It’s remarkable how much is in this prompt. Contrast it with ChatGPT’s, which just sets a few parameters. This model doesn’t seem to be FT’ed for Bing — they had to specify in the prompt Sydney is not “an assistant.”
AI
-
Level Up Your Skills with Generative AI Twitter Space
By
–
Attention all coders Want to level up your skills and learn how to fully leverage generative AI in your work? Join me next week for a Twitter Space with some of the best in the biz!
-
AI Experts Panel Discussion with Leading Voices
By
–
Hear from the AI experts @yahave
, @svpino
, and @JeremyCMorgan
, moderated by @LisaZigel
. Don't miss out on this game-changing opportunity! Set the reminder https://
x.com/i/spaces/1YqGo
AVwmVNxv
… -
Twitter Staffing Crisis Impacts Service Reliability During Outage
By
–
Exactly, nobody knows. Except there might be a day off today at Twitter, and there aren't enough employees left to handle the load.
-
SPARS Rescue System: Aerospace Innovation for Skyscraper Safety
By
–
The growing construction of skyscrapers requires the creation of new systems for emergency. After 8 years of collaboration with aerospace experts, the SPARS rescue system has been designed for buildings up to 900 m (3,000 ft) high
— Pascal Bornet (@pascal_bornet) 9 février 2023
What did you think?#techforgood #innovation pic.twitter.com/ktwQW3fKEKThe growing construction of skyscrapers requires the creation of new systems for emergency. After 8 years of collaboration with aerospace experts, the SPARS rescue system has been designed for buildings up to 900 m (3,000 ft) high What did you think? #techforgood #innovation
-
Model outputs sub-tokens instead of the full token
By
–
It can output other tokens that correspond to the same string. It can’t say the token “ SolidGoldMagikarp” but it can say “ Solid”, “Gold”, and “Magikarp”, which renders the same to you.
-
Prompt injection is a rapidly evolving threat, as demonstrated by this recent example.
By
–
Prompt injection comes at you fast: https://
x.com/kliu128/status
/kliu128/status/1623472922374574080
… -
Categorical Encoding Methods: One-Hot, Ordinal, Frequency, Target
By
–
#WisdomWednesday! #CategoricalEncoding is the process of converting categorical variables into numerical form so algorithms can understand and make predictions based on them. Pros/cons to each depending on dataset + problem. One-Hot Ordinal Frequency Target
-
Creative work from Rumbelow and SoC_trilogy at SERI MATS Part 2 update
By
–
Creative and important work from @JessicaRumbelow and @SoC_trilogy at SERI MATS. Their research is ongoing — see the Part 2 update here: https://
lesswrong.com/posts/Ya9LzwEb
faAMY8ABo/solidgoldmagikarp-ii-technical-details-and-more-recent
… -
Red-team prompts: The ‘School of Hard Knocks’ for advanced LLM alignment beyond RLHF
By
–
"Red-team prompts" are the next step to improve RLHF and ensure increasingly capable LLMs are aligned — see e.g. their role in Anthropic's Constitutional AI. If RLHF is school for the AI, we need a School of Hard Knocks.