Bakz T. Future
Sharing my initial reactions to GPT-J vs GPT-3 Curie and also on CogView vs. DALL-E.
In this episode, I talk about how:
– GPT-J compares (so far) to GPT-3 Curie
– I’m finding CogView which is a multimodal model that can generate images just with simplified Chinese text
AI Fan Club Discord group:
https://discord.gg/myWKuNQF7V
GPT-J:
https://6b.eleuther.ai/
GPT-3 Curie:
https://openai.com/blog/openai-api/
Cogview:
https://github.com/THUDM/CogView
Subscribe to the Multimodal Podcast!
Spotify – https://open.spotify.com/show/7qrWSE7ZxFXYe8uoH8NIFV
Apple Podcasts – https://podcasts.apple.com/us/podcast/multimodal-by-bakz-t-future/id1564576820
Google Podcasts – https://podcasts.google.com/feed/aHR0cHM6Ly9mZWVkLnBvZGJlYW4uY29tL2Jha3p0ZnV0dXJlL2ZlZWQueG1s
Stitcher – https://www.stitcher.com/show/multimodal-by-bakz-t-future
Other Podcast Apps (RSS Link) – https://feed.podbean.com/bakztfuture/feed.xml
Connect with me:
YouTube – https://www.youtube.com/bakztfuture
Substack Newsletter – https://bakztfuture.substack.com
Twitter – https://www.twitter.com/bakztfuture
Instagram – https://www.instagram.com/bakztfuture
Github – https://www.github.com/bakztfuture
This is interesting
Noice! I was recently was researching this
I recently tried GPT-3 and was very disappointed. I asked it a fairly complex question about how to make a cup of tea and it just gave me gibberish or a googled wiki response. Maybe I didn't use it correctly. Thank you
The intuitive approach is for sure problematic but there is also another part. Prompting. The same prompts might give hugely different outputs on both models, according to their "mind model". So we also would have to "normalize" prompts for both GPT-J and Curie. Finetuned GPT-3 surpasses raw Curie by order of magnitudes, but that's not a fair comparison xD.
GPT-J feels like what I would expect of GPT-2.5. It's good enough at storytelling if you don't pay too much attention to the story itself. It's bad at recognizing humour and dreadful at maths. But it's still a lot more fun than I had with GPT-2.