AI agents are “lost” when patients ask medical questions, and it will take time to roll agents out effectively in a clinical setting, according to the founders of two healthtech firms.
Ali Parsa, who previously ran the failed Babylon Health and Circle Health and is now heading up Quadrivia, which has built an AI agent for clinicians, and Vladimir Lupenko, the CEO and co-founder of the orthodontic health firm Impress, were speaking on the Tech.eu podcast.
The pair discussed AI, how their respective startups were leveraging AI, AI challenges and the future of AI in healthtech.
AI agents have been one of the hottest subjects in AI this past 12 months, with AI evangelists seeing them as the future of the workforce.
But in a recent podcast appearance, Andrej Karpathy, co-founder of OpenAI, said functioning agents “will take about a decade”.
Lupenko pointed to the limitations of agents, saying Impress had not implemented any agents within patient care. He said: “If patients start asking any medical questions, the agent is lost.” He said it would take time to develop agents.
He added: “If someone tells me that ‘we can build you an agent for patient care for orthodontic services within two weeks’, I say 'no, it’s impossible, you can answer generic questions but if you go outside of 10 standard questions that a patient could ask, it’s not working at all.'"
Echoing this, Parsa said AI agents do 80 per cent of the “basics very well”, but the problem lies with the other 20 per cent. Parsa said too many US AI startups were rushing to market too quickly, making bold claims in a bid to hoover up customers quickly.
He said: “I think that in healthcare it pays to be cautious, it pays to be transparent."
Would you like to write the first comment?
Login to post comments