The worlds of technology and medicine are making big bets on AI playing a central role in the delivery of healthcare in the future. Today a startup out of Durham, North Carolina, called Bionic Health — built by two early movers in the commercializing of AI — is throwing its hat into that ring to build out its approach.
The startup has raised $3 million in seed funding to create what it describes as an “AI health clinic”: people can get bloodwork and other diagnostics carried out and monitored, and then an AI — built on OpenAI’s GPT-4 and other large language and machine learning models — provides personalized insights based on the data points coming out of these assessments.
The initial aim is to build out preventative health services rather than primary care for people who, for example, are experiencing chronic pain or a virus. Longer term, co-founder and CEO Robbie Allen believes Bionic Health has the potential to extend into all aspects of doctor-patient care.
“There are just so many areas that can be improved,” he said, highlighting the shortage of both general practitioners and specialists across both developed and developing world communities. “We are losing primary care doctors every day. It’s unacceptable that you can’t get appointments that easily. We may need to automate more of that out of necessity.” Using AI to take on the work of specialists, meanwhile, could also evolve over time, synthesizing more data from across specializations to deliver more accurate insights. “The first line of specialty care could be tech-driven,” he added.
Bionic Health’s first clinic is also, effectively, its first lab: As the startup trains its AI and figures out where best it can be put to work, it will have actual, human doctors involved working alongside that AI, and providing feedback to better shape it. And Bionic’s co-founder Dr. Jared Pelo, Allen said, is its first doctor. There is already a waitlist, and the clinic is in the process of onboarding its first patients, who will start out by paying $250 per month, covering regular assessments and tests as well as the diagnostic services based on them.
AI-Assisted Care technology is designed to assist clinicians with clinical evaluation tasks, Allen said. “We believe it can significantly improve the efficiency and accuracy of the diagnosis and treatment process.”
IDEA Fund Partners, Studio VC, Alumni Ventures, Tweener Fund, AI Operator’s Fund, and Operator.VC all participated in this round. Allen said that originally the aim was to raise $2 million but interest in the startup was high. That’s not just because “generative AI” is all the rage right now (although that will have definitely figured here); Allen and Pelo have a prescient and proven track record when it comes to building lasting AI startups.
Allen’s previous company, Automated Insights, built one of the first generative AI services back in 2007, creating prose out of data and other prompts. Years before CNET found itself embroiled in an AI-writing controversy, the Associated Press invested in and used Automated Insights to write hundreds of articles. The startup is now owned by Vista Equity and works across a big range of enterprise use cases.
Pelo, meanwhile, founded a company called iScribes, which described itself as an “ambient documentation” company aimed at healthcare. It was eventually acquired by Nuance, which itself was acquired by Microsoft, where Pelo worked as chief clinical product officer just until this month, leaping to found Bionic Health with Allen (who had been on iScribes’ board: Durham AI tech guys stick together, I suppose). The GPT-based health documenting service announced just yesterday by Microsoft was based on technology Pelo developed at iScribes and oversaw at Nuance and then Microsoft.
There are a lot of areas where AI might potentially play a role in the world of healthcare: robotics, administrative tools, drug discovery, pathology and clinical interactions are all areas that have seen activity in recent years. However, not everyone thinks that clinical roles are the most ideal of these — not now and possibly not ever. Alexandre Lebrun, the co-founder of another AI healthcare startup called Nabla — coincidentally Lebrun also sold a previous AI startup to Nuance years ago — believes that there is just too high a risk with AI being wrong, and that’s too serious to consider in healthcare scenarios.
“With all large language models, there is a risk,” Lebrun told TechCrunch. “It’s incredibly powerful, but 5% of the time it will be completely wrong and you have no way to control that… In healthcare we [literally] can’t live with a 5% error rate.”
Nabla earlier this month released Copilot, focused just on providing administrative assistance, not clinical advice, to clinicians and patients.
Allen believes that development will be much more of a continuum, and that the evolution will be pushed not just by market and social forces — healthcare being “broken” while ever more people are demanding ever more services — but also by AI technology that has been rapidly evolving.
Allen said he and Pelo first started thinking about what has become Bionic Health with theoretical ideas about “automated doctor-patient” interactions but things changed with early looks at what GPT-4 could do (recall Pelo was at Microsoft, which backs GPT creator OpenAI, until just this month). “As GPT-4 started to become available, it really changed the dynamic,” Allen said, with “a speed, and a much higher percentage [of accuracy] than even a few months ago.”
Providing personal health data using a model like GPT-4, combined with a model for specific guidance and treatments, is already very high even without improving, he added. “And when GPT-5 comes out it could be another significant step forward. I think Nabla might be underestimating the technology.”