Alt text for video: Audiogram with Mystery AI Hype Theater 3000 logo, title "Episode 31: Science Is A Human Endeavor" and subhead "Even if the hype of 'automated discovery' by 'AI scientists' were valid, would we really benefit from scaling up scientific research?" >>
ETHICS
-
Artificial Scientists: Why the AI Dream Is Impossible
By
–
MAIHT3k Ep 31: Science is Human Endeavor in which Drs. @mollycrockett and @lmesseri join me & @alexhanna to explore why the dream of artificial scientists is not only impossible but actually a nightmare.https://t.co/OK58jhdQDK
— @emilymbender.bsky.social (@emilymbender) 8 mai 2024
Thx to @ctaylsaurus for production! pic.twitter.com/r6C4cZmseIMAIHT3k Ep 31: Science is Human Endeavor in which Drs. @mollycrockett and @lmesseri join me & @alexhanna to explore why the dream of artificial scientists is not only impossible but actually a nightmare. https://
buzzsprout.com/2126417/150200
29-episode-31-science-is-a-human-endeavor-feat-molly-crockett-and-lisa-messeri-april-15-2024
… Thx to @ctaylsaurus for production! -
Sam Altman’s Thoughts May Contain Errors Verification Warning
By
–
“Sam Altman’s thoughts and creations may contain errors. Please check information is correct.”
-
Empêcher les LLMs d’apprendre menace les éditeurs
By
–
I am not sure if preventing LLMs to learn on your content is something you want long term as a publisher. In a scenario where LLMs would take over most of the search traffic, you would just make yourself irrelevant with such move
-
Three Laws of AI: Competence Overestimation and Success Factors
By
–
Back on April 1st I posted my three laws of robotics. Here are my three laws of AI. 1. When an AI system performs a task, human observers immediately estimate its general competence in areas that seem related. Usually that estimate is wildly overinflated. 2. Most successful AI
-
Deterministic Quoting: Preventing LLM Text Alterations
By
–
I saw an interesting proposal today for avoiding those kinds of minor alterations to direct quotes: https://
mattyyeung.github.io/deterministic-
quoting
… "That’s the only way to guarantee that an LLM has not transformed text: don’t send it through the LLM in the first place." -
Vectara Platform Addresses GenAI Hallucinations and Copyright Issues
By
–
Great visit to @Vectara for discussion with @awadallah and team. They have built a great platform for GenAI for developers that enables mitigation of all the issues you see in common models (Mitigating Hallucinations and Copyright Concerns, Minimizing Bias, Enhancing
-
Company credibility threatened by regulatory engagement strategy
By
–
Feels like that could undermine their cozying up to lawmakers and safety community a bit
-
Need for Closed LLM Evaluations and Unknown Benchmarks
By
–
Totally agree. I was just talking recently about the need for closed evaluations and unknown benchmarks. Some kind of LLM comparison tests that can't so easily be gamed.
-
Content and Data Strategy Approach in AI Era
By
–
Our approach to content and data in the age of AI: