Individuals preserve telling me to look at The Pitt, and now that the present is tackling generative AI in healthcare, I have to. As Mashable’s tech editor, I subject numerous questions on generative AI, and I additionally spend numerous time speaking to people who find themselves extraordinarily keen about synthetic intelligence, and people who find themselves extraordinarily hostile to AI.
So, what does the newest episode of The Pitt Season 2 get proper about AI in drugs, and what does it get flawed?
How AI components into The Pitt episode “8:00 AM.”
Credit score: Warrick Web page / HBO Max
AI wasn’t your complete focus of Season 2, episode two, “8:00 AM.” We additionally get to see a nun with gonorrhea in her eye, an unhoused man with a colony of maggots inside his moldy arm solid, and a transparent view of an erection, which, after all, has lasted greater than 4 hours.
Nevertheless, one of many plotlines focuses on newcomer Dr. Baran Al-Hashimi’s (Sepideh Moafi) try to modernize the ER and combine AI into affected person care. She introduces an AI app that robotically listens to affected person visits and summarizes the pertinent particulars of their charts.
When an excited scholar says, “Oh my god, are you aware how a lot time this can save?”, Dr. Al-Hashimi has a solution: With the brand new AI app, the ER docs will spend 80 % much less time charting. Later within the episode, the nice physician tells Dr. Robby (Noah Wyle) that it’ll additionally permit physicians to spend 20 % extra time with sufferers.
Thus far, so good, however the app instantly makes a mistake. It paperwork the flawed treatment within the affected person’s chart, substituting a similar-sounding treatment as an alternative.
This does not diminish Dr. Al-Hashimi’s enthusiasm in any respect. “Generative AI is 98 % correct at current,” she says. “You will need to all the time rigorously proofread and proper minor errors. It is wonderful however not excellent.”
AI transcription is actually good (and it nonetheless makes errors).
Dr. Al-Hashimi states that generative AI is 98 % correct. However is that this actually true?
Truth-checking Dr. Al-Hashimi’s declare is definitely difficult, as a result of it is not completely clear what she means. If she’s referring purely to AI transcription, she’s nearer to the reality. That is a job that generative AI excels at.
Final yr, a bunch of researchers performed a scientific assessment of 29 research that measured AI transcription accuracy particularly in healthcare settings. (You possibly can try your complete assessment, printed within the BMC Medical Informatics and Determination Making journal.)
Mashable Mild Pace
A few of these research did discover accuracy charges of 98 % or increased, and a few discovered vital time financial savings for docs. Nevertheless, these research concerned managed, quiet environments. In multi-speaker environments with numerous crosstalk and medical jargon, such as you’d discover in a crowded emergency room, accuracy charges have been a lot decrease, typically as little as 50 %.
Nonetheless, speedy developments in large-language fashions are bettering AI transcriptions on a regular basis. So, we may very well be beneficiant and say Dr. Al-Hashimi’s declare is near the reality, for the newest LLM fashions, in sure settings.
Generative AI is certainly not 98 % correct.
Let’s take the newest model of ChatGPT for example. When the GPT-5.2 mannequin was launched a number of months in the past, OpenAI printed documentation on the mannequin’s tendency to hallucinate and supply false data.
In response to OpenAI, its GPT-5.2 Considering mannequin has a median hallucination fee of 10.9 %. That’s actually excessive, particularly contemplating that OpenAI desires ChatGPT that will help you with medical questions. (The corporate just lately launched ChatGPT Well being, a “devoted expertise in ChatGPT designed for well being and wellness.”)
Now, when GPT-5.2 Considering is given entry to the web, its hallucination fee drops to five.8 %. However would you belief a physician who’s flawed 5.8 % of the time — and solely after they can use the web to verify their work? And would you need your AI healthcare app to be related to the web in any respect?
Generative AI could at some point be 98 % correct. However we’re not there but.
Generative AI cannot substitute docs.
Credit score: Warrick Web page / HBO Max
Dr. Robby and different characters have a dialog about “intestine” instincts on this episode. That is one thing that generative AI cannot replicate, and one motive that many individuals do not need to see AI instruments changing human staff completely, whether or not within the arts or drugs.
The episode additionally spends numerous time displaying examples of empathy. The very best docs do not simply have an encyclopedic information of their specialty. They do not simply have good intestine instincts. They’re true healers, who know that holding a affected person’s hand on the proper second could be simply as essential as the fitting prognosis.
To be truthful, most AI healthcare instruments I’ve seen aren’t making an attempt to exchange docs. As an alternative, they need to give healthcare staff extra diagnostic instruments and save them time. As Dr. Al-Hashimi says repeatedly within the episode, she desires to provide docs extra time at their sufferers’ bedside.
AI will help save docs time.
Radiology, or medical imaging, is likely one of the most promising functions for generative AI in healthcare settings. When radiologists at Northwestern College carried out a customized generative AI instrument to assist analyze X-rays and CT scans, their productiveness elevated by 40 % — and with out compromising accuracy.
I feel even numerous AI skeptics would agree that is a constructive end result, for sufferers and docs.
The Pitt does appear to be organising Dr. Al-Hashimi as one thing of a villain — or, at the least, a foil for Dr. Robby — however good drugs and generative AI aren’t essentially in battle. As with all instrument, it may be terribly useful — or terribly harmful.
Disclosure: Ziff Davis, Mashable’s dad or mum firm, in April 2025 filed a lawsuit in opposition to OpenAI, alleging it infringed Ziff Davis copyrights in coaching and working its AI techniques.
Subjects
Synthetic Intelligence
HBO
