primordialsoup

This is very interesting work, but it's not really a LLM. It doesn't have language abilities. They should have called it a seq2seq model, but I think that term is not in vogue these days :)

It's like someone asking for a gasoline version of Tesla.

It's pretty much the opposite: Big corporates are the ones training the models (and benefiting from it).

I can do it myself and it's not a huge hill to climb, but I don't necessarily have the time to do it. There is also the time that needs to be dedicated towards maintenance (upgrades etc.).

I have heard this argument from perfectly healthy people, and I couldn't really believe it, since there are just so many other touch points when you go to a restaurant.

If someone is immuno compromised I think this makes sense to extent.

Tixerb. Here I reversed it.

Makes it worse, since you have this understanding and still went with this explanation that it's 1.7 trillion patients

And resources at some level.

Congrats! I went to your demo and asked for words that end in agi. This is what I got:

--

agi, agi, agi, agi, agi, agi, agi

These are some of the words that end in agi. You can also use the word agi in a sentence. For example, "I am going to the grocery store to get some agi."

These are some of words that end in agi.

These are some words that end in agi.

maximize, maximize, maximize, maximize, maximize, maximize, maximize, maximize

These are some words that ends in agi

--

So I think this needs more work to get to "as good as ChatGPT". But having said that, congrats on the landing

I like Substack. Obviously this situation is not good for Substack, but they should be proud that Twitter is scared of them.

Yeah it's a short article and not a long form analysis. So yes, by don't definition it's shallow.

It's expressing the author's discomfort and so it's pessimistic.

I think those are fine.

Who will prompt? Yes, the manager.

When everyone is sick, they stop calling it a disease. - naval.

:(

The first version of Dall-E was mediocre. Look where we are now.

I can think of a list of works from Alphabet but can't think of much from FAIR (other than OPT).

I will be happy to read something nice from FAIR as well, so please elaborate.

That's sad and hope something can be done about that. But that still doesn't mean we should go back and kill the iphone if we can do it. Because then we should go back to get rid of cars (so many accidents) and kill industrial revolution as well (so much pollution)...

HIIT vs. regular cardio is not micro optimization (in your words). Trying to figure out the exact heart rates might count as one.

An analogy to what you are saying would be this: Studying in MIT vs. any random university should be the same. "Go to college, study hard and you will be fine. It's not rocket really science."

Your other points are valid.

It's funny you say that GPUs are general purpose.

It's almost like a universal law that generic stuff is helpful only to some extent, after which you need specialization. That's why we do PhDs I suppose. And that's why you have cardiologists.

Somewhat orthogonal comment. Was reading this book "The coming storm" where the author Michael Lewis talks about chaos theory, and how you could go back to one (small) thing that is at the root of who you are. For example, DJ Patil had dyslexia, and that made him not do well in the school, and so he ended up going the off-beaten track and hacking his school systems etc. Years later, he was hacking into US Weather service's data, and many more incidents later, he became "Head of Data Products and Chief Scientist of LinkedIn" and later worked for Obama.

Parallel here is that one guy's mischievous behavior has such a massive consequence.

If I don't like you, I could give this to you without your knowledge and report you. The point being, it can fail and can be misused.

I have switched to Remnote to take better notes of books I read as well to use practice spaced repetition, but it is hard to make a routine of practicing the cards. :)

Does this apply to non-transformer based architectures as well?

Yes, by "making the switch" I was not referring to just suddenly become director material. It is basically how I can change how I learn. What you have mentioned is useful, thanks, and I am trying my best to do all of those, but how I improve is linear at best. I see others who grow better than linear, and would like to emulate them. Ultimately my goal is not to just strive towards a Director, but to become a solid performer who strategizes/works like one.