Berkeley SkyDeck is leaning more into artificial intelligence with its latest batch of startups, more than half of which are tapping into AI across various sectors.
The accelerator at UC Berkeley selected 20 companies for Batch 17, most of which are expected to pitch during a demo day event on campus Wednesday.
SkyDeck received more than 2,200 applications for this round, a 23% increase over the previous batch, bringing its acceptance rate for this cohort to less than 1%.
Its acceptance rate for Batch 16 was 1.3%.
That ability to be highly selective may belie an increasingly competitive landscape for accelerators.
"The accelerator space is noisy," SkyDeck Executive Director Caroline Winnett told me. "Challenges and opportunities are always two sides of the same coin. So, with the explosion of generative AI, everybody's challenge is to pick those companies that really do something meaningful."
Winnett is also still seeing a lot of hype.
"It is getting very frothy out there. So I think there will be a lot of dead bodies along the road eventually," Winnett said. "We need to continue to ... make sure we're picking founders who understand the pain, who can really build, and who are going to hustle and keep at it until they've really built for that problem."
As long as the AI component has real business applications, founders accepted into SkyDeck can be working on anything from foundational large language models to flower-delivery logistics, Winnett said.
"That's really the core of everything. You can build all sorts of nice AI models, and you can have your lovely (large language models), and you can dive into OpenAI and ask questions. But if you're not really solving a pain point, 20 other startups are going to come do the same thing," Winnett said. "At some point, generative AI ... won't be a differentiator. And in fact, it's already not a differentiator."
Among the current batch is Sunnyvale-based Cerevox which is building tools to help develops transform data for use in large language models, or LLMs.
San Francisco-based MyMagic AI is also tapping into LLMs with open-source models to help businesses use and transform data.
CoGrader, also based in San Francisco, is developing AI-powered software for educators to help speed up grading assignments.
SkyDeck also accepted its second quantum computing-related startup, an Israel-based company called LQL which is developing a quantum lidar system for automotive uses.
View the full list of companies presenting during SkyDeck's Batch 17 Demo Day here.