Weekly Digest 17

Content Plan Updated

This weekly digest has now reached its 17th edition, which means 4 months of reading. I think it is really a lot, and a lot about AI. I have learned tons of ideas from various materials. And reading blogs has become my new hobby.

As I keep practicing my reading skill, I find I can filter irrelevant materials more rapidly, so I can read more and get what matters to me. And I will keep reading and sharing in the weekly digest as usual.

On the other hand, I find that I have read too much but do too little. Reading takes up a large amount of time in my life. So I decided to read less, but do more practice to get real experience on things. Then I will try to share this experience on this weekly digest as well. By the way, I think this can also train my writing skill, which is really the next bottleneck to overcome.

AI

Deep Dive into LLMs like ChatGPT

models need tokens to think

Basically, each token represents a fixed, finite amount of computation in the process of inference. So to get better answers, one way is to distribute computation to more tokens, which means thinking.

Note: ChatGPT web app can use code (Python) to do some calculations.