The education system time and time again proves to be the place where knowledge and creativity go to die.
I dream of the day that so many AI algorithms become publicly available that it is hopelessly impossible to ban them; that no student has to write pointless essays again.
I solved a foundational problem for quantum mechanics and all future physics. I am seeking independent patronage to continue unhurried, foundational research.
https://t.co/MHs4xnhL6z
1/2
People say "to predict the next token well you must understand the underlying reality." Not so.
Ptolemy's epicycles predicted planetary positions to within a few arc-minutes, yet captured none of the underlying physics.
Prediction is not the same as understanding.
The West, or perhaps only the Anglosphere, is the historically unique idea that individuals exist not for the sake of their ruler, their class, their society, their ancestors, their gods, or any other such superstition, but for their own sake.
@Anthony_Bonato There is a reality out there of abstract truths. All we can do is conjecture about those truths. We invent knowledge about what is out there; in doing so we may always be mistaken.
@Yoni_Sch@wellingmax Before we throw ai at it, I think we ought to start by specifying precisely the problems climate change yields (eg more intense weather events), and which of those we can already solve with more wealth.
Our latest paper - Task-Driven Wavelets Using Constrained Empirical Risk Minimization - has been accepted at CVPR for an oral presentation.
Congratulations to @EricMarcus_, Ray Sheombarsing and @jonasteuwen for being in the top 3% of all the accepted papers!
🚨 Our #XAI research in EJR emphasizes the vital role of explanations in AI, with a focus on radiology. Even the darkest of black boxes can be explained. #ExplainableAI@EricMarcus_@jonasteuwen https://t.co/1NTqDbU4FR
Two of our lab’s papers — “Task-Driven Wavelets using Constrained Empirical Risk Minimization” and “Kandinsky Conformal Prediction” — have been accepted at #CVPR2024!
Big congrats to @EricMarcus_, Ray Sheombarsing, @jorenwb, and @jonasteuwen :)
https://t.co/1vajyLJDlh
Respect and awe. A palaeolithic person presented with just one of those road signs (let alone one of the advertising signs – let alone what they advertise) could have spent a day marvelling at its beauty and majesty.
Happy to share our preprint of our latest work "Constrained Empirical Risk Minimization: Theory and Practice". We've developed a novel approach for incorporating constraints directly into the Empirical Risk Minimization (ERM) framework. 👉https://t.co/xG1P4L5Vvr 🧵
Excited to share STAPLER, a language model to predict TCR – pMHC reactivity that outperforms prior models.
And for ML aficionados: Description of a new data leakage problem inherent to a common negative data generation strategy.
https://t.co/dsLeYJoT7v
🧵👇