🙏Anke! On theft, I see a distinction between AI for creative content(art, stories etc - a hard no) and AI for software. In programming there is a strong tradition of open access & sharing - using AI to parse the back catalogue seems a natural evolution to the process?
Such sharing requires consent on the part of the original authors, and frequently comes with some form of obligation to share forward within very specific constraints (FOSS/GNU, BSD etc). I'm not seeing much in the way of consent or concern with licensing? Quite the opposite.
Also 🙏for the Forbes link. What I can't immediately assess is whether the efficiencies of running a faster less energy-intensive AI-improved piece of software outweighs the energy used for the AI query in the first place 🤔. This definitely needs more thought though so thanks for raising it!
Hello! AI has to be trained on something. How that something - the dataset(s) are obtained is a problem. Much is stolen, ie scraped from the internet. Other orgs are using proprietary datasets, are open/transparent/sharing and that’s good. But they are in the minority & this issue has to be resolved