- Tweets 29
- Following 33
- Followers 4
- Likes 97
Inference-scaling lets us trade extra compute for better modeling accuracy. Next to reinforcement learning, it has become one of the most important concepts in today's LLMs, so the book will cover it in two chapters instead of just one.
I just finished the first one. It is a 35-page introduction to inference-time scaling through self-consistency sampling. This chapter was a lot of fun to write because it takes the base model on MATH-500 all the way from 15.2% percent to 52.2% accuracy.
Seeing that jump without additional training is incredibly satisfying.
Submitted the chapter yesterday, and it should appear in the Manning Early Access program in the next few days. (In the meantime the first 176 pages that lead up to this chapter are already available.)
The next chapter will focus on self-refinement techniques, where the model improves its own answers through iterative reasoning.
Happy reading!
23
162
1,006
183,518
话说这几天在瑞典的一线城市斯德哥尔摩和超一线城市哥德堡，频频被认出来，搞得我还怪不好意思的。
今天我们临时决定，下一站德国，再下一站法国，我们会在某一天派王局上街闲逛，第一位认出王局并上来打招呼的粉丝，王局将掏钱请对方吃点儿好的。对方可以提议吃任何当地的饭馆。
但是，接下来说但是，
如果这一天王局没有碰到认识他的观众，作为惩罚，他就不能吃饭。
有没有饭吃就看有没有观众了。
259
12
798
277,690
Artists standing up for themselves and making it “disruptive” for top executives is 👏👏
Disney CEO Bob Iger says "there’s a level of expectation" the Hollywood strikers have "that is just not realistic. And they are adding to the set of the challenges that this business is already facing that is, quite frankly, very disruptive." variety.com/2023/tv/news/bob…
9
3
645
Sorry wrong thread, see
This statement is flawed on so many levels. Going by this, he definitely met consumers who were like “hey, I wish my phone had blinky lights.”
Kudos to Nothing for designing a stand-out, good-looking slab phone but don’t go around parading like actual innovation isn’t useful just because you can’t do it yet. A big screen that folds into a smaller phone is much more appealing than lights.
11
This statement is flawed on so many levels. Going by this, he definitely met consumers who were like “hey, I wish my phone had blinky lights.”
Kudos to Nothing for designing a stand-out, good-looking slab phone but don’t go around parading like actual innovation isn’t useful just because you can’t do it yet. A big screen that folds into a smaller phone is much more appealing than lights.
Carl Pei on foldables, courtesy of @IndiaToday. Can't sell what you don't have, right?
This is also such a weird take. Because people might not wish for a foldable phone but they do wish for a large screen or a small phone, albeit without the drawbacks of each.
21
9
185
30,253
7. Yet no one makes phones with LED because they don’t know if it is what the customers want. Nothing is bold to launch such a simple yet differentiate design.
9
6. Equating the foldable design to nothing’s LED light is also hilarious. Manufacturers make foldable phones because they thought they can and others cannot. That’s why they are still very pricy. The LED lights are so simple that any phone case company can make them.
12
5. If Foldable phone is what customers all really want, it would’ve been what most people looking for when buying their phone. And this doesn’t happen. If this is the case, Apple will definitely roll out their own foldable phone.
8
4. Of course there are some customer love using foldable phone. Maybe you are even one of them. But talking about your subjective experience of foldable phone has nothing to contradict Carl’s point, but leave a bad impression of Carl Pei on people who know nothing about him.
1
15
3 Reality distortion field is all about Steve’s style of management that persuaded his employees to work harder for Apple’s objective, while Carl is simply providing his analysis on a category of product.
8
2 For example, in what sense does Carl Pei’s thought on foldable phone equate Steve job’s reality distortion field? This is almost name calling.
10