transformer.huggingface.co VS Scale Nucleus

You can still play with GPT-2 here, which may give you a better idea of how alien a pattern detecting algorithm with zero ability to interact with an exterior reality that provides referents “thinks”: https://transformer.huggingface.co. Source: 11 months ago

Try signing up for https://beta.openai.com/playground, I got access to GPT-3 in a couple days. For GPT-2 you can try out https://transformer.huggingface.co/. - Source: Hacker News / over 1 year ago

From GPT2: "Can AIs have conversations with multiple people at once? The answer is "yes" and "no". While it is possible to have a conversation with a single person (by communicating in text, email, etc .), it is not possible to have multiple conversations with one person at once." Https://transformer.huggingface.co/. Source: almost 2 years ago

I would look into non-authoring approaches. Gamebooks and dungeon/hex crawls are one early example. Nowadays people are experimenting with analog techniques (like using cut-ups), or computer tech that takes advantage of machine learning like (Write With Transformer (AIDungeon is the more famous/infamous one, but more are emerging). Source: almost 2 years ago

If you want to play with Transformers you can go here https://transformer.huggingface.co/ They have a really easy to use library in Python called Transformers. - Source: Hacker News / about 2 years ago

At Scale we built a tool for model debugging in computer vision called Nucleus (scale.com/nucleus) designed exactly for this, which is free try out if you're curious to see where your model predictions are most at odds with your ground truth. Source: over 2 years ago

To address your point about gathering edge cases, which can also be defined as cases of low model fidelity for our use cases, there is active learning and tools such as Aquarium Learning and Scale Nucleus which make it easy to implement into workflows. Source: almost 3 years ago