fei-fei li health care

  from fei-fei li dialogue

stanford hai 18000 audience

sequoia prof -co-director hai denning family chair sponsor

With the current pandemic accelerating the revolution of AI in healthcare, where is the industry heading in the next 5-10 years? What are the key challenges and most exciting opportunities? These questions will be answered by HAI's Co-Director, Fei-Fei Li and the Founder of DeepLearning.AI, Andrew Ng in this fireside chat virtual event.  This event will be moderated by Curtis Langlotz, Director of the Center for Artificial Intelligence in Medicine and Imaging (AIMI Center). If you'd like to submit and upvote questions for our speakers, please register for the Q&A + General access ticket. Submit your questions through the Slido link in your order confirmation email.

question why do you soend lot of time ai health

fei-fei in this area 10 years ago- personal to me - also daughter taking care of chronically ill mother er ity=u surget=ry rooms been through 30 years - my personal understanding importance toi this industry to every family can ai make health sector better

nb parly because dad = doctor - inspired bt dads stories of patients - wanted to go beyond ai in consumer internet qa fe years ago - data sitting there ripe for ai- got to know kurt and his stanford medical school early in this context area

question to ng - yu have desiugned system real time sexpert support- given that why so few in clinical use question doctts plus ai

just because i can publish a paper showing ai can perform still lot of work to make that operational

data drift from one collection system to another even though radiologist can go into 2 different hospotals and use their slighlty different data systems

is the gap you describe uniquue to health or similar in other industrie

it is quite common across sectors

the businness custmer chain in healtcare is more complex

real work nees more than hold out test

gei-fei you ioneer ambient intel

how do you naigate with physiciansunique space of health care that is not yet digitised - namely the hhuman behavior whether thats pateints art home or doctors hoping procedue folows- we have almost zero data on the human behavior - so thats prof arme milstei entered amvbient data of health sensors - alerts when different actors need help

heatlth care muktistakeholder - take humanistc approach from getgo

privacy fairness integrity part of design of ambient system - so on our team bio-ethocs, legal scholars all from sranford working hand in hand

modality of sensing / encryption of device/data- security of different learning models - ethoics of consent; data intgrity inclusion of dufferent patients...

health care abut human dignity so thats how we do it

===

questins anout collab- what makes great colab, best tem

most ofmy work colaboratestarting with phd- most important thing for me is humility - such a technical lens - we are scientists are humans with lot of responsibilty

arnie taught me a lot - when we first talked 10 years ago only understood half of it but his openeess to showing me helth care--  humility doctos and ai experts can come together

question to ng - work with great/humble people deep knowledge... people who wamted my team to be succesful - doing favors for each other withour expectations; trust;

in ai i find choosing projects difficult - best to sit down with cross functional partners - brainstorm possible projects - then choose one thats feasible as well as value in moving health forward

to both of you

is someone bridging imporant - both sides need emathy; willingess of understanding each ot5her lsangiage and problem spaces

i asked students in our team to close laptop and spend days shadowing hralth care - how tiring nurse shift is, doctor doing rounds, family members; so its that kind of empathy that matters

its interesting how ai data can help wider team make recommendations than jutst the opinion of the seniod=r medic

question to andrew

hiw ensureraxialbiasses not perpetuatedbyaihealth

the deployment of machinesis doneby people

data curation to ald dev to inference to decision making to looping back- every step human bias possible

understand bias befoe you evem collect data

rightbalancetraspaency and explainability

ng - a framei use what info person neds o do their job - s this replicated in deign of ai system

from india-us summit