Researchers prove that AI model embeddings can be "translated" from one model to another without seeing the text or modifying the models themselves. Evidence of a universal geometry of meaning. Impressive for AI, unsettling for vector privacy.
ETHICS
-
We Live in a Prompt Not in a Simulation
By
–
We live in a prompt (not in a simulation)🙂 https://t.co/b0SEcSYtYV
— Marek Rosa | European🇪🇺 | South African🇿🇦 (@marek_rosa) 23 mai 2025We live in a prompt (not in a simulation)
-
IRB Ethics Concerns in AI Clinical Trial Process
By
–
That piece was ~50% of the reason that I immediately suspected IRB shenanigans. (I was told they were the models of professionalism in accelerating the process for the trial and could not have been more helpful, by the PI.)
-
Incompetence in Clinical Trial Decision-Making Processes
By
–
And in this exact case we get partial credit because by some string of coincidences and communications pathways someone whose knowledge of clinical trial practice optimization was due to selling bingo cards got tapped to make the call on how many questions to ask an enrollee.
-
Institutions and Inadequate Equilibria in Decision Making
By
–
I haven’t read a better book on this general issue that Inadequate Equilibria. Because *clearly* society would have wanted that trial to succeed (contingent on the biochemistry working etc etc). But society doesn’t make decisions; the institutions it is consisted of do.
-
Academic System Failures Hide Research Trial Underperformance Issues
By
–
It’s not the grad student’s fault, not in any way that matters. Nobody at the university wants to make an enemy of the PI or alienate potential funders in the future. There is no after action on why the trial underperformed because the system does not know, legibly, that it did.
-
Management Failure Blocks Drug Trial Over PHP Developer Shortage
By
–
And nobody in the management chain felt sufficient levels of care about this issue to immediately find someone who could fix. At, again, an American research university, on a funded drug trial, where the blocker is literally Can Anyone In The World Write PHP Please.
-
Trial bureaucracy undermines research efficiency and accountability
By
–
In an early meeting with the team, I asked why the trial intake form had ~200 questions. I assumed this was for Complicated Medical Reasons or that it had been foisted on them by the bioethics board which was in no way accountable for actual success of the trial.
-
Operational competence challenges in clinical drug development trials
By
–
Listening to @rossry ‘s new podcast about drug development and the first episode about operational competence issues in clinical trials is giving me flashbacks.
-
AI Reshapes Humanity: Richard Susskind on Existential Challenges
By
–
Don't miss my latest #podcast with world-renowned AI expert Richard Susskind AI Could Reshape Humanity And We Have No Plan For It In this mind-blowing conversation, I explore why AI represents humanity's defining challenge, requiring us to completely reimagine our