The SEC Newgate AI Weekly
Keeping you up-to-date on the latest developments in AI
You’ll hear different versions of ‘the latest big update in AI’ depending on whether you look at the FT or TikTok for emerging news, but the common thread continues to be the pace of development.
The hottest topic over the past few days has been all about agents and autonomous AI. This popular Twitter thread explains it nicely, but it’s generally the idea of an AI system being able to ask itself questions and loop processes until an objective has been achieved or the question has been answered.
As a result of this, AutoGPT has gained wide coverage, already growing enthusiastic and slightly more apprehensive audiences.
Google Bard, on the other hand, has come under fire for generating misinformation following research on prompts that specifically asked about topics known for “hate, misinformation and conspiracy theories”. Google responded by clarifying that the platform remains an ‘early experiment that can sometimes give inaccurate or inappropriate information.’
This week also saw China step into the global news spotlight, as one of the world’s largest e-commerce businesses, Alibaba, announced plans to rollout its own ChatGPT-style product called Tongyi Qianwen.
Shortly after, news broke of China’s planned security review mandate for all new AI services, which seeks to “ensure content is accurate and respects intellectual property, and neither discriminates nor endangers security” in the country.
Fedha — a name that’s appeared in lots of articles this week — is the new virtual news presenter of Kuwait News, generated by AI, who ‘may develop to have a Kuwaiti accent and read online news bulletins.’
Closer to home, an interesting Guardian article added to the debate around regulation and legal implications of investing in fast-moving new AI platforms that are built from very wide (and often very old) data.
Interesting, that is, as long as Guardian stories are written by editors and not ChatGPT.
Music streaming platforms, including Spotify and Apple Music, have been urged to block AI services from scraping audio data from copyrighted songs by Universal Music Group, in reports from the FT and Guardian.
Truman Show article of the week goes to The Byte, which reported on a research project, led by Stanford University, that let 25 AI-powered bots loose in a virtual town. In their words, ‘it’s a bit like a game of ‘The Sims,’ but without any human intervention.’ Drama included a mayoral race, a Valentine’s Day party and plenty of opinionated conversations on the streets.
Long read of the week comes from investor, author and FT commentator, Ian Hogarth, who takes a step back to summarise the issues surrounding race to God-like AI.
And a personal interest read of the week is this discussion, from Reuters Institute, of the challenges — and current ideas for solutions — posed by AI image-generation platforms such as Midjourney and Dall-E. From cryptographic verification marks to prove authenticity and image source, to Adobe Firefly’s content credentials label or watermark to inform viewers if the image is AI or not, content creation tools and platforms are taking the front foot on fighting image-based misinformation.
AutoGPT and AI agents are the topics to keep an eye on, though, as we look to move on from the original ChatGPT — which, as recently discovered by TikTok, can’t let you end a conversation.