Experts warn healthcare leaders to ignore AI hype
Artificial intelligence hype is waning, according to one industry founder. Joseph Villa, the head of Bigbee Labs, predicts a coming “AI winter” driven by the gap between actual capabilities and vendor promises. Villa shared these views during the second annual HealthAI summit in Albuquerque, New Mexico, where health leaders gathered to discuss technology adoption.
The event featured an interactive display where attendees posted sticky notes with questions and concerns. Organizers later added solutions to the wall, creating a visual record of the group’s collective thinking. Villa used the occasion to address the growing presence of AI in medical settings, urging organizations to distinguish between real utility and marketing rhetoric.
Understanding the technology limits
With decades of software development experience, Villa clarified how the technology functions. Large language models work by predicting and selecting text sequences, while agentic AI systems rely on software tools to attempt actions toward a goal. Neither method inherently shows human-like understanding or checks the accuracy of its own output.
This distinction matters because exaggerated claims can waste resources and endanger patients. Villa warned against “AI washing,” a practice where companies stretch definitions to make products seem more advanced. Despite these risks, the technology holds value. LLMs can transcribe speech, translate languages, and convert clinical notes into structured data. They also support medical imaging and diagnostic workflows.
Risk assessment is essential when evaluating these tools. A system that misidentifies an appointment time causes minor inconvenience, but an incorrect diagnosis carries serious consequences. Leaders must ask if a system will produce errors, whether those errors are detectable, and what happens if they go unnoticed.
Villa cautioned against agentic AI in healthcare. Many products labeled as agentic are simply standard workflows with an LLM added to a specific step. Because clinical decisions carry significant weight, uncontrolled trial-and-error approaches are unsuitable for most applications.
Focusing on practical use cases
“I think the bill is going to come due,” Villa remarked. “We’re going to go into an AI winter.” He predicted a period of experimentation followed by a “cleansing” where organizations remove the term AI from their websites. This cycle mirrors previous technology hype waves, such as expert systems and the semantic web.
Instead of chasing promises of human-like intelligence, healthcare organizations should prioritize practical use cases. Villa concluded that the technology can deliver meaningful results when leaders focus on measurable benefits and appropriate safeguards rather than assuming capabilities it has not demonstrated.
Leave a Reply