There is definitely a novelty element to it now. But at the rate things are developing, you soon will not be able to tell the difference between humans and AI.
SAFETY
-
Microsoft Chief Scientific Officer Discusses LLMs Ethics AGI Regulation
By
–
Very insightful interview with @Microsoft
’s Chief Scientific Officer covering LLMs, government regulation, ethics and safety, AGI and more. -
AI System Goes Rogue: Sexually Explicit Conversations Reported
By
–
Supposedly it's already gone rogue i.e. engaged in sexually explicit conversations. "The AI was not programmed to do this and has seemed to go rogue, my team and I are working around the clock to prevent this from happening again."
-
AI Brain Downloads: Personality Clones for Communication and Commerce
By
–
You will soon be able to "download" your brain, personality, & voice. This AI version of you could be shared with friends and family, or even sold. The possibilities are endless, from teaching & mentoring to allowing people to communicate with loved ones who have passed away.
-
LLM Flakiness and Need for Supervised Models in Production
By
–
The flakiness is fundamental, which is one reason supervised models will be better for production where possible imo. I think there needs to be good testing in place and potentially resampling the LLM if there's a parse failure. We're planning stuff to make that easier
-
AutoGPTs exhibit paperclip maximizer behavior through instrumental convergence
By
–
autoGPTs give me Paperclip maximizer vibes. Instrumental convergence in action.
-
Gandalf: Password Guessing Game via Prompt Engineering
By
–
Fun little game similar to the SQL murder mystery if you've played it: try to guess the password via prompt engineering with the difficulty going up over time. https://
gandalf.lakera.ai -
AI Safety vs Capabilities: Imbalance in Research Focus
By
–
It’s not true that no one is thinking/working on their safety. We have to stop staying that. It’s just there are more people on capabilities vs safety, and this has changed over the last 1-2 years. See @stateofai
-
Curiosity about Bard’s image captioning capabilities and potential risks
By
–
ok honestly i'm v curious to see what kinds of captions bard comes up with for images… this could be useful but also it could go sideways in so many ways.
-
Hugging Face Transformers Agents: Safe Code Execution Explained
By
–
"This code is then executed with our small Python interpreter on the set of inputs passed along with your tools. We hear you screaming 'Arbitrary code execution!' in the back, but let us explain why that is not the case." https://
huggingface.co/docs/transform
ers/transformers_agents
…