Deploying AI systems requires explicit evidence behind recommendations
Michael Karasick · Michael Karasick, IBM Watson // Data Driven #33 // Jan 2015 (Hosted by FirstMark Capital) · Jan 15, 2015 · at 15:49
Michael Karasick (VP of Innovation at IBM Watson) explaining lessons learned while deploying Watson in clinical oncology settings at Memorial Sloan Kettering.
0:00 / 0:09exact quote · 9.9s
▶ Watch the full episode on YouTube → 720p mp4 · rendered on demand · StarZero watermark
“One of the things we learned about systems like Watson, if you're going to deploy it is people want to know why. You know, give it an answer, and they go, they want evidence.”
quote is from the automated transcript, cleaned for reading: filler sounds and stutters are removed, nothing is rephrased. names can be misheard (the analysis reads context, assessments check outside sources). how →
More from Michael Karasick
IBM Watson was engineered as modular specialized bots, not one model
“The right way to think of Watson is, And the engineering is, is a lot of cooperating little bots. So think of all of the Watson engines as having an NLP component which is trained according to the domain. So there isn't one Watson, there's a bunch of them.”
IBM Watson used Apache UIMA and did not replace Hadoop
“I wouldn't assert that it replaces Hadoop. In fact, it's based on UEMA. It's an Apache project.”
IBM created a $100 million Watson investment fund
“And we created a hundred million dollar investment fund.”
IBM Watson has made three minority investments in startups
“And we've done, I guess, three minority investments in startups who have decided to work with us.”