3 ms·

I'm not sure it's clear what viewpoint you're referring to.

• PaulHoule 4y ago
  The viewpoint described in the article, particularly the emotion of envy. I have been working on this stuff since 2005 or so and I can do more than I ever did and if there is one thing I am certain of it is that other people are going to research things and publish even better models in the future that I'm going to download from huggingface or whatever comes next. (On the other hand the fact that I know how to pose well-defined problems and I am willing to work hard to make training sets really does put me in the top 5%. If I am kicking myself it is for the times when I didn’t have the courage to ask clients to put the resources in to make training sets.)
  • propogandist 4y ago
    how do you convince clients to improve data collection?
    • PaulHoule 4y ago
      These people wanted an audiovisual matching engine a bit beyond the state of the art. The only way they were going to get that then (or now for that matter) is hiring people to label the data. They were funded and had a lot of traction and I think I could have talked them into it if I’d been thinking a little bit.