Singapore Spotlight

Following up on the "Shout-outs" section at the September MeetUp, we've rustled up a super selection of speakers for this one, demonstrating the overall strength of the Singaporean AI ecosystem!

Talks:

"Concepts for Reliability of LLMs in Production" - Jun Yu Tan
LLMs are adept at writing emails, but how might we ensure that they are appropriate for sending out to customers? Jun Yu will explain some of the risks and pitfalls for automating emails using LLMs, and the techniques that can be used to safeguard ourselves when building AI-enabled products.

"An approach to solving The ARC Challenge with a System of Multiple Expert Agents" - John Chong Min Tan
The ARC Challenge is a difficult because it tests for adaptability to various unknown rules and requires a versatile program to solve it. John's team attempts to solve the ARC Challenge using LLMs as a system of multiple expert agents, with different text-based abstraction spaces - grid, pixel, object - achieving 50 out of 111 training set solves (45%). John will share insights and potential improvements on this new piece of work!

"No Free Lunch with Open-endedness" - Dr. Raymond Chan
Some of the techniques that we've talked about at previous MeetUps may sound like they are too-good-to-be-true. Dr. Raymond Chan will discuss what a Free Lunch is, and how (for instance) Open-endedness techniques may not be be in violation of the No Free Lunch Theorem.

"Token-Crisis and OpenMoE, scaling LLM with less data and computation" - Fuzhao Xue
Token-Crisis is a LLM scaling paper accepted at NeurIPS 2023. And OpenMoE was announced a couple of months ago on Twitter : there's now an active effort to create a 8Bn parameter Mixture-of-Experts. Fuzhao Xue is the lead contributor of these two projects. Currently a PhD student at NUS, and having interned at Google and Nvidia, our speaker will discuss his innovative research.

Talks will start at 7:00pm and end at around 8:45pm, at which point people normally come up to the front for a bit of a chat with each other, and the speakers.

As always, we're actively looking for more speakers - both '30 minutes long-form', and lightning talks. For the lightning talks, we welcome folks to come and talk about something cool they've done with keras_core, TensorFlow, PyTorch, JAX and/or Deep Learning for 5-10mins (so, if you have slides, then #max=10). We believe that the key ingredient for the success of a Lightning Talk is simply the cool/interesting factor. It doesn't matter whether you're an expert or and enthusiastic beginner: Given the responses we have had to previous talks, we're sure there are lots of people who would be interested to hear what you've been playing with. If you're interested in talking, please just introduce yourself to Sam or Martin at one of the events.