On Mon, Feb 20, 2023 at 12:33 PM Jimmy Wales <jimmywa...@wikitribune.com> wrote: > Speaking only for myself, out of curiosity, some real world examples might be > helpful here. I don't have access to Bing's > version yet, but I do have access to chat.openai.com which is very impressive > but deeply flawed.
I've found ChatGPT most useful for small coding tasks (with a lot of scrutiny). Most of the other practical applications I've heard of have been of the creative variety, or in writing mundane letters, emails, proposals, summaries, etc. As an example, please find a ChatGPT-generated summary of this email at the end. I think it's best to view ChatGPT (and its like) at this stage at, at its best, a useful assistive technology and, at its worst, a distributed denial of service attack on our collective ability to understand our world. The attempts to quickly commercially exploit these technologies tend to push their impact more towards the latter, at least until those deep flaws you mention are addressed. It's a technology that requires a high degree of literacy in its responsible use, while suggesting to the user that it requires none: a dangerous combination. The grand vision is to create human-level artificial intelligence. "AGI" (Artificial General Intelligence) is now an explicit stated goal of major players in the field. Of course, if AGI is in fact realized, it _will_ change everything: a dream as big as SETI or limitless energy generation. But for now we just have sparkling autocomplete. It's easy to enumerate potential positive applications (assisted editing, Wikidata query generation via natural language, automatic summaries of open access citations, ...). For any one of them, I think the challenge is to figure out a way towards _responsible_ integrations that don't proliferate misinformation and add value. I do think that it is strategically vital for Wikimedia to understand and explore this space, to look for low-risk/high-reward applications, and to be dispassionate and objective in the face of both AI hype and anti-AI backlash. Erik --- ChatGPT summary of this email: The email discusses the practical applications of ChatGPT and warns about the negative consequences of quickly commercializing AI technology. The writer suggests responsible integration of AI to avoid misinformation and add value, and recommends that Wikimedia explore low-risk/high-reward AI applications while remaining objective in the face of AI hype and backlash. _______________________________________________ Wikimedia-l mailing list -- firstname.lastname@example.org, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l Public archives at https://email@example.com/message/SDRJGERONJY7FCQYRJ3ZVX74OIUX4LPJ/ To unsubscribe send an email to wikimedia-l-le...@lists.wikimedia.org