| |
|
| |
A MiniMax Research Night during SF Tech Week
Intelligence in the Open is a recurring MiniMax research series bringing together AI researchers, engineers, technical founders, & investors for candid conversations about the work shaping the next generation of intelligent systems.
For our SF Tech Week edition, we're exploring one of the most consequential questions in AI:
What happens when AI systems begin contributing to their own improvement?
This edition will focus on recursive self-improvement (RSI) & the research paths leading toward it - from autonomous research & post-training to self-improving agents, synthetic data, verifiers, evaluation, & systems capable of running increasingly complete experimentation loops.
We'll bring together researchers working across areas including:
Recursive self-improvement
Autoresearch & autonomous experimentation
AI-driven post-training
Synthetic data & self-play
Reinforcement learning
Verifiers & automated evaluation
Self-improving agents & coding systems
Long-horizon agentic research
One area we'll explore comes directly from recent MiniMax research. In PostTrainBench, MiniMax M3 was given four pretrained base models & tasked with autonomously running the full data synthesis training evaluation iteration process, deciding what data to generate, how to train the models, & how to adjust subsequent experiments based on evaluation results.
The broader question goes beyond any single experiment:
How close are we to AI systems that can meaningfully improve the systems that built them?
Each researcher will present recent work in a focused 10-15 minute talk, followed by an open discussion connecting different approaches to self-improvement & autonomous research.
Expect technical conversations, real experiments, open questions, food, & plenty of time to connect directly with researchers & builders working at the frontier.
Speakers
Speakers will be announced soon.
We're currently inviting researchers working across recursive self-improvement, autoresearch, post-training, reinforcement learning, agents, synthetic data, & evaluation.
Schedule
6:00-6:30 PM - Arrival & Check-in
Meet fellow researchers, founders, engineers, & investors before the program begins.
6:30-6:40 PM - Welcome & Opening Remarks
Introduction to MiniMax, Intelligence in the Open, & the evening's focus on recursive self-improvement.
6:40-7:40 PM - Research Presentations
Three to four focused talks presenting recent work across RSI, autoresearch, post-training, agents, & evaluation.
7:40-8:00 PM - Open Discussion & Panel Q&A
Speakers return for a moderated discussion around what actually constitutes self-improvement, where current approaches succeed, & what still stands between today's systems & meaningful RSI.
8:00-9:00 PM - Food & Networking
Continue the conversation directly with speakers & connect with others across the AI research & startup ecosystem.
About MiniMax
MiniMax is a frontier AI lab building multimodal foundation models & AI products for developers, enterprises, & creators.
Presented by MiniMax during SF Tech Week.
Additional speakers, research topics, partners, & venue details will be announced as the program develops.
This event is a part of #SFTechWeek-a week of events hosted by VCs & startups to bring together the tech ecosystem. Learn more at www.tech-week.com.
|
|
|
|
|
|
|
|
