Lessons from Alexandre Lebrun, Nabla, on building a defensible and safe Language Learning Model in healthcare
Pear Healthcare Playbook
00:00
Building Defensible Healthcare AI Companies
It's not to use a public LML and just hope that it won't have to be made too much. But having an open source, large language model and fine tune it with your own very high quality data is how we remove from just like text nodes towards more actual automations of lots of clinical tasks. And our job at AI has care of startups is to fine tune these terms. It's difficult. You need to, you need a very strong technical team and you need proprietary data sets. And you need to understand the users of the product to see what's working or not with them. I think this is the most. amazing.
Transcript
Play full episode
