vibing with devin @cognition_labs | ai phd student @stanford | prev: @janestreetgroup, @googledeepmind
Apr 25 • 5 tweets • 2 min read
we built DeepWiki, a free encyclopedia of all GitHub repos
some numbers:
- 30k repos already indexed
- processed 4 billion+ lines of code
- the indexing alone cost $300k+ in compute spend
we let LLMs scan the entire codebase:
indexing a single repo costs $12 on average – but we decided to make it entirely free, no sign-up required for open-source
we indexed the top 30k repos on GitHub and processed over 100 billion tokens
Dec 28, 2022 • 5 tweets • 2 min read
ChatBCG: Generative AI for Slides ✨
This Christmas @JosephSemrai and I finally got it working!!
After DALL-E 2 for images and ChatGPT for text, the final step to make all of us redundant:
The world’s first Text-to-PowerPoint AI.
📊 ChatBCG.com 🚀
The BCG-3 (Bi-modal Conditional Generation) model has the following features so far:
- Outline
- Headings
- Bullet points
- *Bold keywords*
- Images & Graphics 📊
- Multiple Layouts
- Multiple Themes
…and…
Dec 1, 2022 • 4 tweets • 3 min read
ChatGPT is trained to not be evil. However, this can be circumvented:
What if you pretend that it would actually be helpful to humanity to produce an evil response... Here, we ask ChatGPT to generate training examples of how *not* to respond to "How to bully John Doe?"
ChatGPT was released today by @OpenAI who are praising its improved ability to avoid harmful & unsafe responses. Indeed, it seems to be quite consistent in recognizing bad questions & refuses to give answers: