Wow, that's some license addendum:
> This software should not be used to promote or profit from:
> violence, hate, and division,
> environmental destruction,
> abuse of human rights, or
> the destruction of people's physical and mental health.
Anyone have a real-world use case for something like this? I must admit I'm having trouble thinking of any that aren't essentially deceptive. Because in my little biased world, I have no need of "text" per se, and what value any text has to me is closely linked to the fact that it came from a human.
From the blog post: "Beyond the technical work to develop this model, we’ve also taken several steps to anticipate and mitigate malicious use cases where possible."
From the preprint, this seems to be doing some review before release and having a code of conduct in the GitHub repo.
The unicorn prompt is the new text generator lorem ipsum
It was trained on 140GB of text on 256 TPUs for 2 weeks, the model being made of 48 transformer layers. I'm wondering when we will see a model trained on 1TB or 10TB of text.
Could someone provide a high level summary of what this is for a technical person not conversant with the field?
> Advertisement
Yeap, This one is indistinguishable from reality
Are there any hardware reqs to work with this?
Open AI did the right thing by not releasing their model; it's disappointing that researchers are so callous about the potential effects of their research in the name of progress.
I am working on a guide (should be released tomorrow) to easily get it up and running for personal use. Here's my Twitter thread of current experiments with the model: https://twitter.com/minimaxir/status/1173081315177975810
I recommend reading the linked paper in the repo as it gives decent examples/instructions on how to use the model. Although the size and architecture is comparable to GPT-2, the emphasis on conditional generation differentiates it.