A week in Generative AI: Apple Car, Genie & Figure AI
A week in Generative AI: Apple Car, Genie & Figure AI
News for the week ending 3rd March 2024
Apple is shifting focus away from developing an automated car towards generative AI (and potentially robotics 🤖), whilst Google continues it’s 2024 streak of a big weekly announcement, but also details emerge of how it’s asking smaller publishers to test a troubling new GenAI product. Oh, and robots. You know I’m big on robots this year!
Apple to Wind Down Electric Car Effort After Decade-long Odyssey
I wasn’t surprised that this happened, the project has been troubled for years and I could never envisage Apple manufacturing and shipping a $100k car.
It’s been widely reported that Apple will be reassigning many of the engineers from Project Titan to work on AI. Many commentators believe this is a sign of Apple doubling down on its commitment to generative AI and new features that will be arriving in iOS 18 later this year.
I think that the reality is a little more nuanced than this though. Apple has a lot of engineers from Project Titan that are experts in hardware systems as much as software systems and I think the big pivot is more to do with what Apple is seeing in the market around robotics than it has to do with pure generative AI.
Could I see Apple shipping a $100k autonomous car in 5 years time? No.
Can I see Apple shipping a $20k humanoid robot in 5 years time? Absolutely!
I could be completely wrong, but I think we’ll start to hear rumours around Apple’s efforts on robotics over the next couple of years, replacing all the rumours we’ve had on Project Titan for the last decade.
Google announces Genie, a generative model for creating interactive environments
Not wanting to break a 2024 streak of announcing something big and significant every week, Google this week announced Genie which can generate an endless variety of playable worlds from a single image.
Similar to how OpenAI’s Sora isn’t just creating a video, it’s modelling reality, Genie isn’t just animating an image it’s making it interactive and playable. It’s not difficult to see how all these capabilities will come together over the next few years to create photo-realistic, playable virtual environments that at some point might become indistinguishable from the real world.
Figure AI secures $675m funding from Nvidia, Microsoft, OpenAI and Bezos
Figure AI is one of many interesting robotics companies that are now getting huge amounts of investment to build humanoid robots that can replace human labour at many tasks. They’re still a bit slow (around 20% of human speed) but this will improve, and the ability to work 24/7 will offset that. I think we’ll start to see humanoid robots deployed at scale in factories and warehouses in the next couple of years.
Google is paying publishers to test an unreleased GenAI platform
This does not sound like a good, ethical use of generative AI, especially in light of some many news organisations scaling back right now. Essentially, this tool (according to reports) indexes recently published articles generated by other sites, summarises them and publishes that as a new article. That’s not new, reporting, or journalism. That’s just repurposing existing content to try and generate site traffic.
Mistral AI releases new model to rival GPT-4 and its own chat assistant
This is some really impressive work from Mistal AI, who have been going from strength-to-strength the last few months. It’s great to see a Europe-based AI start up going toe-to-toe with the big US companies whilst simultaneously having such a strong commitment to the open-source community. There’s a lot OpenAI could learn from them!
Meta plans to launch Llama 3 in July
There’s not a huge amount of interesting information available on Llama 3 yet, but a July release would be 12 months after the release of Llama 2, so makes sense. It’s good to see more progress being made in open source models!
“The future is already here, it’s just not evenly distributed.“
William Gibson