SXSW 2025 Day 1: Small Language Models, Quantum Computing, and the AI Revolution

This is my favorite time of year: the start of South by Southwest.

I’ve been coming to this little conference and festival since 1994, when I was covering the local music scene in Cincinnati and stumbled upon the SXSW Film and Multimedia Conference. 

Throughout the next 31 years, SXSW Interactive would shape my professional life more than I could ever explain. In that time, I’ve spoken on half a dozen panels, moderated dozens more, hosted the SXSW Pitch for its first seven years, written for SXSW magazine, evaluated thousands of panel pitches, and evangelized about the power of connection here. 

I’ve watched this little conference, which had 500 attendees in its first year, grow into a must-attend event attended by tens of thousands of people from around the world. I’ve watched the ebbs and flows of content—some years, it skews hard science, some years, it bends toward content, and some years, it reaches in so many directions it’s hard to see what’s coming next.

This year feels like it’s leaning into the hard sciences: Robots. Quantum Computing. GenerativeAI. Space.

And in my experience, when SXSW Interactive builds around the hard sciences, we’ll see wondrous things in the next 3-5 years as people begin getting their hands on these new tools.

What I Think I Learned Today

• Biggest AI Shift: Small Language Models are going to become wildly important for many industries, including robotics, medicine, and other hard sciences. Combined with GenerativeAI, these models will allow specific fields to build AI tools that solve specific problems.
• Smarter Robots, Sooner: It sounds like robotics companies are using these Small Language Models to help teach robots more generalized skills, which may allow them to be deployed for personal use much sooner than expected.
• The Future of AI-Powered Labs: Medical research labs using Generative AI will soon be run by robots, which will handle everything from extracting materials from experiments to writing early drafts of research papers using these Small Language Models.
• Almost…Quantum Computing: I thought Quantum Computing was on the precipice of arriving in a very real way, but today’s discussions popped that bubble. Quantum is very good at logistics (creating efficient air traffic schedules, for instance), but the limitations are still very real. We are generations of computers away from real use cases beyond logistics and optimization.
• Generative AI in medicine: The good news is that GenAI is very good at developing early-stage drugs and selecting participants for studies. However, the models aren’t powerful enough to replace (or even speed up) clinical trials with humans.

Panel Summaries

*These summaries were written by Claude.AI, which I trained on my writing style and gave my notes from the sessions. I’ve edited them, added links, corrected spelling, and tweaked a few points. The factual mistakes below are from me. The notes are taken from what the speaker said during the session. I didn’t record anything, so please fact-check!

MIT Technology Review: 10 Breakthrough Technologies

The “MIT Technology Review: 10 Breakthrough Technologies” panel highlighted several groundbreaking innovations poised to reshape our future. The Vera C. Rubin Observatory in Chile, housing the largest digital camera ever built, is set to capture first images later this year as part of a 10-year study creating a timelapse of the night sky and a 3D image of the Milky Way. Generative AI search is revolutionizing how we navigate the internet, while Small Language Models offer specialized solutions that can run offline and on personal devices. The panel also discussed fast-learning robots advancing AI capabilities in physical systems. Other highlighted technologies included methane-reducing solutions for cattle, robotaxis beginning global testing in 2025 with companies like WayMo already providing paid trips in San Francisco and cleaner jet fuel initiatives from companies like Twelve and LanzaJet that are moving the aviation industry toward sustainability.

Behind the Quantum Tools Fueling Discovery

The “Behind the Quantum Tools Fueling Discovery” panel featured Trevor Lanting from D-Wave Quantum Systems Inc. and Alejandro Lopez-Bezanilla from Los Alamos National Laboratory, who discussed the revolutionary potential of quantum computing. D-Wave’s technology leverages thousands of interconnected qubits (5,000) with 40,000 interactions between them, using principles from material science to create more precise quantum “crystals” and utilizing quantum tunneling to reach ground states—a distinctive approach that sets them apart in the field. The experts outlined three primary applications: quantum optimization for logistics solutions, providing multiple answers at low energy costs compared to traditional computer clusters, and quantum simulation that enables the creation of “physical systems” within quantum computers to test new materials rapidly without laboratory work, dramatically accelerating materials science research.

The Power of GenAI to Transform the Future of Medical Research

The “Power of GenAI to Transform the Future of Medical Research” panel brought together Alex (Aleksandrs Zavoronkovs) Zhavoronkov (Insilico CEO), Tatyana Kanzaveli (Open Health Network), and Samuel “Sandy” Aronson (Mass General) to discuss AI’s impact on drug discovery and clinical research. Insilico has developed over two dozen new drugs currently in various clinical testing stages, focusing on the intersection of aging and disease. Zhavoronkov tempered expectations by noting that despite AI’s promise, drug development remains extraordinarily challenging—with only about 50 FDA approvals annually and a 90-99.9% failure rate—emphasizing that human biology’s complexity means AI can accelerate drug discovery but not instantly cure diseases. Aronson highlighted a practical application at Mass General, where they’re using AI to reduce the approximately 35% of time researchers spend identifying eligible clinical trial participants, with studies showing generative AI performs well at this task.

Main Photo by Andy Kelly on Unsplash