SOPA Images/Getty Images
- Lux Capital's Josh Wolfe is bullish on AI investments in biology and robotics.
- AI's role in biology includes startups like Recursion and EvolutionaryScale.
- Wolfe said the value in robotics lies in data repositories for training, not humanoid forms.
With the AI arms race underway and a host of industries investing in the technology, Lux Capital venture capitalist Josh Wolfe identified the areas he's particularly keen on.
In a recent episode of "The Logan Bartlett Show," the Lux Capital cofounder and biotech VC said that his firm has invested in two main sectors within AI over the past two years — biology and robotics.
Wolfe predicted that in a couple of years, these fields, alongside aerospace and defense technology, will have people "very bullish" about investing.
AI in biology
"AI into biology is really interesting because most computer scientists underappreciate how hard biology is," he said. "Biology is really complex — it is not linear, it is not 'programmable biology' the way many people think."
In turn, Wolfe added that most biologists similarly "underappreciate how sophisticated computer science and AI" are.
"There's people that are versed at the interstices between those two things that I think are poised to do really well," he said.
Wolfe said that the first stage of AI and biology included biotech companies like Recursion — which Lux Capital invested in — that uses AI to translate pictures of cells into computable data and identify biologically meaningful relationships, according to the company's website.
"AI 2.0, which is where we are today, is actually using large language models which very few people have other than EvolutionaryScale," he said, giving a shoutout to an AI startup founded by ex-Meta researchers, one of whom previously started a biotech company with Wolfe.
After Meta axed the group's project, the researcher approached Wolfe, the VC said, who offered him $25 million on the spot.
"He has the only frontier language model in biology, and he can really, from prompt to protein, produce things," he said. "So it's jumping out of the two-dimensional space of code into the three-dimensional space of actually producing a protein."
The company, which has raised over $142 million in seed funding led by Nat Friedman, Daniel Gross, and Lux Capital, partnered with Amazon Web Services to launch its AI model, ESM3, in June.
ESM3 is "capable of generating entirely new proteins that have never existed in nature," AWS wrote in a press release. And the company has already created its first protein, a non-previously evolved version of a fluorescent protein.
"There's many ways that evolution has found this way, but they found a path that evolution never found, which is pretty cool," Wolfe said. "So the implication for being able to do antibodies, small molecules, drugs, proteins is pretty profound."
AI in robotics
While "robots" typically invoke images of futuristic humanoid tech like Elon Musk's Optimus, Wolfe said he thinks "humanoid robots are not going to be the way."
"Focusing on the training data, the repository, for universal robot intelligence is what they're focused on," he said. "I think that's where the value is going to be."
Although Wolfe thinks humanoid robots will attract money for their intrigue, he said that robots mimicking human capabilities are "really inferior to what robots can do."
Rather, he said that the value of robotics will be the repository of data that can train robots to exist in an unconstrained environment. Similar to how LLMs are trained off large batches of known information from the internet to develop predictive chatbots.
"I think that the winning models will be highly specialized robots that do one or two specific things, and they do it extraordinarily well," he said.
Wolfe said he thinks mass competition will emerge for creating a universal operating system between humanoid robotic companies, including China's Unitree Robotics.
"But I think that the value will accrue to the intelligence," he added.