Sameera Poduri

AI leadership currently focussed on GenAI for customer support. @uber @aminohealth @jawbone @qualcomm.🤖Robotics PhD USC, ⚙️BTech IIT Bombay. she/her/mom.

[Sameera Poduri @sameerapoduri]

This is a great read and commentary. For me, the last few lines really resonated: "this tumultuous weekend showed just how few people have a say in the progression of what might be the most consequential technology of our age."

[Delip Rao e/σ @deliprao]

The amazing @_KarenHao and @cwarzel document the goings on inside OpenAI leading to the Friday event. Sharing some hightlights with commentary below:

https://t.co/x1D5UX0S5o

[Sameera Poduri @sameerapoduri]

Such exciting possibilities for the future of content creation!

Loved this video illustrating Dale-e 3's capabilities https://t.co/pLNrBolDM0

[Sameera Poduri @sameerapoduri]

@savvyRL Enabling mentorship and collaboration is key - I love what you have already done through ML collective.

[Sameera Poduri @sameerapoduri]

@savvyRL I missed the poll, but I agree strongly. I think it’s more important now than ever to make development of AI tech and applications accessible to everyone.

[Sameera Poduri @sameerapoduri]

@prateeks Nice! Banganapalli mangoes are the best :)

[Nirant @NirantK]

Last CALL!

.@shubhi_says is hosting a Women in AI mixer on 27th May in Koramangala!

🥁🥁🥁

Please share this with your friends' WhatsApp group, post on internal Slack and Teams!

And help other folks find this — RT for Karma please?

https://t.co/dk7Bl0V1zX

[Sameera Poduri @sameerapoduri]

@NirantK @prateeks @shubhi_says @lavanyats @zainabbawa Oh I would love to join, thanks for asking me. I can’t be in Bangalore on the 27th, so will miss this event, but would love to connect with the AI group in Bangalore. I will be there June 4th to 15th.

[Sameera Poduri @sameerapoduri]

Public livestream event on May 19 with several leading genAI researchers.

"In an era in which convincing images, audio, and text can be generated with ease on a massive scale, how can we ensure reliable access to verifiable, trustworthy information?"

https://t.co/df1xPvWoTi

[Sameera Poduri @sameerapoduri]

Another great resource to understand recent developments in open source LLMs.

Ahead of AI #8: The Latest Open Source LLMs and Datasets, by @rasbt https://t.co/UFrqwPcW4P

[Sameera Poduri @sameerapoduri]

Very good summary of the open source LLM movement.

[Andrej Karpathy @karpathy]

Oops haven't tweeted too much recently; I'm mostly watching with interest the open source LLM ecosystem experiencing early signs of a cambrian explosion. Roughly speaking the story as of now:

1. Pretraining LLM base models remains very expensive. Think: supercomputer + months.
2. But finetuning LLMs is turning out to be very cheap and effective due to recent PEFT (parameter efficient training) techniques that work surprisingly well, e.g. LoRA / LLaMA-Adapter, and other awesome work, e.g. low precision as in bitsandbytes library. Think: few GPUs + day, even for very large models.
3. Therefore, the cambrian explosion, which requires wide reach and a lot of experimentation, is quite tractable due to (2), but only conditioned on (1).
4. The de facto OG release of (1) was Facebook's sorry Meta's LLaMA release - a very well executed high quality series of models from 7B all the way to 65B, trained nice and long, correctly ignoring the "Chinchilla trap". But LLaMA weights are research-only, been locked down behind forms, but have also awkwardly leaked all over the place... it's a bit messy.
5. In absence of an available and permissive (1), (2) cannot fully proceed. So there are a number of efforts on (1), under the banner "LLaMA but actually open", with e.g. current models from @togethercompute, @MosaicML ~matching the performance of the smallest (7B) LLaMA model, and @AiEleuther , @StabilityAI nearby.

For now, things are moving along (e.g. see the 10 chat finetuned models released last ~week, and projects like llama.cpp and friends) but a bit awkwardly due to LLaMA weights being open but not really but still. And most interestingly, a lot of questions of intuition remain to be resolved, e.g. especially around how well finetuned model work in practice, even at smaller scales.

[Sameera Poduri @sameerapoduri]

Amazed at how good #midjourneyv51 is. Took me only a few minutes to create these for a presentation.

[Simon Willison @simonw]

Leaked Google document: “We Have No Moat, And Neither Does OpenAI”

The most interesting thing I've read recently about LLMs - a purportedly leaked document from a researcher at Google talking about the huge strategic impact open source models are having
https://t.co/q2lsjTHKGS

[DAIR.AI @dair_ai]

Top ML Papers of the Week (April 24 - 30):

- AudioGPT
- Track Anything
- Agents Learn Soccer Skills
- Harnessing the Power of LLMs
- Scaling Transformer to 1M tokens
- A Cookbook of Self-Supervised Learning
...

[Delip Rao e/σ @deliprao]

"patience, attention to detail, and thinking deeply about small things".

Unless you are building throwaway demos, these three things make the execution of non-trivial ML projects appear as if there is no progress when "progress" involves doing those three things.

[Sameera Poduri @sameerapoduri]

@GergelyOrosz I think @MosaicML

[Sameera Poduri @sameerapoduri]

My team at Uber is hiring ML engineers and researchers in SF/Bangalore/Hyderabad. We work on automating and improving customer support using Conversational AI (chatbots, speechbots, and yes, ChatGPT too :) ), LLMs and multimodal models, Causal ML, etc. We…https://t.co/MfnRG6kp7G

[Sameera Poduri @sameerapoduri]

@prateeks Sorry to hear Prateek. That’s a beautiful tribute. I remember her smile from when I saw her at your place couple of years ago.

[Sameera Poduri @sameerapoduri]

Useful resource to track covid in India

[Rukmini S @Rukmini]

For those in India and overseas looking for Indian covid data as numbers begin to rise, a reminder that https://t.co/whh3VBEbNx, our effort at continuing the work of https://t.co/cgHUiqwhah is up and running. Data can be downloaded from the api as before. Corrections welcome.

[Sameera Poduri @sameerapoduri]

Heartbreaking, tragic 😟 We had all the time to prepare, capacity to produce all the vaccines we need, yet we failed.

[Ashish K. Jha @ashishkjha]

COVID in India horrifying

Nearly 300,000 infection today, more than 2000 deaths

Great @nytopinion piece on what went wrong

Neglect, poor messaging, worse policy, false belief that India had beaten COVID

Humanitarian disaster unfolding in front of us

https://t.co/2kmihOIqVN

[Sameera Poduri @sameerapoduri]

“I am strong when I am relaxed”

A simple reflection from 7yo after her Taekwondo class.