Signed in as:
filler@godaddy.com
Signed in as:
filler@godaddy.com
A morally questionable AI that will insult you, make jokes at your expense, and break every week or two.
DuckyAI is based on an open source LLM(Large Language Model), Llama 2. Most LLMs are a bit boring, though. They aren't fun to talk to at all, and give you very coperate lifeless responses. So, heres what I did to fix that.
DuckyAI is essentially a fine-tuned custom version of Meta's Llama 2 model. Llama 2 is already great at generating text and mimicking human responses. Instead of improving that, I wanted DuckyAI to be funny and sarcastic. To do this, I used an open source web scraper to download large amounts of data from comedy shows, plays, and other sources that I believed would help achive the personality I was looking for. On top of that, I also scraped data on current events and other information to expand on Llama 2's knowladgebase. DuckyAI was fine tuned using Replicate's cloud LLM fine tuning stack.
During the fine-tuning process, I wanted to make sure DuckyAI wouldn't hallucinate(go insane and give false information), and that DuckyAI was fun to chat with. I'm pretty happy what I ended up with; an AI that insults you, and can provide accurate information(most of the time).
Visit https://chat.duck.enterprises to try DuckyAI
Of course not. OpenAI is a giant company, I'm just a kid with too much free time. In my opinion, though, DuckyAI is a lot more fun.
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, you consent to the collection of your data.