"Age of Miracles"

Anton Teaches Packy AI | E1

17 snips
Nov 7, 2022
Ask episode
Chapters
Transcript
Episode notes
1
Introduction
00:00 • 3min
2
The Future of B2B SAS and AIML
03:01 • 3min
3
Is Attention All You Need?
05:32 • 3min
4
Is Parallelization a Good Idea?
08:24 • 2min
5
The Intersection of Crypto and AI?
09:58 • 2min
6
The Weirdest Furthest Out in a Little While
11:36 • 2min
7
Machine Learning Model Weights - What's Happening There?
13:30 • 2min
8
The Basic Algorithm of Machine Learning Models
15:07 • 2min
9
How to Do a Machine Translation Task?
16:49 • 3min
10
Are We Still in the World of Neural Networks?
19:41 • 2min
11
Neural Networks - The First AI Winter
22:09 • 3min
12
The Core Architecture in a Transformer
25:24 • 2min
13
The Position Encoding in a Recurrent Model
27:33 • 2min
14
Is Attention a Multi-Head Mechanism?
29:06 • 4min
15
Is There a Best Explanation for Why It Works?
32:49 • 2min
16
The Chinchilla Scaling Laws
34:45 • 3min
17
Is There a Need for New Architectures?
37:23 • 3min
18
What's Order Regressive?
40:17 • 2min
19
Is It Order Regressive?
42:01 • 3min
20
The Multi-Headed Attention Layer in the Encoder
45:22 • 3min
21
Is GPT 3 Parallelisable?
47:59 • 5min
22
Instruction GPT Is Better at Attending to What You're Asking It to Do
52:56 • 3min
23
How Do You Train a Model?
56:24 • 2min
24
Train a Model
58:10 • 4min
25
GPUs for Machine Learning
01:01:41 • 2min
26
Is the TPT So Expensive to Train You Cost $6 Million?
01:03:50 • 3min
27
Is There a Paper That You Should Read on Chinchillas
01:06:23 • 2min