วีดีโอ

A+, have been curious about ColPali and this was both insightful and easily understood by a non-technical

Thank you for making the video!

ReBase: Training Task Experts through Retrieval Based Distillation

Thanks for uploading!

Great interview but encourage the host to give a little more rope to the speaker to explain themselves fully

❤

Hi please improve the audio quality by speaking closer to the mic , I listen to you guys every month, keep up the good work

Great interview thanks ! 🎉

the discussion at 17:20 is mostly referencing results in table 1, not table 2 as mentioned but otherwise great ep!

That wasn’t horrible

Today I missed the webinar, but I'm happy to find out it's available on TH-cam.

Thanks. Any news on OpenAI's mysterious "Q" project? When asked about it, Sam said they weren't ready to talk about it yet but it is rumored it was behind part of the big split that happened at OpenAI.

salute 🫡

thank you for the update:)

I miss this podcast format! Can't get these discussions anywhere

These are always GOLD !

Thanks! :)

Great work! BTW how to generate this paper topic distribution visualization?

Wooow !! Thank you so much its like falling in a gold mine while looking for copper

Lovely talk

nice:)

Good discussions and good depth. Thanks for sharing!

i really love your roundups i helps me soo much

Thanks for the roundup as always! Monthly round ups are long cycles for the field! :D

Always love the monthly roundup! Thanks guys, your TH-cam channel is awesome

Love this series and channel, please keep putting out awesome content!

Very interesting session! Love it, keep going!

Awesome interview! Love listening to Sarah and thanks for putting out awesome content ZA -- excited to see the series on IR methods with Colbertv2 coverage. Great job.

are emergent abilities of LLMs still there tho? i just saw jason weis rebuttal to those arguments of it being a mirage

Can the paper chat handle tables? Also using gpt multimodal to handle the figures would be awesome!

Great video and visualization!

Can you please share the slides??

Great conversation

Adding timestamps would make these videos much more accessible. Maybe y'all can use AI for this?

very interesting tool.

Exciting times.

Express yourself through a comment on my channel. I can't wait to read what you think! 📝

Express yourself through a comment on my channel. I can't wait to read what you think! 📝

Great interview! Angela was a bit misleading in the "open source" section though, by omitting what is not open-sourced and instead focusing on what is open-sourced. Namely none of their models are open sourced (nor datasets to the best of my knowledge), meaning you can't use any of them for commercial applications and you have to retrain the models from scratch (due to the CC-BY-NC 4.0 license). Which means you have to have 100+ GPUs if you want to train 54B MoE in any reasonable timeframe. With 100 A100 GPUs you need roughly a month. :) (assuming of course you have MT experts and a crystal ball at hand otherwise a lot of additional inefficiencies creep in)

how to retrain this on custom data?

Great interview! Looking forward to the developments with agents and chaining (ChatGPT Plugins like Zapier, AutoGPT, etc.), wonder if/how lots of what has been developed for traditional ML methods can be reused here. Pipelining for complex Data Science use cases, including lots of data engineering pre-processing and multiple models in parallel or series has become a commodity, we can do this in code (e.g. Kedro) or in tools such as Alteryx, Dataiku, Sagemaker, etc.

it seems noteable the mentioned AI optimists have career interests in AI. the mentioned regulations seem not to address some other concerns, like use in surveillance and the military.

Thank you for your effort I find these updates extremely helpful

please add the timestamps

this is such a good interview. so informative and cutting edge.

how can we watch the full speech given by Professor Wang, during the conference event?

I wish the audio was better. TH-cam should provide subtitles when audio is not that great.

For quality vs quantity. I feel the conversation is a little superficial, I hope the host will spend more time on more technical stuff.

Thank you so much for the helpful and interesting informations about this amazing topic! You definately gained another subscriber :) btw. I'm a student and tried to sign up for the free account on zeta-alpha. But when I click on the link in the FAQ section, I just get the message "unauthorized". Can you help me get access?

great breakdown and explanations, thank you!