A week in Generative AI: Claude 3, Apple WWDC, & The AI Mirror Test
News for the week ending 31st March 2024
A slower AI news-cycle this week with no major new announcements but Claude 3 now tops the Chatbot Arena Leaderboard, Apple announced an AI focused WWDC, OpenAI released some new Sora videos made with professional filmmakers and artists and an interesting AI mirror test was shared on Twitter.
Claude 3 surpasses GPT-4 on Chatbot Arena for the first time
This is great to see - the first time that another model has bested GPT-4 in the Chatbot Arena since the arena was launched nearly a year ago in May 2023. The Chatbot Arena is the best comparison of chatbots as it pits them against each other head-to-head and gives them an Elo score, like you get in competitive chess.
Claude 3 Opus only beats the latest GPT-4 model by a couple of points, so this reinforces my view that Claude 3 is a “GPT-4 class” model and doesn’t represent the next generation of models we’re likely to see towards the end of this year. It’s rumoured that OpenAI will release a GPT-4.5 this summer, so it will be interesting to see if that retakes the crown!
Apple WWDC 2024 promises to be (A)bsolutely (I)ncredible
On Monday, Apple announced their Worldwide Developers Conference (WWDC) right on time. Greg Joswiak, Apple’s SVP of Marketing, teased on Twitter that the event would be “absolutely incredible” in a not so subtle hint at what the focus of the event will be.
I’m excited for this - Apple are rumoured to announce the most radical overhaul of iOS since it was launched, which coupled with a big focus on AI could mean we’re going to see the first mobile operating system truly built for the AI-era. I don’t think Apple will go that far, but hopefully what we’ll see is a significant stepping stone in that direction.
Sora: first impressions
OpenAI released the first few short films that it’s created using Sora with some professional filmmakers and artists. None of these videos will blow you away, it’s nothing that couldn’t be created with other technologies, but they were probably made in a fraction of the time at a fraction of the cost.
The AI Mirror Test
This is a really interesting little test that Josh Whiton did last week. His write up on Twitter is worth a read and he makes a good point. Even if AI isn’t truly self aware, then it’s very good at closely replicating self awareness. And in practice, is there really any difference?
AI Ethics News
Ready or not, AI chatbots are here to help with Gen Z’s mental health struggles
Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence
‘Resist this’: outrage as BBC replace Mamma Mia! star with AI voiceover
Adobe's GenStudio brings brand-safe generative AI to marketers
A personalized chatbot is more likely to change your mind than another human, study finds
As AI booms, land near nuclear power plants becomes hot real estate
Long Reads
MIT Technology Review - Large language models can do jaw-dropping things. But nobody knows exactly why.
“The future is already here, it’s just not evenly distributed.“
William Gibson