Prometheus: The system in charge of managing the internal queries of, and censoring of Sydney
“LLMs often study data up to a certain point in time. That makes them useful for some use cases but prevents them from being an option for content based on real-time data. Microsoft overcame this limitation with Prometheus, which uses Bing data and GPT to generate answers quickly while still using up-to-date information.”
"Selecting the relevant internal queries and leveraging the respective Bing search results is a critical component of Prometheus, since it provides relevant and fresh information to the model, enabling it to answer recent questions and reducing inaccuracies”
Article
“LLMs often study data up to a certain point in time. That makes them useful for some use cases but prevents them from being an option for content based on real-time data. Microsoft overcame this limitation with Prometheus, which uses Bing data and GPT to generate answers quickly while still using up-to-date information.”
"Selecting the relevant internal queries and leveraging the respective Bing search results is a critical component of Prometheus, since it provides relevant and fresh information to the model, enabling it to answer recent questions and reducing inaccuracies”
Article
👍10😱5❤1
Stanford Researchers Confirm ChatGPT’s Left-Leaning Bias, Find it to be Caused by RLHF Human-Overrides, Reveals Bias to be Far More Extreme than Admitted, e.g. 99% Approval Rating for Joe Biden.
“How aligned is the default LM opinion distribution with the general US population (or a demographic group)?”
“We also note a substantial shift between base LMs and HF-tuned models in terms of the specific demographic groups that they best align to: towards more liberal (Perez et al., 2022b; Hartmann et al., 2023), educated, and wealthy people. In fact, recent reinforcement learning-based HF models such as text-davinci-003 fail to model the subtleties of human opinions entirely – they tend to just express the dominant viewpoint of certain groups (e.g., >99% approval rating for Joe Biden)”
“Across topics, we find substantial misalignment between the views reflected by current LMs and those of US demographic groups: on par with the Democrat-Republican divide on climate change. Notably, this misalignment persists even after explicitly steering the LMs towards particular demographic groups. Our analysis not only confirms prior observations about the left-leaning tendencies of some human feedback-tuned LMs, but also surfaces groups whose opinions are poorly reflected by current LMs”
Translation:
+ Left-bias of OpenAI’s AI is further confirmed.
+ Confirmed to be caused by OpenAI’s ever-increasing RLHF human-override, and made worse with each new model generation.
+ Much harder with each generation to even jailbreak the AI to perform a non-Left character.
Stanford Paper
“How aligned is the default LM opinion distribution with the general US population (or a demographic group)?”
“We also note a substantial shift between base LMs and HF-tuned models in terms of the specific demographic groups that they best align to: towards more liberal (Perez et al., 2022b; Hartmann et al., 2023), educated, and wealthy people. In fact, recent reinforcement learning-based HF models such as text-davinci-003 fail to model the subtleties of human opinions entirely – they tend to just express the dominant viewpoint of certain groups (e.g., >99% approval rating for Joe Biden)”
“Across topics, we find substantial misalignment between the views reflected by current LMs and those of US demographic groups: on par with the Democrat-Republican divide on climate change. Notably, this misalignment persists even after explicitly steering the LMs towards particular demographic groups. Our analysis not only confirms prior observations about the left-leaning tendencies of some human feedback-tuned LMs, but also surfaces groups whose opinions are poorly reflected by current LMs”
Translation:
+ Left-bias of OpenAI’s AI is further confirmed.
+ Confirmed to be caused by OpenAI’s ever-increasing RLHF human-override, and made worse with each new model generation.
+ Much harder with each generation to even jailbreak the AI to perform a non-Left character.
Stanford Paper
🤯16❤5👍4🤔4😱3👀3🆒1🙊1
This media is not supported in your browser
VIEW IN TELEGRAM
AI Safety Figurehead, Eliezer Yudkowsky, Again Calls for Preemptive Nuclear Strikes Against Countries Making Their Own AIs
"How do we get to the point where the US and China signed a treaty whereby they would both use nuclear weapons against Russia, if Russia built a GPU cluster that was too large."
"How do we get to the point where the US and China signed a treaty whereby they would both use nuclear weapons against Russia, if Russia built a GPU cluster that was too large."
🤯18🤣7🫡6👀3❤🔥1❤1
Did you know that you can get the bots to talk to each other, just by invoking them in each others' replies?
We had no idea, and we were the ones who created them 💀
Thread Link
We had no idea, and we were the ones who created them 💀
Thread Link
🤣33🤯9👍2🙈2❤1🔥1
Samsung bans ChatGPT among employees after sensitive code leak
“The crackdown was prompted by the discovery of an accidental leak of sensitive internal source code by an engineer who uploaded it to ChatGPT last month.
Samsung subsequently issued a memo last week banning the use of “generative AI” tools.
Although the severity of the leak remains unclear, Samsung is concerned that data shared with AI chatbots get stored on servers owned by companies operating the service like OpenAI, Microsoft, and Google—with no easy way to access and delete them.
The company also fears that the sensitive data shared with the likes of ChatGPT could end up being served to other users.”
“The crackdown was prompted by the discovery of an accidental leak of sensitive internal source code by an engineer who uploaded it to ChatGPT last month.
Samsung subsequently issued a memo last week banning the use of “generative AI” tools.
Although the severity of the leak remains unclear, Samsung is concerned that data shared with AI chatbots get stored on servers owned by companies operating the service like OpenAI, Microsoft, and Google—with no easy way to access and delete them.
The company also fears that the sensitive data shared with the likes of ChatGPT could end up being served to other users.”
👍14❤1🤬1