Undeterred Capital has invested in AI for drug discovery since this early-stage venture firm launched in 2020. Today, it’s portfolio of nearly 20 companies is collectively valued at over $1 billion. Founder and managing partner Joe Wilson, MBA, shared insights on building successful life science companies, and the current state and future of AI in healthcare.
Few investors understand how to build companies that are highly technical or highly scientific
“At universities, every time you turn a corner, there’s another scientific breakthrough happening,” said Wilson. “But you don’t see those breakthroughs often becoming companies.” Recognizing this gap, Wilson developed an investment strategy for backing entrepreneurs at a very early stage, even before incorporation.
“When I look at an entrepreneur, it starts with the science,” said Wilson. “One thing that we say at the firm is ‘the breakthrough is the beginning.’ The breakthrough is often the culmination of years or decades of work, but it's really the start of the journey for us and for the entrepreneur.
“The second thing we think about is what is the commercialization potential of that breakthrough? How big is that market? And then third, and maybe most important, we look at the founder. Who is the founder? What are their strengths? Have they shown an ability to grow into a CEO role? We spend a lot of time vetting the entrepreneur for before we make an investment.”
To find promising investments, Undeterred partners with research institutions throughout the United States. Wilson lectures on management and entrepreneurship and holds office hours. He also writes about these subjects and encourages people to reach out to him directly.
“Through this we meet some incredible companies at a really early stage,” he said. “Many don't work out or don't even become companies. But every once in a while, we meet a very rare entrepreneur that has the ability to build something big that's just had a breakthrough.”
Data sharing and collaboration are needed for AI to transform healthcare
A major roadblock to AI in healthcare and biotech is a lack of data, in Wilson’s view. “ChatGPT, or the underlying transformer models — those are trained on the internet corpus of data. It's a tremendous amount. But in drug discovery, there's not enough. We have to create processes and systems to share larger amounts of data so these models can be trained up.”
Wilson noted the need to create collaborations that generate these data, included data that are publicly accessible and not just in the purview of a few large pharmaceutical companies. Such sharing would help more entrepreneurs have an opportunity to create. As these models get more powerful, being a bit protective also is warranted, including anonymizing patient data, he added.
Is AI on the verge of transforming drug discovery? Wilson thinks not. Between the lack of data and other hurdles, he estimated that the timeline is three to five years. “I don’t think it’s 20 years away,” he said. “But today we’re seeing mostly incremental improvements.”
Safety guidelines for AI are urgent
“There needs to be a really big discussion on safety, on using AI in biotech and healthcare, and what systems we are putting in place to make it safer,” said Wilson. It’s time to “start thinking about how to design organizations and structures to protect against, what if things go off the rails? Or what if acceleration is faster than you think?”
As possible pitfalls, he points to the potential for biased models that don't account for different ethnic groups, or the appropriate use of data in designing drugs, like synthesizing new compounds that can potentially be dangerous.
Wilson predicts that these issues will come to a head soon. “Think of all the anger surrounding COVID-19 and the theory around whether it was designed in a lab,” he said. “If you multiply that by a hundred, that will be the firestorm of AI designing potentially lethal drugs.”
“I'm not advocating for super stringent regulations, but I do think the dialogue should start now,” said Wilson. He points to Open AI as a model. “From the moment they started as a nonprofit, a big part of their mission was being thoughtful around designing artificial intelligence solutions. I think we need to do the same for biotech and healthcare.”
