I use a continuous glucose monitor, but I'm not sure I see what the weight loss payoff Kennedy imagines from…
Healthcare AI News 4/30/25
News
Meta rolls out Meta AI, featuring voice chat, a Discover feed that shows how others use AI, optional personalization that draws from the user’s Facebook and Instagram accounts, and support for Meta’s AI glasses.
Google’s NotebookLM can now create its podcast-like Audio Overviews in 50 languages.
Children’s Hospital Los Angeles launches a sensor-based home apnea risk evaluation that uses a newly created data collection app and Apple Watch algorithms.
AI company Infinitus Systems releases patient- and provider-facing voice AI agents that automate outbound phone calls.
In the UK, Great Ormond Street Hospital for Children evaluates ambient documentation systems in an NHS-funded project.
Business
Healthcare AI workflow automation platform vendor Plenful raises $50 million in a Series B funding round.
Cleveland Clinic will use coding assistant and CDI AI tools from Akasa.
Ambulatory health IT vendor CareCloud launches a domestic and offshore AI Center of Excellence to further integrate AI across its technologies and processes. The company says the 50-member team will have 500 employees by the end of 2025.
A drug company and biotech firm will pay precision medicine technology vendor Tempus $200 million in data licensing and model development fees to create a cancer drug development model. Tempus says it has spent billions of dollars over the past decade to develop a database of the de-identified clinical data of cancer patients.
Research
UCSD researchers determine that a gene that was thought to be a biomarker for Alzheimer’s disease may actually cause the condition, which they determined using AI to analyze protein structures.
University of Zurich researchers secretly tested the persuasiveness of AI-generated Reddit comments, some of them falsely claiming to be from rape victims or a trauma counselor. Their 1,700 bot-written posts were 3–6 times more effective at changing user opinions than human ones, raising concerns about AI-created disinformation.
Other
A 19-year-old Case Western pre-law student and a law journal editor write a 50-page paper on the need for healthcare-specific AI regulations in Ohio, after which they were invited by lawmakers to help draft a bill on the topic.
Today I learned from a sign that veterinarians are using AI-powered ambient documentation. ScribbleVet’s digital scribe generates SOAP notes in real time, which the company says reduces end-of-day charting from two hours to 20 minutes. Pricing starts at $40 per user per month, while the full-featured plan offers unlimited SOAP notes, dental charts, callback summaries, customizable templates, and a medical record summary for $150 per full-time vet.
A woman who asked ChatGPT for a palm reading was surprised when it instead flagged a mole on her hand as potentially acral lentiginous melanoma, a rare skin cancer, and recommended that she see a dermatologist. She hasn’t yet posted the result.
Contacts
Mr. H, Lorre, Jenn, Dr. Jayne.
Get HIStalk updates.
Send news or rumors.
Follow on X, Bluesky, and LinkedIn.
Sponsorship information.
Contact us.
Re: ChatGPT flags concern over skin cancer. Even though not asked to do so.
I’ll bet $20 it’s just an ordinary mole, and ChatGPT is way out of it’s depth here. Why?
Standard “is this mole skin cancer” advice goes as follows:
– has the mole (recently) increased in size, or changed shape in any way?
– is the edge of the mole fuzzy?
– is the shape of the mole irregular?
If the answer to all these questions is “No”, then you can be reassured that it’s “just an ordinary mole”. I mean, lots of people have lots of moles, and the vast majority of those are not cancerous. If we had physicians chasing every mole out there, then they’d not have enough time to do real work.
There is also the matter that this particular mole in on the individual’s hand. People know their hands. Intimately! I’ll bet this mole has been there since birth, is unchanged in all that time, and has never caused any problems or concerns.
ChatGPT is very likely giving this person a cancer scare, without any justification. ChatGPT is more qualified to do the palm reading.
I find it interesting that “WebMD will tell you it might be cancer” became a meme that everyone used as shorthand to reference the difficulty of diagnosing something without an actual clinician, but now “ChatGPT says it might be cancer” is being (attempting to be) touted as a life saving paradigm shift.
I think this really illustrates the phenomenon where the fact that computers (in general, but specifically LLMs of late) are automatically seen as an “authority”. To many people, something coming from an LLM is automatically considered to be “probably right”. It really underscores the risk that tools like this will further ingrain biases and reinforce the status quo – “deepen the ruts” so to speak.