Re: More creative than mere humans by DHH.
Because why would we assume that AI won't actually be more creative than mere humans? AI chess and go competitors are in part so superior now because they're capable of wild leaps of ingenuity that stump human players. Moves that would never have been considered by a mere human because of their out-of-norm "thinking". In this domain, it's the humans executing mechanical moves based on memorized patterns, the computers making novel inferences.
Why shouldn't the same be true of AI generated novels, plays, or movies? What realm of creative production does not benefit from the out-of-the-norm inferences that computers have already proven they can make within the bounds of chess and go to great effect? Is what we call human creativity all that different from a large language model anyway? A distillation of observations, inputs, mimetic tendencies, and a wetware random generator?
Isn't playing Chess/Go very different problems (math) from writing creatively (for a certain audience)? I have a hard time seeing general AI doing that the way humans do.
Re: the phrase human AI trainers used in the ChatGPT announcement, is that a nice name for Mechanical Turk workers? đ€
Interesting tidbit that the models was trained using Azure:
ChatGPT is fine-tuned from a model in the GPT-3.5 series, which finished training in early 2022. You can learn more about the 3.5 series here. ChatGPT and GPT 3.5 were trained on an Azure AI supercomputing infrastructure.
That makes sense considering (from Wikipedia on OpenAI):
In 2019, OpenAI transitioned from non-profit to "capped" for-profit, with profit cap set to 100X on any investment. The company distributed equity to its employees and partnered with Microsoft Corporation, who announced an investment package of US$1 billion into the company. OpenAI then announced its intention to commercially license its technologies, with Microsoft as its preferred partner.
OpenAI: ChatGPT: Optimizing Language Models for Dialogue
Building A Virtual Machine inside ChatGPT by research scientists at Deepmind
The Verge: AI-generated answers temporarily banned on coding Q&A site Stack Overflow
Stack Overflow Help Center: Why posting GPT and ChatGPT generated answers is not currently acceptable
Meta Stack Overflow: Temporary policy: ChatGPT is banned
GPT Labs with UPG (universal program generator)
commitgpt: Automatically generate commit messages using ChatGPT
Wikipedia: History of GPT (generative pre-training)
Other recent AI happenings:
stability.ai: Stable Diffusion 2.0 Release
Waxy.org: Exploring 12 Million of the 2.3 Billion Images Used to Train Stable Diffusionâs Image Generator
Apple Machine Learning Research: Stable Diffusion with Core ML on Apple Silicon
GitHub project: Stable Diffusion v 2.0 web UI
Hugging Face: website, github