Wow, this has just become my favorite LLM test. I missed that this doesn't work but it really doesn't, even for SOTA LLMs. Seems to be a bit hit and miss, e.g. with GPT4o which failed 1/3 times, Claude failed 3/3 times.
ETHICS
-
Basic Income Study Shows Impact on Financial Stability
By
–
fascinating story from @shiringhaffary and @sarahsholder about a big @sama
-funded US basic income study that gave people $1k/month for 3 years. the results: people spent the $ on basic needs and helping others, they spent more time with family, and had more financial flexibility. -
Musk’s Influence: Questioning Tech Project Promotion
By
–
lots of people cant get past the headline on this one (fair!) the piece is really about the influence Musk purchased two years ago and how everyone who posts here has to ask themselves some uncomfortable questions about whose project they're furthering
-
Adversarial Training Benchmarking: Average Case vs Attacker Strategy
By
–
As one of the co-inventors of adversarial training I endorse Carlini’s take. The issue is not so much which attack transformations are allowed, the issue is using average case benchmarks when an attacker will not randomly sample from average starting points
-
LLMs Demonstrate Capabilities Beyond Stochastic Parrot Theory
By
–
Oops, looks like LLMs aren’t stochastic parrots after all.
-
Political Divide: Millions Caught Between Ideological Extremes
By
–
then there are the millions of us stuck in the middle politically homeless
-
French Olympics tech barriers spark debate on technocratic governance
By
–
Vous savez comme je suis optimiste et enthousiasmé par les JO. Mais franchement, ces QR codes et barrières qui empêchent les gens de rentrer chez eux, d’aller bosser, et qui tuent les restaurateurs, c’est insupportable ! Le monstre technocratique français est en roue libre.
-
CrowdStrike Poses Greater Risk Than Fictional Terminator Threat
By
–
Fake danger: Terminator
Real danger: CrowdStrike -
Healthcare AI Systems Creating Problems for Patients and Staff
By
–
This is nuts. I’m so sorry they have to deal with this, and patients have to deal with this, and you, too, have to deal with this. Thank you for helping them.
-
Software Issues: System Prompts and Adversarial Examples
By
–
Of course, it's software.
Easy mode: a bad system prompt update.
Hard mode: an adversarial example in the context.