4 ms·

LLM’s synthesize new material that looks most like material they’ve been trained on. In practical terms, that means they do a genuinely good job of synthesizin

LLM’s synthesize new material that looks most like material they’ve been trained on. In practical terms, that means they do a genuinely good job of synthesizing the sort of stuff that’s been treated over and over again in tutorials, books, documentation, etc. The more times something’s been covered. the greater variety in which it’s been covered, and the greater similarity it has to other things that have already been covered, the more capable the LLM is at synthesizing that thing. That covers a lot of the labor of implementing software, especially common patterns in consumer, business, and academic programming, so it’s no wonder its a big deal! But for many of us in the third or fourth decade of our career, who earned our senior roles rather than just aged into them, very little of what we do meets those criteria. Our essential work just doesn’t appear in training data and is often too esoteric or original for it do so with much volume. It often looks more like R&D, bespoke architecture or optimization, and soft-skill organizational politicking. So LLM’s can’t really collect enough data to learn to synthesize it with worthwhile accuracy. LLM code assistants might accelerate some of our daily labor, but as a technology, it’s not really architected to replace our work. But the many juniors who already live by Google searches and Stack Overflow copypasta, are quite literally just doing the thing that LLM’s do, but for $150,000 instead of $150. It’s their jobs that are in immediate jeopardy.

bdangubic 3y ago
Every senior person thinks just like you do... The fact that you "earned (y)our senior roles rather than just aged into them" has nothing to do whether or not your skills can be replaced technology like LLM's. Chances are that you most likely earned your senior role in a specific company / field and your seniority has less to do with your technical skills but more with domain knowledge. Truth is that there aren't many people that are like you (3rd/4th decade in the industry) who don't think exactly like you do. And truth is that most of you are very wrong ;)
- pdimitar 3y ago
  Care to clarify why is your parent wrong? They said that LLMs can't be trained on what's not publicly available, and a lot of it is deeper knowledge. What's your retort?
  - FeepingCreature 3y ago
    Not parent, but this presumes that the current split between training and inference will hold forever. We're already seeing finetunes for specific domains. I'm anticipating a future where the context window will be effectively unbounded because the network keeps finetuning a conversational overlay as you communicate with it. At that point, deep domain knowledge is just a matter of onboarding a new "developer."
    - pdimitar 3y ago
      I know enough about ML/DL but never worked it. Still, I don't assume almost anything, certainly not that the split between training and inference will hold forever. Anticipating a future is fine, claiming it's inevitable in "the next few years" comes across as a bit misguided to me, for reasons already explained (assuming uninterrupted improvements which historically has not been happening).
  - mikepalmer 3y ago
    Context: LLMs learn all the amazing things they do by predicting the next token in internet data. A shocking amount can be inferred from the internet by leveraging this straightforward (I won't say "simple"!) task. There was not explicit instruction to do all that they do - it was implied in the data. The LLM has seen the whole internet, more than a person could understand in many lifetimes. There is a lot of wisdom in there that LLMs evidently can distill out. Now about high level engineering decisions: the parent comment said that high level experience is not spelled out in detail in the training data, e.g., on stack overflow. But that is not required. All that high level wisdom can probably also be inferred from the internet. There are 2 questions really: is the implication somewhere in the data, and do you have a method to get it out. It's not a bad bet that with these early LLMs we haven't seen the limits of what can be inferred. Regarding enough wisdom in the data, if there's not enough, say, coding wisdom on the internet now, then we can add more data. E.g., have the LLMs act as a coding copilot for half the engineers in the world for a few years. There will be some high level lessons implied in that data for sure. After you have collected that data once, it doesn't die or get old and lose its touch like a person, the wisdom is permanently in there. You can extract it again with your latest methods. In the end I guess we have to wait and see, but I am long NVDA!

bdangubic 3y ago

Every senior person thinks just like you do... The fact that you "earned (y)our senior roles rather than just aged into them" has nothing to do whether or not your skills can be replaced technology like LLM's. Chances are that you most likely earned your senior role in a specific company / field and your seniority has less to do with your technical skills but more with domain knowledge. Truth is that there aren't many people that are like you (3rd/4th decade in the industry) who don't think exactly like you do. And truth is that most of you are very wrong ;)

pdimitar 3y ago
Care to clarify why is your parent wrong? They said that LLMs can't be trained on what's not publicly available, and a lot of it is deeper knowledge. What's your retort?
- FeepingCreature 3y ago
  Not parent, but this presumes that the current split between training and inference will hold forever. We're already seeing finetunes for specific domains. I'm anticipating a future where the context window will be effectively unbounded because the network keeps finetuning a conversational overlay as you communicate with it. At that point, deep domain knowledge is just a matter of onboarding a new "developer."
  - pdimitar 3y ago
    I know enough about ML/DL but never worked it. Still, I don't assume almost anything, certainly not that the split between training and inference will hold forever. Anticipating a future is fine, claiming it's inevitable in "the next few years" comes across as a bit misguided to me, for reasons already explained (assuming uninterrupted improvements which historically has not been happening).
- mikepalmer 3y ago
  Context: LLMs learn all the amazing things they do by predicting the next token in internet data. A shocking amount can be inferred from the internet by leveraging this straightforward (I won't say "simple"!) task. There was not explicit instruction to do all that they do - it was implied in the data. The LLM has seen the whole internet, more than a person could understand in many lifetimes. There is a lot of wisdom in there that LLMs evidently can distill out. Now about high level engineering decisions: the parent comment said that high level experience is not spelled out in detail in the training data, e.g., on stack overflow. But that is not required. All that high level wisdom can probably also be inferred from the internet. There are 2 questions really: is the implication somewhere in the data, and do you have a method to get it out. It's not a bad bet that with these early LLMs we haven't seen the limits of what can be inferred. Regarding enough wisdom in the data, if there's not enough, say, coding wisdom on the internet now, then we can add more data. E.g., have the LLMs act as a coding copilot for half the engineers in the world for a few years. There will be some high level lessons implied in that data for sure. After you have collected that data once, it doesn't die or get old and lose its touch like a person, the wisdom is permanently in there. You can extract it again with your latest methods. In the end I guess we have to wait and see, but I am long NVDA!