Advertisement · 728 × 90

Posts by Nicholas Tatonetti, PhD

Accent related errors in clinical speech transcription and a LLM-based remedy - npj Digital Medicine npj Digital Medicine - Accent related errors in clinical speech transcription and a LLM-based remedy

We then tested a simple correction pipeline: WhisperX transcription → LLM correction

Passing transcripts through an LLM told to “adjust the errors” reduced word error rates by ~4.5–6.9 pts

This also narrowed the gap between groups

www.nature.com/articles/s41...

1 month ago 1 0 0 1

Some patterns stood out
1) the gap was largest in base Whisper with ~11 point higher word error rates for non-native speakers
2) errors clustered in clinically dense language
3) more biomedical terminology > higher error rates - medication/disease names were especially vulnerable

1 month ago 1 0 1 0

We evaluated Whisper/WhisperX
Participants read clinical texts covering typical scenarios (colonoscopies or desc. clinical history)
We include
• native & non-native English speakers
• with and without medical training
Across models, errors were higher for non-native speakers

1 month ago 1 0 1 0
Accent related errors in clinical speech transcription and a LLM-based remedy - npj Digital Medicine npj Digital Medicine - Accent related errors in clinical speech transcription and a LLM-based remedy

GI fellow Jamil Samaan noticed something odd when using AI ambient scribes and speech transcription tools
Clinicians with non-native English accents appeared to confuse the model
He and Yassi Fatapour decided to investigate, now out in npj Digital Medicine
www.nature.com/articles/s41...

1 month ago 2 0 1 1
Preview
A crisis of overconfidence: Why confidence, not accuracy, is the real risk in clinical AI - BioData Mining BioData Mining - Language models today are trained to convey confidence in their outputs, regardless of whether those outputs are correct. The alignment methods we use to make them helpful also...

New editorial in BioData Mining: A crisis of overconfidence in clinical AI led by @cedarssinaihsu.bsky.social PhD student, Jacob Berkowitz.

Why sounding confident can be more dangerous than being inaccurate, and why calibration needs to be central in clinical AI.

link.springer.com/article/10.1...

2 months ago 1 1 0 0
Post image

You can now connect ChatGPT to an Apple Watch.
So I imported 29 mil steps & 6 mil heartbeats into the new ChatGPT Health.
It graded my heart health an F. ⁉️
Cardiologist @erictopol.bsky.social called it “baseless.”
Any bot claiming to give health insights shouldn’t be this clueless. Even in beta. 🧵

2 months ago 94 34 7 12

I suspect Holly Mitchell is going to find in the coming years that she shares Henry Waxman's legacy as someone who blocked a whole part of the LA region from getting rail for decades due to needless, pointless capitulation to fearmongering. Her time on the Metro Board will not be remembered kindly.

2 months ago 77 6 1 0
Preview
Cedars-Sinai Will Pioneer AI in Drug Safety With New Funding Cedars-Sinai receives an up to $5,054,235.00 award to develop KronosRx, a platform using AI and 'patient avatars' to predict adverse drug reactions, improve cli

Excited to announce our new project with ARPA-H which advances methods to artificially accelerate the aging of organoids to study drug toxicities over time!! A current huge gap in our understanding and our models.

www.cedars-sinai.org/newsroom/ced...

2 months ago 0 0 0 0
Preview
From prompt engineering to agent engineering: expanding the AI toolbox with autonomous agentic AI collaborators for biomedical discovery - BioData Mining

Is "Agent Engineering" the next big thing? Getting so-called "AI Agents" to work together effectively isn't trivial. @moorejh.bsky.social and I discuss the need for rigorous development in our latest editorial in BioData Mining

biodatamining.biomedcentral.com/articles/10....

4 months ago 3 2 0 0
Advertisement
Post image

Deskilling? cacm.acm.org/news/the-ai-... #deskilling #ai

5 months ago 1 1 0 0
Microsoft Forms

👀 Health AI PhD Information Session hosted by Cedars-Sinai! This event is an opportunity for interested students to learn more.

November 21 @ 8 AM PT
Register: forms.office.com/r/vMreBMZQak

5 months ago 0 0 0 0
Preview
PhD in Health Artificial Intelligence

Application period for the Cedars-Sinai PhD Program in Health AI is now OPEN! Come study cutting-edge AI in a top hospital with clinical rotations, clinical collaborators, & easy access to clinical data - please RP! www.cedars-sinai.edu/education/gr... #artificialintelligence #bioinformatics #phd

6 months ago 12 7 0 0
Preview
Opinion | There’s Nothing Magical in the Machine

Nice piece in the NYT about the risks of "magical thinking" with AI that allow big tech to hype their products www.nytimes.com/2025/09/25/o... #artificialintelligence #magic #hype #llms

6 months ago 7 1 0 0
Preview
California’s $23 Billion Plan to Restore Federal Cuts to Scientific Research

Loving my state rn.

