I find it interesting that people actually want to build on something that might already be obsolete in 6-12 months when the next better, bigger model comes along and smashes GPT-3. As all the effort you probably have put into prompt engineering you can throw straight in the bin as every new model will behave completely different.
https://www.copy.ai/ seems interesting.
Could any copywriters chip in on how good it is in comparison to actual copywriters?
I find it interesting that people actually want to build on something that might already be obsolete in 6-12 months when the next better, bigger model comes along and smashes GPT-3. As all the effort you probably have put into prompt engineering you can throw straight in the bin as every new model will behave completely different.
Is it possible to use GPT-3 on Azure? I tried looking the other day but couldn't find it :(
I'm not sure the model itself is publicly available.
And even so, the model has 175 billion parameters; it's too big to fit on a GPU (the largest GPU in the world has 48 GB of VRAM)
For Azure, it will probably cost from $1500 to $3000 per month to host it on GPU VM, like 4xK80
As far as I understand all of the GPT-3 projects out there are using OpenAI API (which is hosted on Azure btw)
Ah I see, thanks for this. I've now signed up to the OpenAI API waitlist :)