --> 00:13:39.509
the tech is powerful, but Understanding the details,

00:13:39.730 --> 00:13:41.730
the strengths, the weaknesses, even the learning

00:13:41.730 --> 00:13:44.909
habits, that's the key to actually using it well.

00:13:45.049 --> 00:13:46.830
It definitely raises a big question, doesn't

00:13:46.830 --> 00:13:49.330
it? Yeah. If AI is learning so much from current

00:13:49.330 --> 00:13:52.029
content, especially blogs and news, what does

00:13:52.029 --> 00:13:54.289
that mean for how information lasts over time

00:13:54.289 --> 00:13:57.029
and how trustworthy it is in this digital age

00:13:57.029 --> 00:13:58.909
we're building? Yeah, something to think about

00:13:58.909 --> 00:14:01.230
next time you get an AI answer. Where did that

00:14:01.230 --> 00:14:03.750
really come from and what does that imply about

00:14:03.750 --> 00:14:06.750
its context, maybe even its accuracy? We really

00:14:06.750 --> 00:14:09.240
hope this deep... dives give you some fresh perspective,

00:14:09.500 --> 00:14:12.299
maybe some useful insights you can act on. You've

00:14:12.299 --> 00:14:14.379
been listening to The Deep Dive. Thanks so much

00:14:14.379 --> 00:14:16.139
for joining us. We'll be back soon with another

00:14:16.139 --> 00:14:17.440
one. Out to your music.