UC Berkeley EPIC Lab

Effective Programming, Interaction, and Computation with Data Lab @UCBerkeley

[Çağatay Demiralp @CagatayDemiralp]

This was a great retreat put together by the @UCBEPIC team. As expected, applications of LLMs were central to many talks and discussions. Identifying the core problems invariant to the particularities of the next big LLM is key for research ROI in these applications.

[UC Berkeley EPIC Lab @UCBEPIC]

Can you explore the space using LLMs - but do it in a way that is efficient? How do we find the high error regions?

[UC Berkeley EPIC Lab @UCBEPIC]

Fereshte Khani from Microsoft describes how to collaboratively develop NLP models, ensuring alignment and safety

[UC Berkeley EPIC Lab @UCBEPIC]

It’s hard to robustly test edge cases in a model and make user defined concepts explicit

[UC Berkeley EPIC Lab @UCBEPIC]

A new system they are working on is Humboldt for data discovery. You shouldn’t have to ask experts about what data you should explore!

[UC Berkeley EPIC Lab @UCBEPIC]

Alex Bauerle from Sigma Computing tells us about what’s hard when building a spreadsheet for cloud data warehouses

[UC Berkeley EPIC Lab @UCBEPIC]

Can you fuse structural understanding of API programs with LLM techniques? Naman provides a way! Parametric templates for the win!

[UC Berkeley EPIC Lab @UCBEPIC]

Naman Jain explores how to summarize data transformation scripts using a template-based approach, informed by LLMs

[UC Berkeley EPIC Lab @UCBEPIC]

LLMs by themselves are insufficient for this task - brittle and hard to control

[UC Berkeley EPIC Lab @UCBEPIC]

Flor allows users to travel back in time to help debug ML training. You can also inspect and “jump into” another user’s training history. Time travel and shapeshifting!

[UC Berkeley EPIC Lab @UCBEPIC]

Rolando Garcia @rogarcia_sanz describes the next generation of Flor, a tool for rapid iteration during ML training via a live notebook demo!

[UC Berkeley EPIC Lab @UCBEPIC]

Haotian leverages large language models to identify visualization intent (variants of BERT) and prior work on automatically translating visualization intent into actual visualizations (eg Lux).

[UC Berkeley EPIC Lab @UCBEPIC]

Haotian Li describes how to support conversation with data via visualization - why write code when you can just talk to your data!

[UC Berkeley EPIC Lab @UCBEPIC]

Can we check extensional equality (ie two programs have similar outputs) for constrained domains like biology? So that we can automatically rewrite and make code more performant — component by component?

[UC Berkeley EPIC Lab @UCBEPIC]

Justin Lubin @jplubin embedded himself in a “wet lab” biology group to identify their programming challenges

[UC Berkeley EPIC Lab @UCBEPIC]

There is a trade off between easy to understand code (eg one that loops through arrays) and those that are performant (eg one that manipulates arrays in NumPy)

[UC Berkeley EPIC Lab @UCBEPIC]

Yet more challenges in Machine Learning - operationalizing, explaining and trusting it.

[UC Berkeley EPIC Lab @UCBEPIC]

Barbara Stanley from Google BQ tells us about user challenges in end to end data analytics/science

[UC Berkeley EPIC Lab @UCBEPIC]

More open challenges in helping novice users through the data science workflow - so that one can go from “zero to hero”