How To Read AI Research Papers Effectively

According to a recent survey, over two-thirds (66.9%) of developers and machine learning teams are planning production deployments of LLM apps in the next 12 months or “as fast as possible” - and 14.1% are already in production! Given the rapid rate of progress and constant drumbeat of new foundation models, orchestration frameworks and open source libraries - as well as the workaday challenges of getting an app into production - it can be difficult to find the time to digest and read the dizzying array of cutting-edge AI research papers hitting arXiv.
That task has never been more critical, however, as the time between academic discovery and industry application moves from years to weeks. How can teams discover and read AI research papers quickly without losing nuance, with an eye toward pragmatic application, while balancing real-world challenges?
In this session, Aparna Dhinakaran - who blends a background in academia with experience overseeing AI in production and troubleshooting real-world AI systems as co-founder and Chief Product Officer of Arize AI - will be joined by data scientist and machine learning engineer Amber Roberts to talk through strategies for understanding and applying the latest research, reducing mean time to application. The session will include an exercise of digesting 1-2 to be announced papers (will be a recent release!) in real-time.
Survey papers:
- A Survey of Large Language Models:
arxiv.org/pdf/2303.18223v12.pdf
- Retrieval-Augmented Generation for Large Language Models:
A Survey, arxiv.org/pdf/2312.10997.pdf
- Benchmarking paper:
HellaSwag: Can a Machine Really Finish Your Sentence,
arxiv.org/pdf/1905.07830.pdf
-Breakthrough paper (deep dive):
Mistral AI Mixture of Experts, arxiv.org/pdf/2401.04088.pdf
- Slides:
docs.google.com/presentation/...
About DeepLearning.AI:
DeepLearning.AI is an education technology company that is empowering the global workforce to build an AI-powered future through world-class education, hands-on training, and a collaborative community. Take your generative AI skills to the next level with short courses help you learn new skills, tools, and concepts efficiently.
About Arize:
Arize AI is an AI observability and LLM evaluation platform. The company’s LLM observability tools - including its popular task-based LLM evaluation libraries and tools for troubleshooting LLM traces and spans, RAG, and prompt iteration - are counted on every day by top enterprises. Learn more about the company’s platform and open source libraries at Arize.com and phoenix.arize.com.
Speakers:
Aparna Dhinakaran Co-Founder and Chief Product Officer
/ aparnadhinakaran
Amber Roberts Machine Learning Engineer
/ amber-roberts42

[@mrsscharff]

Refreshing! As an academic, I was interested to attend and love the idea of paper reading sessions. Thank you to the 2 amazing speakers today.

[@seyiakadri]

Thank you so much. The categorization of papers was intuitive

[@omanshsharma6796]

Beautiful video, thank you so much!!

[@reakal7740]

Thank you for this insightful session!

[@BishalPaudel0]

Thank you so much, this was really helpful.

[@cemtekesin9033]

Thank you this was really helpful

[@pablocabrales8]

What a great session, congrats!

[@benamorzahia3087]

Thanks for the amazing session

[@ramzirebai3661]

Thank you so much

[@samsondawit]

This was great. Thanks for organizing it and hopefully you'll do more of the same.

[@user-if8rt2is9l]

Where can we get the slide that was presented in here ?

[@pierremariani9842]

slides from link shared in the email. Not sure if it will work for other people docs.google.com/presentation/d/18u-Xk-oVI9kmlQUAKlXszXz2nmxu8Zzp0zBVGFcbRmg/edit?_hsmi=299309630#slide=id.p

[@sanderhahn1556]

docs.google.com/presentation/d/18u-Xk-oVI9kmlQUAKlXszXz2nmxu8Zzp0zBVGFcbRmg/edit#slide=id.p

[@douglasdrumond]

They sent it to the email used to register for this event.

[@ishareef03]

Is there a way to get the slides if we didn't register? Can someone share the link here?

[@bohdanforostianyi6369]

They are also available in the description
docs.google.com/presentation/d/18u-Xk-oVI9kmlQUAKlXszXz2nmxu8Zzp0zBVGFcbRmg/edit#slide=id.p

[@surenkhalsa]

great video..can we have more such one on regular frequency plz..

[@ShadowD2C]

Great information

[@Stewz66]

was this video filmed a while ago? I'm surprised people aren't aware that 8x7b is the hottest open-source model on HF.

[@dahiruibrahimdahiru2690]

Any advice about looking at the code? I usually find myself jogging between the code and the paper.

[@mariyanzarev6423]

The trademark DeepLearningAI 720p videos :D

[@MuhammadOmar-qx6nh]

59:00

[@MuhammadOmar-qx6nh]

44:00

[@MuhammadOmar-qx6nh]

11:39

[@MuhammadOmar-qx6nh]

32:40

[@MuhammadOmar-qx6nh]

20:18

[@flaviorabelobarros6417]

Summary:
The session, aimed at helping attendees navigate the complex landscape of AI research, includes discussions on the importance of staying updated with the latest research, strategies for identifying and digesting valuable papers, and leveraging community resources like social media and paper reading groups.
Key points covered include:
- An overview of how AI research papers can be categorized into surveys, benchmarks, and breakthrough papers, each serving different purposes in the field.
- Practical advice on how to approach and digest papers, including where to find them and how to use tools and resources to understand complex topics.
- The role of paper reading in keeping pace with rapid advancements in AI, highlighting the shift from academic discovery to industry application.
- A detailed walkthrough of reading a specific research paper on the "Mixture of Experts" model, demonstrating how to extract and comprehend key information and implications from the paper.
The event also features contributions from experts like Apara and Amber from Arise AI, who share insights on AI observability and evaluation platforms, as well as personal experiences from their careers in machine learning and AI research. The session emphasizes the importance of community engagement and continuous learning to stay abreast of evolving AI technologies.

[@Cloudyraman]

Arize ai documents.

[@AC-tn4it]

I’m sorry what

[@MuhammadOmar-qx6nh]

55:57