WEBVTT

00:01.410 --> 00:02.520
-: Hi, I'm just gonna walk you

00:02.520 --> 00:05.130
through ChatGPT Deep Research.

00:05.130 --> 00:07.650
So it's a button here that you press

00:07.650 --> 00:09.480
and you could do it with the different models.

00:09.480 --> 00:14.370
In particular, if you have the $200 a month Pro account,

00:14.370 --> 00:15.720
then it works really well

00:15.720 --> 00:18.240
because the Pro model is much better

00:18.240 --> 00:19.350
at thinking through things,

00:19.350 --> 00:22.290
and then it can come up with better things to research.

00:22.290 --> 00:25.530
Just going to ask a question here.

00:25.530 --> 00:29.910
I use this primarily for researching scientific papers

00:29.910 --> 00:31.920
or topics that I need to understand deeply.

00:31.920 --> 00:36.150
So I'm gonna say, why didn't memetics

00:36.150 --> 00:40.623
become a science versus cultural evolution,

00:42.060 --> 00:43.840
which seems to have

00:46.110 --> 00:47.340
become the dominant path

00:47.340 --> 00:51.633
for studying evolution of ideas.

00:52.470 --> 00:53.520
And when you submit,

00:53.520 --> 00:56.340
quite often, it'll ask you specific questions

00:56.340 --> 00:57.510
'cause it wants to get it right

00:57.510 --> 01:00.270
and it is quite a lengthy process, so it is important

01:00.270 --> 01:01.170
that you answer it.

01:01.170 --> 01:03.210
Normally, I would actually spend a lot more time

01:03.210 --> 01:06.450
on the prompt, pointing it in a few different directions,

01:06.450 --> 01:11.450
but specifically why from an incentives point of view,

01:11.640 --> 01:12.970
cultural evolution

01:15.750 --> 01:17.860
adopted over memetics

01:20.870 --> 01:22.053
by scientists.

01:29.310 --> 01:31.323
Okay. And now it's doing the research.

01:34.260 --> 01:36.750
It typically can take anywhere from a few minutes

01:36.750 --> 01:39.333
to 10-20 minutes in some cases.

01:45.330 --> 01:47.250
If you click on it, you can usually see

01:47.250 --> 01:48.510
while it thinks, it's red.

01:48.510 --> 01:50.910
On the right hand side here, you'll be able to see

01:50.910 --> 01:52.623
what sources it looks at too.

01:57.240 --> 01:59.910
Okay, you can see here that it's gone through one

01:59.910 --> 02:04.590
of the sources already, this MIT source.

02:04.590 --> 02:07.410
And it's telling you what it's looking at,

02:07.410 --> 02:10.200
and it's gonna keep doing that for a while, keep thinking

02:10.200 --> 02:12.300
and then keep going back to different sources.

02:12.300 --> 02:15.450
It's actually writing the Google search queries in order

02:15.450 --> 02:16.830
to find these sources as well,

02:16.830 --> 02:19.110
which is what makes it so powerful.

02:19.110 --> 02:21.750
And then it will decide when it has enough information

02:21.750 --> 02:23.520
to give you your answer.

02:23.520 --> 02:25.380
So it will think a lot longer for things

02:25.380 --> 02:27.330
that are a bit more difficult

02:27.330 --> 02:29.490
and it'll come back much more directly

02:29.490 --> 02:30.993
if it gets an answer quicker.

02:33.150 --> 02:35.670
Okay, it looks like we have everything back

02:35.670 --> 02:38.040
and it did an astounding amount of work.

02:38.040 --> 02:42.300
It looked at 28 different sources from Stanford

02:42.300 --> 02:47.130
to like Wikipedia and neuroanthropology.net.

02:47.130 --> 02:49.650
Tons of really interesting sources that you could always go

02:49.650 --> 02:51.570
and check for yourself,

02:51.570 --> 02:55.650
but it also gives you all the thinking that went into this

02:55.650 --> 02:59.430
and I can understand its thought process as it comes on.

02:59.430 --> 03:01.290
All right, so we have

03:01.290 --> 03:04.560
like now a pretty comprehensive report.

03:04.560 --> 03:05.970
One of the main things

03:05.970 --> 03:09.780
is that memetics was less testable.

03:09.780 --> 03:12.390
There's also more interdisciplinary integration

03:12.390 --> 03:14.250
in cultural evolution.

03:14.250 --> 03:16.890
And also, there were better institutional

03:16.890 --> 03:18.030
and funding incentives

03:18.030 --> 03:21.330
'cause it was easier to get at some of the big questions

03:21.330 --> 03:23.010
of cultural evolution

03:23.010 --> 03:25.830
with these existing funding mechanisms.

03:25.830 --> 03:27.510
Yeah, pretty interesting.

03:27.510 --> 03:29.790
You could go read all these references if you want.

03:29.790 --> 03:31.590
You could also ask it more information

03:31.590 --> 03:34.050
now it has these sources that it's found.

03:34.050 --> 03:36.930
It could also go and add more stuff to it as well.

03:36.930 --> 03:39.530
You could also ask it to summarize if you needed to.

03:40.380 --> 03:42.240
Typically, I'll take this and I'll copy it

03:42.240 --> 03:45.180
and I'll paste it into a Claude

03:45.180 --> 03:47.790
and then I will ask Claude

03:47.790 --> 03:50.610
to summarize it in a tweet or something like that.

03:50.610 --> 03:52.080
Claude is a little bit more creative

03:52.080 --> 03:54.930
and these reports end up being relatively dry,

03:54.930 --> 03:57.240
but they are amazingly well researched

03:57.240 --> 03:58.800
and that's the main reason

03:58.800 --> 04:01.503
to use something like deep research.