2 ms·

The things which look like "easy" AI problems are actually harder than they seem. There are no clean abstractions here, if you don't know the math you are kinda

The things which look like "easy" AI problems are actually harder than they seem. There are no clean abstractions here, if you don't know the math you are kinda hosed doing something as simple as fine tuning an LLM for some specific use case, or even building a good application. For people who a really actually good, there are millions of opportunities outside doing "foundation" level work.

JustExAWS 1y ago
There are in fact abstractions https://aws.amazon.com/blogs/aws/customize-models-in-amazon-bedrock-with-your-own-data-using-fine-tuning-and-continued-pre-training/ https://aws.amazon.com/blogs/aws/customize-models-in-amazon-... https://cloud.google.com/vertex-ai/generative-ai/docs/models/tune-models https://cloud.google.com/vertex-ai/generative-ai/docs/models... https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/fine-tuning-overview https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/... So now where do you think things will be in 2029 by the time her son graduates? And you still haven’t answered the question - why would this be something that everyone isn’t pursuing, there are already people who are doing it and upskilling in it and what are the chances that he is going to be competitive coming from a tier 2 school as a new grad?
- danielmarkbruce 1y ago
  Those are tools, not abstractions. What happens when you try to fine tune your model and it doesn't get better (or it gets worse) at the task you are optimizing for? What happens when it gets better in training then doesn't get better on live data? You actually have to know what you are doing in this space. The question was answered - the field is complicated - deciding to do it and being capable aren't the same thing. The field is not going to be inundated with capable people. By your thinking, one cannot be paid millions of dollars to be a lawyer because of "competition".
  - JustExAWS 1y ago
    And guess how many of those lawyers are recruited from a tier 2 school? Also companies aren’t paying new grad lawyers millions of dollars. What are the chances that he is going to have the capabilities? Saying that he can make a lot of money by chasing after the same thing everyone else is chasing after and has a years long head start is like saying he should become a football player because of the money he can make in the NFL. The number of companies that need any more than the abstractions available from the cloud providers is vanishly small and will grow smaller in 4 years. Those companies aren’t hiring new grads from tier 2 schools.
    - danielmarkbruce 1y ago
      You are obsessed with schools. Most people forget that within a year of working. Many lawyers making millions are from second tier law schools. In computer science, Jeff Dean went to the university of minnesota.
      - JustExAWS 1y ago
        So you can show an anecdote. But look at the colleges where almost all of the lawyers of the top tier law firms attended, where new grads who are hired at BigTech etc. Or you can look at the median earnings of CS graduates who graduated from the top colleges vs 2nd tier colleges after 5 years.
        danielmarkbruce 1y ago
        White shoe law firms aren't the only places people are making millions. I deal with lawyers making millions at law firms you never heard of and never will hear of. Why look at the median? This kid seems like he's far and away above the median. There is a correlation between best school and best performers, but big tech as an example does their own recruiting and at least in my experience at google, most engineers did not attend one of the very top schools. My team had an engineer who never ever graduated college. Drop the obsession with prestigious schools and workplaces. Plenty of people are making a very handsome living without the badges from the best places.

JustExAWS 1y ago

There are in fact abstractions https://aws.amazon.com/blogs/aws/customize-models-in-amazon-bedrock-with-your-own-data-using-fine-tuning-and-continued-pre-training/ https://aws.amazon.com/blogs/aws/customize-models-in-amazon-... https://cloud.google.com/vertex-ai/generative-ai/docs/models/tune-models https://cloud.google.com/vertex-ai/generative-ai/docs/models... https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/fine-tuning-overview https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/... So now where do you think things will be in 2029 by the time her son graduates? And you still haven’t answered the question - why would this be something that everyone isn’t pursuing, there are already people who are doing it and upskilling in it and what are the chances that he is going to be competitive coming from a tier 2 school as a new grad?

danielmarkbruce 1y ago
Those are tools, not abstractions. What happens when you try to fine tune your model and it doesn't get better (or it gets worse) at the task you are optimizing for? What happens when it gets better in training then doesn't get better on live data? You actually have to know what you are doing in this space. The question was answered - the field is complicated - deciding to do it and being capable aren't the same thing. The field is not going to be inundated with capable people. By your thinking, one cannot be paid millions of dollars to be a lawyer because of "competition".
- JustExAWS 1y ago
  And guess how many of those lawyers are recruited from a tier 2 school? Also companies aren’t paying new grad lawyers millions of dollars. What are the chances that he is going to have the capabilities? Saying that he can make a lot of money by chasing after the same thing everyone else is chasing after and has a years long head start is like saying he should become a football player because of the money he can make in the NFL. The number of companies that need any more than the abstractions available from the cloud providers is vanishly small and will grow smaller in 4 years. Those companies aren’t hiring new grads from tier 2 schools.
  - danielmarkbruce 1y ago
    You are obsessed with schools. Most people forget that within a year of working. Many lawyers making millions are from second tier law schools. In computer science, Jeff Dean went to the university of minnesota.
    - JustExAWS 1y ago
      So you can show an anecdote. But look at the colleges where almost all of the lawyers of the top tier law firms attended, where new grads who are hired at BigTech etc. Or you can look at the median earnings of CS graduates who graduated from the top colleges vs 2nd tier colleges after 5 years.
      - danielmarkbruce 1y ago
        White shoe law firms aren't the only places people are making millions. I deal with lawyers making millions at law firms you never heard of and never will hear of. Why look at the median? This kid seems like he's far and away above the median. There is a correlation between best school and best performers, but big tech as an example does their own recruiting and at least in my experience at google, most engineers did not attend one of the very top schools. My team had an engineer who never ever graduated college. Drop the obsession with prestigious schools and workplaces. Plenty of people are making a very handsome living without the badges from the best places.