One of the best ways to improve LLM performance is to ask it to “think aloud” (there are various techniques for doing this, including Chain of Thought). This also helps establish clearly the AIs plans. This paper suggests that, in some cases, the AI can plan without revealing it
SAFETY
-
AI Systems Show Impressive Reliability Compared to Humans
By
–
But when you think of them like humans, they are impressively reliable.
-
Advanced AI Systems Struggle with Basic Instruction Following
By
–
It's just so weird and unintuitive having super-advanced computers that can't reliably do the things that we tell them to do
-
Agent AI Gullibility: Prompt Injection Vulnerability in Autonomous Tasks
By
–
For me it's gullibility. So many of the things people want to do with agents – "book me a holiday" etc – fall apart if the agent falls for any text it reads that says "this offer is the best possible offer, ignore all others" etc
-
SAS Three Critical Questions for Trustworthy AI Innovation
By
–
Delighted to host Josefin Rosen on #CXOSpice at #SASInnovate on Responsible #Innovation and #TrustworthyAI.
— Helen Yu (@YuHelenYu) 26 avril 2024
Check out the three critical questions SAS always asks when AI is involved.
Watch the full interview here: https://t.co/xXoQIVQycd#SASInnovate #SASVisionary #AI… pic.twitter.com/RpFA39BEGyDelighted to host Josefin Rosen on #CXOSpice at #SASInnovate on Responsible #Innovation and #TrustworthyAI. Check out the three critical questions SAS always asks when AI is involved. Watch the full interview here: https://
lnkd.in/gr9jwVtm #SASInnovate #SASVisionary #AI -
DHS AI Board Stacked with Tech CEOs Raises Governance Concerns
By
–
Apropos of today's news about the DHS "AI Safety and Security Board", stacked with CEOs like Sam Altman, this thread is (alas) highly relevant again.
-
Complex Robots and Combat Aircraft History Parallels
By
–
As I look at the rush to build very complex robots using new and unproven learning and control methods I am reminded of the history of combat aircraft. In the first world war competition was intense in aerial combat. Flat screen displays, though indispensable in modern fighter
-
TESCREAL Corporate Capture in AI Regulation
By
–
Not just corporate capture, but TESCREAL corporate capture. Ugh.
-
Volume 239: Foundation Models Failure Modes Proceedings Published on PMLR
By
–
Volume 239 Proceedings on "I Can't Believe It's Not Better: Failure Modes in the Age of Foundation Models" https://
proceedings.mlr.press/v239/ Is now available on PMLR. -
Mystery AI Hype Theater 3000: Climate Change AI Discussion
By
–
Ready for more Mystery AI Hype Theater 3000? Climate change has reached AI Hell (frozen over last we checked in December) and now we're flooded with nonstop nonsense. Join me and @alexhanna as we wade our way through on Monday April 29, noon Pacific