|
Umang Bhatt
@
umangsbhatt
Cambridge, England
|
|
Research Fellow @PartnershipAI. PhD Student @CambridgeMLG @LeverhulmeCFI. prev: @CarnegieMellon, @Microsoft. forever: @BAPS.
|
|
|
315
Tweetovi
|
399
Pratim
|
448
Osobe koje vas prate
|
| Tweetovi |
|
Umang Bhatt
@umangsbhatt
|
4 h |
|
Loving this summary by @ecekamar on work with @bansalg_, @dsweld, and @erichorvitz regarding backwards compatibility, dissonance in model errors, and designing for complementarity #SafeAI2020 #AAAI20
Also: good morning, New York :)
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
1. velj |
|
Love-hate relationship with Barcelona.
PROs: experiencing a mix of urban grunge and rich Spanish history was refreshing; plus @FAccTConference was lovely
CONs: my laptop bag (and its contents) found a new home, thanks to ladrones. I hope the new owners enjoy my (ex-)machine 😭
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
30. sij |
|
Fascinating, any pointers on where to start in the social sci literature? Is it safe to say that the social context of a decision (or intervention as a result of a model output) in the real world would be a precursor to the need for a justification or explanation?
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
30. sij |
|
Ooo love the hot take. Would you say explanation is a social construct, in most cases?
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
30. sij |
|
Would love to hear the thoughts of @alicexiang on this, given her legal and statistics superpowers
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
30. sij |
|
Huge shoutout to @alicexiang as well who helped shape the work all throughout :)
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
30. sij |
|
Sooo cute!!! I hope the name came to you at first sight :)
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
30. sij |
|
Thanks! Would love your thoughts. Huge shoutout to @alicexiang who helped shape the work all throughout :)
|
||
|
|
||
| Umang Bhatt proslijedio/la je tweet | ||
|
Ayşe Naz Erkan
@naz_erkan
|
30. sij |
|
One of my favorite talks at @fatconference by @umangsbhatt summarizing the state of explainable machine learning boldly. Paper: arxiv.org/abs/1909.06342 pic.twitter.com/8ZBHEt2QvK
|
||
|
|
||
| Umang Bhatt proslijedio/la je tweet | ||
|
Varoon Mathur
@VaroonMathur
|
30. sij |
|
Fantastic work by @umangsbhatt and co. on explainability here at #FAT2020. Can’t wait to dive deeper into this! twitter.com/umangsbhatt/st… pic.twitter.com/J7xV0xi44i
|
||
|
|
||
| Umang Bhatt proslijedio/la je tweet | ||
|
linnet taylor
@linnetelwin
|
30. sij |
|
#AI explainability is mainly used for internal debugging, its goals are not clearly defined, and it has serious technical limitations. 'we suggest everyone cast doubt on those who are claiming to work with explainability technologies' @umangsbhatt at #acmfat2020
|
||
|
|
||
| Umang Bhatt proslijedio/la je tweet | ||
|
Adrian Weller
@adrian_weller
|
30. sij |
|
@umangsbhatt presenting recent work at FAT* pic.twitter.com/L9SsjtGYKD
|
||
|
|
||
| Umang Bhatt proslijedio/la je tweet | ||
|
Partnership on AI
@PartnershipAI
|
30. sij |
|
PAI research reveals that, in practice, machine learning explainability techniques fall short of transparency goals (arxiv.org/abs/1909.06342). Presenting at #FAT2020 in the Plenary room at 11:20 today, @umangsbhatt & @alicexiang will discuss key findings: partnershiponai.org/xai-in-practic…
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
27. sij |
|
Nice to see LinkedIn using fair rank aggregation in search and deploying feature importance explanations for engineers during model selection; h/t @kkenthapadi in the @fiddlerlabs tutorial #FAT2020
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
26. sij |
|
Please let me know if you're in town and want to connect: currently interested in all things explainability, uncertainty, and robustness #FAT2020
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
26. sij |
|
Will be talking about our paper with @PartnershipAI on XAI in deployment (arxiv.org/abs/1909.06342)
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
26. sij |
|
Looking forward to a fun, productive week in Barcelona for the @fatconference pic.twitter.com/R2qxCr0Nmn
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
24. sij |
|
Gave a huge shoutout to @dkaushik96, his policy work, and their joint exciting work on counterfactual datasets from human-in-the-loop learning 💪🏽
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
24. sij |
|
**sarcastically** “we should use convnets to detect pneumonia in chest X-rays since ofc deep learning sees 20/20” 😬
|
||
|
|
||
|
Umang Bhatt
@umangsbhatt
|
24. sij |
|
|
||
|
|
||