A week in Generative AI: Gemini, Mistral & Suno
News for the week ending 24th November 2024
This week has mostly been filled with some small feature releases and new versions of models being released. First up we have memory being added to Google’s Gemini chatbot. We then have some big updates from Mistral, who have added Web Search, Canvas, and many other new features to their Le Chat platform. Lastly, we have a new music model, v4 from Suno, which greatly improves on v3. Oh, and a nice little update video from Figure AI on their robotics work with BMW.
On the Ethics front, there’s more news on Trumps anti-regulation stance, reports that OpenAI is funding research into ‘AI morality‘ which sounds very promising, and a new web standard that could improve how LLMs read and process online content.
In Long Reads, Ben Evans has released his annual technology presentation that he shared at Slush Festival this week - AI Eats The World is worth a look.
Enjoy!
Google’s Gemini chatbot now has memory
I think memory is one of the most underrated features of ChatGPT and its surprised me that none of the other frontier AI companies had released a similar feature until now. Memory is a really important feature to move generative AI chat models on from giving generic answers to more personal answers and a key building block for them to become more like digital companions.
Gemini’s memory feature works just like ChatGPT’s where it will remember things from previous chats to help add more context to your future conversations. You can also ask it remember specific things for you.
The feature is currently being rolled out and is only available for premium subscribers via the web client.
Mistral has entered the chat
Lots of new features dropped by Mistral this week for their LeChat platform:
Web search with citations
Canvas for ideation, in-line editing, and export
State of the art document and image understanding, powered by their new Pixtral Large model.
Image generation, powered by Black Forest Labs Flux Pro
This mostly brings Le Chat up to par with the likes of Claude, ChatGPT, and Gemini with the exception of a memory feature which both ChatGPT and Gemini now have. It’s great to see the French start up continuing to compete with the US frontier AI companies when it comes to both models and features!
Suno releases v4 music generating model
It’s hard to believe that at the beginning of the year there were no generative AI models that could create usable music and now we have v4 of Suno released that produces incredibly high quality music.
Not only has the music generation come along way this year, but the tools and interface has evolved a lot too. There are lots of great features now available in the Suno platform to change lyrics, add instrumental breaks, have more control over the style of music generated and much more.
I think it’s safe to say that music generation is now ‘ticked off’ the list of things for generative AI to master. Yes, there’s still lots of room for improvement, but the quality and the tools are now at a level that anyone can create high quality music just with their imagination.
Figure shares update on Figure 02
This is a great little update video, showing the progress Figure AI have been making with their Figure 02 humanoid robot and the testing they’re doing with BMW. The video shows the robots working in a mocked-up factory setting and performing various manual tasks.
The video claims a 7x increase in reliability and a 4x speed increase. It still looks quite slow to me, but bear in mind that these robots can in theory work 24x7, only pausing to charge, so working at human speeds won’t be necessary to start off with.
AI Ethics News
‘An AI Fukushima is inevitable’: scientists discuss technology’s immense potential and dangers
Trump’s Anti-Regulation Pitch Is Exactly What the AI Industry Wants to Hear
New "llms.txt" web standard could fundamentally change how LLMs read and process online content
New AI Tool Tracks Your Steps by Reading the Bacteria You Carry
AI could cause ‘social ruptures’ between people who disagree on its sentience
Marissa Mayer just laid out a possible business model for ad-supported AI chatbots
Long Reads
Benedict Evans - AI Eats The World
“The future is already here, it’s just not evenly distributed.“
William Gibson