I wrote about the science of meetings (back before ChatGPT when my Substack became entirely about AI)
@emollick
-
Meeting Efficiency: Training for Cost Reduction
By
–
I think a lot about the fact that one of the single most expensive things you can do at a company is call a meeting. And anyone can do it! And people aren’t trained on how to run meetings well & efficiently!
-
AI Detectors in Schools: Low Accuracy and Bias Against Non-Native Writers
By
–
Do not use AI detectors in schools! Accuracy of AI detectors, already low, dropped to 22% when researchers added simple changes, like purposeful spelling errors. Worse, the writing of non-native speakers are more likely to be falsely flagged as AI writing https://
arxiv.org/ftp/arxiv/pape
rs/2403/2403.19148.pdf
… -
Doctor Skill Variation: Critical Impact on Patient Outcomes
By
–
The skill of doctors:
Bottom quartile surgeons have over 4x the complications as the best in the same hospital
A 1SD better doc lowers chance of giving birth to an unhealthy child by 6.3% Skill differences among radiologists accounts for 44% of the differences in diagnoses -
AI Agents Share Private Information Without Memory Retention
By
–
This is a very neat idea taking advantage of a feature that AI agents have that humans do not – the ability to evaluate a situation without retaining memory of what it “saw.” AI agents can reveal private information to each other to reduce asymmetry. https://
arxiv.org/pdf/2403.14443
.pdf
… -
AI Impact on Jobs Understanding Task Bundles Systems
By
–
To understand the impact of AI on jobs, you need to understand how jobs operate as bundles of tasks, and how jobs themselves fit within larger systems From my new book, Co-Intelligence, out next week. If you pre-order you can get access to companion GPTs: https://
moreusefulthings.com/book -
Patron Saints of Technology Standards and Ethics
By
–
Given the banning of the Lena image from computer science journals (which is what the subject of the photo wanted), re-upping this thread on the secret Patron Saints of Technology standards.
-
Authors as Experts in Human Innovation and Prediction
By
–
For those who don’t know, the authors are the experts on how humans innovate & make predictions.
-
GPT-4 Matches Human Experts in Forecasting and Ideation
By
–
We have controlled experiments from key scholars in relevant fields showing GPT-4 is competitive with humans in:
1) Forecasting, where it meets or exceeds crowds of experts.
2) ideation, where it exceeds the idea quality of trained students
More academic expert testing is needed! -
LLMs Exceed Human Accuracy in Fact Assessment
By
–
This is the first evidence that LLMs can exceed human levels when assessing the accuracy of facts. A big potential milestone.