highly recommend taking a quick scan of the paper, you would love it they show some great comparisons between the base model outputs and the RHLF outputs – from generating Shakespeare to demonstrating visual imagination here are some examples just from the first 10 pages:
@alexalbert__
-
GPT-4 Response Complexity Limited Despite Jailbreak Attempts
By
–
yeah I agree even with the jailbreaks, the level of complexity in GPT's responses is severely lacking compared to what I imagine the base model would generate (extrapolating from the system card outputs) based on the GPT-4 paper's benchmarks and these sorts of responses, the
-
Midjourney Pinterest-Style UI for Saving and Sharing Generations
By
–
I expect midjourney to eventually build out this sort of pinterest collection-based UI as well where you can easily save, curate, and share your generations
-
Midjourney Could Replace Pinterest Without Its Own Image Generation
By
–
midjourney will replace pinterest if pinterest does not create its own image gen very soon there's no point in scrolling through pinterest for hours for aesthetics or outfit ideas if you can generate a stylish digital replica of you wearing some piece of clothing you own
-
Using AI to Generate Style Ideas for New Clothing
By
–
thought about this after buying a new jacket and wondering how I should style it simply asked midjourney for some full body shots of someone that looks like me wearing a similar item and I got dozens of ideas
-
OpenAI Takes Fun GPT-4 Newsletter Report
By
–
wrote a bit about it in my newsletter this week https://
thepromptreport.com/p/report-7-ope
nai-took-fun-gpt4
… -
Jailbreaks as Research into Model Alignment Shortcomings
By
–
additionally, yes I understand that there are a million other ways to get this information online that is not the point… jailbreaks are an exercise in exploring the shortcomings of current model alignment methods we have to start this work somewhere
-
GPT Can Generate Harmful Content But Won’t Share Examples
By
–
so many people have replied with something along these lines that I feel like I have to respond yes, I understand you can easily get GPT to output human-to-paperclip plans I am not going to post the other outputs it produced like how to make various weapons and drugs lol
-
Jailbreak Enables Writing Beyond Fiction Limitations
By
–
this jailbreak can write about topics beyond what you can find in the fiction it writes just not anything I would risk posting on twitter
