Our previous interpretability research found that we could artificially dial up or down certain “features” within a model to modify its behavior. Here, we ran larger-scale tests of feature steering, focusing on measuring social biases.
ETHICS
-
Anthropic Studies Feature Steering in AI Systems
By
–
New Anthropic research: Evaluating feature steering. In May, we released Golden Gate Claude: an AI fixated on the Golden Gate Bridge due to our use of “feature steering”. We've now done a deeper study on the effects of feature steering. Read the post: http://
anthropic.com/research/evalu
ating-feature-steering
… -
Math Papers Face Extreme Scarcity of Qualified Reviewers
By
–
ML: "We don't have enough qualified reviewers"
Math: "Hold my beer" For some math papers, the number of reviewers who can actually understand and engage with the paper is in the single digits. -
Stem cell transplants show no increased cancer mutation risk decades later
By
–
Reassuring: decades after stem cell transplants, no indication of increase in cancer related gene mutations https://
science.org/doi/10.1126/sc
itranslmed.ado5108
… @ScienceTM https://
nature.com/articles/d4158
6-024-03450-x
… @Nature -
Claude Privacy Policy: Legal Restructuring Over Substantive Change
By
–
Yes definitely better for privacy – though for fairness in the case of Claude it's more moving legal pieces around than a change of policy apparently
-
Nadella’s Copyright Remarks and AI Dehumanization Ethics
By
–
A quick newsletter post on the dehumanization behind Satya Nadella's remarks about copyright law
-
OpenAI’s Reasoning Claims Ridiculed on MAIHT3k Stream
By
–
On the next MAIHT3k live stream, @alexhanna & I will have a good laugh at OpenAI’s recent claims that they’ve trained their synthetic text extruding machines to “reason”, actually. Join us for the ridicule as praxis on Monday Oct 28, noon Pacific:
-
AI Won’t Destroy World: Illiteracy and Misuse Are Real Risks
By
–
AI Will Not Destroy The World—AI Illiteracy And Misuse Could
#AI #AIio #BigData #ML #NLU #Futureofwork @TunstallAsc @StrategyFintech @TamaraMcCleary @TerenceLeungSF @psb_dc @thomaspower @vinod1975 @ylecun http://
ow.ly/ZrO230sH90J -
Local Court Ruling on AI Web Scraping Raises Procedural Concerns
By
–
Fascinating that they consider a local court (non-federal) which does not set precedents to be "paving the way" for legitimizing web-scale scraping for AI—all while there were significant problems w/ the Court's procedure itself. Such PR doesn't help the verdict look above board. https://
t.co/bhpbR4Npui -
AI Anthropomorphization Risks: Liability and Child Safety Concerns
By
–
Yes I keep saying anthropomorphization of AI is a bad idea but it feels so often like shooting in the void « negligence & liability (plus wrongful death, etc) due to intentional anthropomorphization, which was targeted specifically to minors who have a harder time