“It’s just encouraging them to think about it in a different way. And in a way, I think that the pandemic has accelerated this, because we have people all over the world working together. And if you do have that structure sorted out, it makes your life a lot easier.”
Leader in biomedical and epi-informatic data science, Paul Agapow turns big data into tremendous medical value to help mankind. We had the opportunity to chat with him and discover the impact that AI is having on pharmaceuticals.
AI Gospel Embraced by Pharmaceuticals
“… The gospel has been adopted, if you will. Like a lot of industries, pharma spoke a lot about AI and data. And particularly data for many years before it really became part of business-as-usual. But the culture has changed”.
“I’ve been able to see that over the last couple of years, we’ve gone from skepticism to “We must do this. What can we do with this data?”. I think success in a lot of ways comes down to not saying “let’s do AI, but what problems do we have to solve?”.
The Obstacle of Fungible Computation
“There is still that tradition of “We gave you guys a computer cluster. That should be enough. We built one and it’ll do you forever”. Whereas, you’re wanting units with GPUs, TPUs, things like that. You might want memory optimisation to run image analysis”.
“That cultural shift and appreciating that computation now is not something that you just buy as a bit of an infrastructure and you’ve solved it forever. Really, you need some sort of flexible structure where you can set up computation units, tear them down, configure them, so on and so forth”.
A Wider Reason for ML Interpretability
“Of course interpretability has gotten a lot of, you know, got a lot of talk and Headspace around the world… I see a kind of wider reason for using interoperability. It can serve as a smoke test on your conclusions. “Why does this model work the way that it does?” And since we’re dealing with such complicated things, people’s health, false data, it can show why those conclusions were reached and you can do a dummy check on them”.
Catch A Whole Lot More On This Week’s Episode!
There’s a lot more to catch on AI, ML and data science in the medical space.
Be sure to give this episode a listen. And feel free to share any of your thoughts, opinions, or comments with us.