Johannes Allgaier

Data Science | Machine Learning Explainability | Learning Health Systems | AI in Healthcare

PhD Student in Machine Learning Explainability and Data Science

[Johannes Allgaier @jallgaier_]

@ikeb_wuerzburg Thanks for the wishes!
Proud on my summa cum laude honors.
And thankful for @kdpsinghlab‘s feedback on the last year of my phd journey.

[Johannes Allgaier @jallgaier_]

@ikeb_wuerzburg Thanks for sharing! So excited that this got finally published.

[Johannes Allgaier @jallgaier_]

https://t.co/cBJpYQjb3l
One big step closer to PhD!

Our biggest project so far has now been published. Together with my colleagues Lena Mulansky, Rachel Draelos, and Rüdiger Pryss we carried out a literature review about #XAI in #healtcare.

[Johannes Allgaier @jallgaier_]

@ikeb_wuerzburg Great work! Congrats to the presenters!

[Johannes Allgaier @jallgaier_]

@kdpsinghlab For Ann Arbor, definitely Rent!

[Karandeep Singh @kdpsinghlab]

Statistician: How come you didn’t report the 95% CI in your paper?

ML researcher: So sorry I left it out. Didn’t realize you cared about that. The 95% CI is 3.8.

Statistician: … ?

ML: Well I’m using 4 GitHub Actions for CI. And 95% of that is…

Statistician: Get out.

[Karandeep Singh @kdpsinghlab]

Of all the things related to implementing AI in clinical care, the “Confidential” tag placed on things as simple as which variables go into a model is the silliest.

Transparency should be a precondition to clinical use — let’s remove the Confidential tag from basic model facts.

[Karandeep Singh @kdpsinghlab]

In a press release, OpenAI announced that “ChatGPT can now correctly generate credit card numbers, greatly reducing the number of hallucinated numbers as compared to prior versions.”

[Johannes Allgaier @jallgaier_]

@eotles @TheAirKings So I know whom to call if my heater doesn’t work.

[Johannes Allgaier @jallgaier_]

@arjunmanrai @JamesADiao Such a great time there with you guys! Looking forward to seeing you next year!

[Isaac Kohane @zakkohane]

Emphasis by @peteratmsr that GPT 4 is not fine tuned for medicine and perhaps avoidance of premature fine tuning is essential for attaining it’s general reasoning performance #SAIL2023

[Johannes Allgaier @jallgaier_]

@peteratmsr @SebastienBubeck @srikanthbelwadi @vivnat I agree with @SebastienBubeck and @peteratmsr that this is "an early stage of AGI" (https://t.co/r1rYkkRQD0)

[Johannes Allgaier @jallgaier_]

Closing Keynote with Nina Tousch at #sail23 is live NOW

[Johannes Allgaier @jallgaier_]

@A_U_Kale @kdpsinghlab @SAILhealth Wonderful opportunity to talk and meet the Birmingham team at #sail23

[Arjun (Raj) Manrai @arjunmanrai]

“I don’t believe in inductive bias, I believe in scale.” - ⁦@vivnat⁩ #SAIL2023

[Johannes Allgaier @jallgaier_]

@arjunmanrai @kdpsinghlab @JamesADiao Such a great talk!

[Johannes Allgaier @jallgaier_]

@eotles Runners have 6.5 pace per mile but then run 10k. Welcome to America? ;)

[Johannes Allgaier @jallgaier_]

Having a great time with @kdpsinghlab at #sail23 , meeting great people and learning from giants.

[Johannes Allgaier @jallgaier_]

@eotles Especially, if these solutions are well documented, transparent, explainable and maintainable.

[Karandeep Singh @kdpsinghlab]

Why does a proprietary sepsis model “work” at some hospitals but not others?

Is it generalizability? Measurement? Intervention? Patient population? Margin for improvement? Resource constraints?

Working with a team led by @_plyons, we looked at a 9-hospital network.

A story.