4 ms·

The marketing has made it very hard to have a real conversation about IBM Watson. There is no such singular thing as "Watson". IBM offers a ML solution for heal

The marketing has made it very hard to have a real conversation about IBM Watson. There is no such singular thing as "Watson". IBM offers a ML solution for health, for NLP, chatbots, etc. They all have very different capabilities and require different levels of machine learning. The marketing is BS, but most of the tech is real - if you give IBM your data, let them train a model on it, and communicate what you want, you will get an end-to-end custom solution. It's just not the magic IBM sells in its marketing videos. Disclaimer: SWE at Watson Health

shostack 9y ago
How do they market it to a tech audience who presumably see through this sort of thing due to a technical understanding of the subject matter?
- pgodzin 9y ago
  For me, Watson Health is actually working on a lot of interesting work in the healthcare space that will eventually lead to improvement in care and outcomes. And hospitals are exactly the kind of clients that IBM handles well - complicated data requirements, government regulations, etc. It's a very difficult, and often frustrating, space to get into.
- saghm 9y ago
  From other responses in this thread, it sounds like they don't, but instead sell it to the executives who make purchasing decisions.
peterburkimsher 9y ago
"if you give IBM your data, let them train a model on it, and communicate what you want, you will get an end-to-end custom solution" If you give a summer intern your data and communicate what you want, they can also give you a custom solution at a fraction of the price. IBM's model might be more advanced, but the real challenge is applying the conclusions, rather than just building a model. Communicating what you want is the hardest part. Most managers who are excited about AI (including my boss) have no idea what they want to use it for. "Making things more efficient" or "analysing what's going on in the factory" is way too vague. Getting data is not so hard, but getting good metadata is a challenge. For example, downloading some music in Chinese and tagging the artist/play count/lyrics/etc. When I have all that metadata, I might not need the machine learning part - I can just find the top 1000 songs and process them manually in an afternoon. Before spending millions on Watson, I think companies would have a higher return on investment by hiring a summer intern to come up with ideas for further research, and gather the data. Then they can apply AI if they really need it.
agumonkey 9y ago
Good to hear from someone working on W.Health. Just yesterday I saw the latest ad. on Watson which is now promising to help Basketball recruiters spot future stars .. which is quite a let down compared to the original promises of Watson. Anyways, any way for someone outside IBM to read about the health division effort and collaborations ?
- pgodzin 9y ago
  https://www.technologyreview.com/s/607965/a-reality-check-for-ibms-ai-ambitions/ https://www.technologyreview.com/s/607965/a-reality-check-fo... was a good recent review. The only IBM source I know of are the press releases announcing new collaborations: https://www-03.ibm.com/press/us/en/presskit/27297.wss https://www-03.ibm.com/press/us/en/presskit/27297.wss
  - agumonkey 9y ago
    Thanks, that's a nice and long list. I still wish I could get a more technical view of what you're doing in medicine, are there conferences or symposium ?
    - pgodzin 9y ago
      I also found https://www.ibm.com/blogs/watson-health/category/case-study/ https://www.ibm.com/blogs/watson-health/category/case-study/ and https://www.ibm.com/blogs/watson-health/category/white-paper/ https://www.ibm.com/blogs/watson-health/category/white-paper... that show case studies and white papers. There are also some publications linked at https://www.research.ibm.com/healthcare-and-life-sciences/ https://www.research.ibm.com/healthcare-and-life-sciences/ and http://researcher.watson.ibm.com/researcher/view_group_subpage.php?id=2420 http://researcher.watson.ibm.com/researcher/view_group_subpa... http://researcher.ibm.com/researcher/view_group_pubs.php?grp=6743 http://researcher.ibm.com/researcher/view_group_pubs.php?grp... I'm on the SWE side, while the research side does the conferences. I know that there were health conferences that IBM researchers have presented at but I'm not aware of which ones they were. Hope this helps!

shostack 9y ago

How do they market it to a tech audience who presumably see through this sort of thing due to a technical understanding of the subject matter?

pgodzin 9y ago
For me, Watson Health is actually working on a lot of interesting work in the healthcare space that will eventually lead to improvement in care and outcomes. And hospitals are exactly the kind of clients that IBM handles well - complicated data requirements, government regulations, etc. It's a very difficult, and often frustrating, space to get into.
saghm 9y ago
From other responses in this thread, it sounds like they don't, but instead sell it to the executives who make purchasing decisions.

peterburkimsher 9y ago

"if you give IBM your data, let them train a model on it, and communicate what you want, you will get an end-to-end custom solution" If you give a summer intern your data and communicate what you want, they can also give you a custom solution at a fraction of the price. IBM's model might be more advanced, but the real challenge is applying the conclusions, rather than just building a model. Communicating what you want is the hardest part. Most managers who are excited about AI (including my boss) have no idea what they want to use it for. "Making things more efficient" or "analysing what's going on in the factory" is way too vague. Getting data is not so hard, but getting good metadata is a challenge. For example, downloading some music in Chinese and tagging the artist/play count/lyrics/etc. When I have all that metadata, I might not need the machine learning part - I can just find the top 1000 songs and process them manually in an afternoon. Before spending millions on Watson, I think companies would have a higher return on investment by hiring a summer intern to come up with ideas for further research, and gather the data. Then they can apply AI if they really need it.

agumonkey 9y ago

Good to hear from someone working on W.Health. Just yesterday I saw the latest ad. on Watson which is now promising to help Basketball recruiters spot future stars .. which is quite a let down compared to the original promises of Watson. Anyways, any way for someone outside IBM to read about the health division effort and collaborations ?

pgodzin 9y ago
https://www.technologyreview.com/s/607965/a-reality-check-for-ibms-ai-ambitions/ https://www.technologyreview.com/s/607965/a-reality-check-fo... was a good recent review. The only IBM source I know of are the press releases announcing new collaborations: https://www-03.ibm.com/press/us/en/presskit/27297.wss https://www-03.ibm.com/press/us/en/presskit/27297.wss
- agumonkey 9y ago
  Thanks, that's a nice and long list. I still wish I could get a more technical view of what you're doing in medicine, are there conferences or symposium ?
  - pgodzin 9y ago
    I also found https://www.ibm.com/blogs/watson-health/category/case-study/ https://www.ibm.com/blogs/watson-health/category/case-study/ and https://www.ibm.com/blogs/watson-health/category/white-paper/ https://www.ibm.com/blogs/watson-health/category/white-paper... that show case studies and white papers. There are also some publications linked at https://www.research.ibm.com/healthcare-and-life-sciences/ https://www.research.ibm.com/healthcare-and-life-sciences/ and http://researcher.watson.ibm.com/researcher/view_group_subpage.php?id=2420 http://researcher.watson.ibm.com/researcher/view_group_subpa... http://researcher.ibm.com/researcher/view_group_pubs.php?grp=6743 http://researcher.ibm.com/researcher/view_group_pubs.php?grp... I'm on the SWE side, while the research side does the conferences. I know that there were health conferences that IBM researchers have presented at but I'm not aware of which ones they were. Hope this helps!