WEBVTT

1
00:00:00.000 --> 00:00:02.900
Hey, I'm Skyler.

2
00:00:02.900 --> 00:00:08.330
I've been building AI systems for over 10 years, and there's one thing I've

3
00:00:08.330 --> 00:00:08.940
consistently

4
00:00:08.940 --> 00:00:17.100
noticed separates the merely good AI engineers from the great ones.

5
00:00:17.100 --> 00:00:23.860
The products that take off, and the products that don't meet expectations.

6
00:00:23.860 --> 00:00:29.180
It's not about the shiniest new model architectures, it's not about having a

7
00:00:29.180 --> 00:00:31.600
complex system.

8
00:00:31.600 --> 00:00:39.390
It's about having comprehensive and robust evaluations that drive your

9
00:00:39.390 --> 00:00:41.260
development.

10
00:00:41.260 --> 00:00:50.600
The best AI engineers invariably know where their model excels, where it falls

11
00:00:50.600 --> 00:00:51.760
short.

12
00:00:51.760 --> 00:01:01.260
They continuously experiment based on evidence rather than guesswork.

13
00:01:01.260 --> 00:01:07.610
With foundation models and large language models, AI is more accessible than

14
00:01:07.610 --> 00:01:08.960
ever, and we have

15
00:01:08.960 --> 00:01:19.120
a whole suite of amazing engineers coming and building AI for the first time.

16
00:01:19.120 --> 00:01:21.420
But they haven't built EVALS before.

17
00:01:21.420 --> 00:01:28.520
They don't know when their system doesn't work.

18
00:01:28.520 --> 00:01:31.040
And that's dangerous.

19
00:01:31.040 --> 00:01:36.060
Your system kind of works, but you don't know when it's going to make a mistake

20
00:01:36.060 --> 00:01:36.760
that's going

21
00:01:36.760 --> 00:01:43.840
to cost you users or credibility or worse.

22
00:01:43.840 --> 00:01:50.910
If you're serious about building AI products, Hamel and Shreya's AI EVALS

23
00:01:50.910 --> 00:01:52.580
course will equip

24
00:01:52.580 --> 00:01:57.940
you with the practical knowledge to deliver AI products successfully.

25
00:01:57.940 --> 00:02:07.940
[BLANK_AUDIO]