California’s $23 Billion Plan to Restore Federal Cuts to Scientific Research www.nytimes.com/2025/09/13/u...

6 months ago 2 0 0 0
Preview
Leveraging Large Language Models in Extracting Drug Safety Information from Prescription Drug Labels - Drug Safety Introduction Adverse drug reactions (ADRs), including those resulting from drug interactions, remain a leading cause of morbidity and mortality. Structured product labels (SPLs) serve as a primary sou...

AI methods are great tools for advancing drug safety surveillance. We show that LLMs can accurately identify and extract essential concepts from messy product drug labels. Just published in _Drug Safety_!

Great work led by @undinagislad!

link.springer.com/article/10.1...

7 months ago 7 1 0 1
Preview
An LLM Analysis of Reporting Quality of Randomized Clinical Trial Articles This study uses a zero-shot large-language model to automatically assess compliance with the Consolidated Standards of Reporting Trials guideline for English-language randomized clinical trials publis...

We used AI to audit 21,000+ RCTs. The result? -- only 15% of trials fully comply with CONSORT guidelines. Transparency in clinical trials has a long way to go.

Amazing work by Apoorva in my group!

jamanetwork.com/journals/jam...

7 months ago 2 0 0 2
Advertisement

The top three science funders are the US, China, and Japan. The three fund more science than every other country, combined.

In 2024, the US funded more scientific research than China and Japan, combined. The US also funds more science than every other country except for Japan and China, combined.

7 months ago 92 46 1 0
Video

The 2025 volume of the Annual Review of Biomedical Data Science is now online. Take a look at the most read article from this new volume, "Spatial Transcriptomics Brings New Challenges and Opportunities for Trajectory Inference" arevie.ws/45D2NWD
@marylynritchie.bsky.social @tatonetti.bsky.social

7 months ago 2 1 0 0

This creates a fun game that rewards superfluous, outlandish (dare I say outre?) verbiage.

8 months ago 3 2 1 1

With the correct rhetorical device you can make complete jibberish sound like poetry. Just ask Shakespeare (or better read Forsyth's The Elements of Eloquence)

8 months ago 0 0 0 0
From left to right. Patient facing left and looking at their phone. Doctor facing right looking at a tablet.

From left to right. Patient facing left and looking at their phone. Doctor facing right looking at a tablet.

Just came across this fail. Literally everything that's wrong with AI/tech in healthcare. Not only is the patient and doctor buried in their devices, but they're facing AWAY from each other 🤦‍♂️!!

Technology should facilitate and center the humanity of healthcare. Not do whatever this is.

8 months ago 2 0 0 0

I’m petrified about today’s science news. Genetically modifying crabs to have cheetah genes? This could go sideways fast.

9 months ago 22508 4085 798 310
Preview
Vibe coding: a new paradigm for biomedical software development - BioData Mining

What is Vibe Coding and how can it be used to advance biomedical research? @tatonetti.bsky.social & I explain in this editorial biodatamining.biomedcentral.com/articles/10.... #bioinformatics #datascience #coding #vibecoding

9 months ago 8 4 1 3
Preview
Vibe coding: a new paradigm for biomedical software development - BioData Mining

Vibe coding is coming for biomedical software development, my first editorial as Co-Editor-in-Chief of @BioDataMining with @moorejh.bsky.social

biodatamining.biomedcentral.com/articles/10....

9 months ago 6 2 0 0

If anyone is wondering, yes we will be monitoring the safety and side effects of drugs from the moment they hit the market

9 months ago 1 0 1 0

So the market for our drug safety monitoring algorithms just opened up 👀

9 months ago 0 0 0 0
Advertisement
Preview
Biomedical text normalization through generative modeling A large proportion of electronic health record (EHR) data consists of unstructured medical language text. The formatting of this text is often flexibl…

Excited that our paper on GenAI-based text normalization is out at Journal of Biomedical Informatics!

If you ever had a bunch of messy text and need to map it to a controlled terminology, then this is the method for you.

www.sciencedirect.com/science/arti...

Special 👏 to Jacob Berkowitz!!

10 months ago 1 0 0 0
Preview
TBD3 Online registration for Columbia OHDSI Summer School starts on April 9th 2025. Online registration ends July 11th 2025.

Learn observational data from the best. @ColumbiaDBMI and @OHDSI are offering summer school to learn all the real-world-data analysis tips and tricks.

columbiauniversity1.regfox.com/ohdsi-summer...

10 months ago 1 0 0 0
Post image

✅ PASSED SENATE TRANSPORTATION COMMITTEE: SB 445 by @scottwiener.bsky.social that streamlines permit approval process for transit projects to tackle cities that cause costly delay against transit.

11 months ago 72 11 0 0

What a fabulous discussion and meeting! Off to a fun start.

1 year ago 3 0 0 0