Gregory's Server|
9 comments
Computer programming has already been massively shaken up by GitHub Copilot, which is based on the same tech as ChatGPT. There are copyright considerations to work out, but it's already helping developers be more productive. One example that we both shared is our mutual loathing of the syntax of the Unix shell, Bash. It's obscure and peculiar, and almost no one uses it frequently. With AI in your toolkit, programming is about thinking rather than syntax. This is how it should be. Beyond improving the lives of people who already know how to program, AI can also help people program who know nothing about coding. Census GPT is a fantastic example of this. It's a site that writes SQL queries of U.S. Census data from regular text. There will be many more sites like this: https://censusgpt.com The next episode of this mini-series on #AI and the rest of us is a lengthy discussion with With his colleagues, Stephen did an important survey gauging public opinion on AI and also talking about some of the potential issues. Here's the link to the survey if you want the full details beyond this very brief summary: https://changeresearch.com/wp-content/uploads/2023/03/National-AI-Memo-March-2023-FINAL.pdf People are very concerned about their own ability to discern AI-generated content versus human-generated content. And with good reason! ChatGPT passed the famous Turing Test in the survey. People couldn't tell the difference between Stephen's words and ChatGPT's. Beyond their value to foreign propaganda campaigns, bots have long played a role in U.S. politics. So in some sense, people may be overestimating the impact that AI-driven bots could have. Republicans have long used them to fake support. I saw this in my former career as a right-wing media strategist. The bigger possible effect of AI is likely to be in the realm of "deep fakes," simulated video that often looks very, very real. They've been around for a while but have not yet had a political impact. This is certain to change in the near future, especially since making deep fakes is now a form of entertainment as the technology has improved. https://www.fastcompany.com/90876748/ai-celebrity-deepfakes-are-over-hoaxes-as-entertainment-are-just-beginning Be sure to subscribe to Theory of Change to get more episodes sent to you each weekend. And please boost this thread and let your friends know. It's very helpful! /end YouTube: https://youtube.com/channel/UCkmucd07dnIOY9Gf2HZ5Y5w Apple Podcasts: https://podcasts.apple.com/us/podcast/theory-of-change-podcast/id1486920059 |
Big business & govt are already using AI for their purposes, and the rest of us can figure out how we can as well.
We're at a very important moment for early adopters of large language model technologies. Most people and organizations haven't figured out what AI can do for them. They've got enormous flaws.
But nonetheless, they are very, very useful to those who know their capabilities and limitations.