3 ms·

If you are someone who was working in LLMs/Generative AI before it got "cool", The market has fundamentally changed and changed extremely in your favor. Talent

If you are someone who was working in LLMs/Generative AI before it got "cool", The market has fundamentally changed and changed extremely in your favor. Talent is extremely scarce and expensive right now in this space. Ask for well above FAANG compensation. Try to shop around for offers. I had 4 offers and a bidding war for my talent (with no leetcode - but I have an extensive github and publications about LLMs), when previously these same companies would have brutally leetcoded me. I thought that there would be a ton of people getting really good really fast with Generative AI. It turns out that most people are terrible at using the tools, and there's even research about this right now - https://dl.acm.org/doi/10.1145/3544548.3581388 https://dl.acm.org/doi/10.1145/3544548.3581388

sanderjd 3y ago
I'm someone who would like to get "really good really fast", but have found that the on-ramps remain pretty weak. For instance, I have yet to find a good book on the subject. There are a ton of tutorials and articles, but it's maddening to try to cobble together any depth of understanding from those little nuggets. And there are tons of good papers on how the systems actually work, but these are not very useful for people who are new to this set of tools to figure out how to use them to create useful things. But I think this will hole will close up incredibly quickly over the next six months. (And it's yet another really good opportunity for people like you to be involved in that gold rush!) Edit to add: For instance, the article lightly lambasts peoples' recommendations to read up on AI / ML fundamentals, and contains this: > n the near future, nobody will recommend starting in AI Engineering by reading Attention is All You Need, just like you do not start driving by reading the schematics for the Ford Model T. But, amazingly, it doesn't actually suggest an alternative starting point*. There is this call to action about the conference, but presumably that will mostly be people who have already figured this out to some degree. But (to carry on the author's analogy) what is the recommendation for driver's ed?
- swyx 3y ago
  we haven't launched it widely yet but if you peek at the top level nav you'll find the course we are working on :) https://www.latent.space/s/university https://www.latent.space/s/university
  - sanderjd 3y ago
    How about that! :) One interesting thing is that there do seem to be courses available for this, but I still haven't come across any books. Maybe this is just because I'm a dinosaur, but I really feel like what I'm missing is a book about this, with a good Introduction and Chapter 1 motivating the subject and giving a lay of the land. I'm sure every techie publisher will have one of these by the end of the year, but so far I really haven't seen what I think I'm looking for in this space. (But having said that, you can bet I'll check out your course.)
    - swyx 3y ago
      thank you! yeah i guess its easier to iterate on a course than a book, but ofc we are also effecitvely writing and market testing the book contents. having written my own book before i'm not particularly keen on doing that again but am trying to work with partners to do this :)
      - sanderjd 3y ago
        Ha, totally get it.
petercooper 3y ago
It turns out that most people are terrible at using the tools, and there's even research about this right now Thanks – this was a surprisingly fun paper to read. The authors are all CS researchers (and good ones – Björn Hartmann does a lot of great work in this area in particular) but I'd love to see some behavioral scientists tackle it at some point too. I haven't got a citation but the concept initially reminded me of some studies from ~20 years ago where academics were wondering if the public would ever be able to use search engines effectively and how people's mental models could adapt to do so. All these years later, I'm not entirely sure if the public ever really did adapt, or if it was just the search engines that did..

sanderjd 3y ago

I'm someone who would like to get "really good really fast", but have found that the on-ramps remain pretty weak. For instance, I have yet to find a good book on the subject. There are a ton of tutorials and articles, but it's maddening to try to cobble together any depth of understanding from those little nuggets. And there are tons of good papers on how the systems actually work, but these are not very useful for people who are new to this set of tools to figure out how to use them to create useful things. But I think this will hole will close up incredibly quickly over the next six months. (And it's yet another really good opportunity for people like you to be involved in that gold rush!) Edit to add: For instance, the article lightly lambasts peoples' recommendations to read up on AI / ML fundamentals, and contains this: > n the near future, nobody will recommend starting in AI Engineering by reading Attention is All You Need, just like you do not start driving by reading the schematics for the Ford Model T. But, amazingly, it doesn't actually suggest an alternative starting point*. There is this call to action about the conference, but presumably that will mostly be people who have already figured this out to some degree. But (to carry on the author's analogy) what is the recommendation for driver's ed?

swyx 3y ago
we haven't launched it widely yet but if you peek at the top level nav you'll find the course we are working on :) https://www.latent.space/s/university https://www.latent.space/s/university
- sanderjd 3y ago
  How about that! :) One interesting thing is that there do seem to be courses available for this, but I still haven't come across any books. Maybe this is just because I'm a dinosaur, but I really feel like what I'm missing is a book about this, with a good Introduction and Chapter 1 motivating the subject and giving a lay of the land. I'm sure every techie publisher will have one of these by the end of the year, but so far I really haven't seen what I think I'm looking for in this space. (But having said that, you can bet I'll check out your course.)
  - swyx 3y ago
    thank you! yeah i guess its easier to iterate on a course than a book, but ofc we are also effecitvely writing and market testing the book contents. having written my own book before i'm not particularly keen on doing that again but am trying to work with partners to do this :)
    - sanderjd 3y ago
      Ha, totally get it.

petercooper 3y ago

It turns out that most people are terrible at using the tools, and there's even research about this right now Thanks – this was a surprisingly fun paper to read. The authors are all CS researchers (and good ones – Björn Hartmann does a lot of great work in this area in particular) but I'd love to see some behavioral scientists tackle it at some point too. I haven't got a citation but the concept initially reminded me of some studies from ~20 years ago where academics were wondering if the public would ever be able to use search engines effectively and how people's mental models could adapt to do so. All these years later, I'm not entirely sure if the public ever really did adapt, or if it was just the search engines that did..