A week in Generative AI: Character.ai, Apple Intelligence & Robots
News for the week ending 4th August 2024
It’s been another big week in generative AI this week, with big changes at Character.ai, initial availability of OpenAI’s advanced voice mode for ChatGPT 4o, and the first beta testing of Apple Intelligence. There’s also been lots gf great robotics news and a glimpse of the next generation humanoid robot from Figure AI.
On the ethics front, OpenAI have pledged to give the US Safety Institute early access to their next model, the UK is investigating ties between Google and Anthropic, and the UK government has shelved £1.3bn of funding for technology and AI projects.
In the long reads, there’s a great interview between Jensen Huang of NVIDIA and Mark Zuckerberg of Meta, and a really good comparison from Ethan Mollick between GPT-4o’s new voice mode with the new Siri powered by Apple Intelligence.
Character.AI CEO Noam Shazeer returns to Google
I’ve written regularly about Character.ai and how promising a startup it was, genuinely offering something different in the generative AI space and growing a large user base amongst the under 25s. So this move by Noam Shazeer, and his fellow co-founder Daniel De Freitas is quite a surprise!
They both left Google 4 years ago to found Character.ai and are now returning to Google DeepMind along with a non-exclusive agreement for Google to use their tech. This is similar to the move by Microsoft when they bought in Mustafa Sulleyman back in March and effectively acquired his generative AI start-up Pi.
Ultimately, I think this is all about consolidation in the generative AI sector - it’s incredibly expensive to build companies that compete at the frontier and sustainable revenues haven’t yet been found, meaning it’s only the Big Tech companies that are able to buy up the talent and the technologies.
I hope Character.ai carries on operating and pushing the boundaries of what’s possible with generative AI as I think they really bring something different to the party that shows a lot of promise.
OpenAI releases advanced voice mode to small number of users
After a bit of a delay, OpenAI has started giving access to GPT-4o’s advanced voice mode to a small number of users. Originally demoed back in May, advanced voice mode allows GPT-4o to have much more nuanced, and emotional, conversations with users. It also allows users to interrupt and have a much more natural conversation with ChatGPT. I haven’t had a chance to try it yet, but the feedback from commentators has been very good. Ethan Mollick has a good overview of it vs. the new Apple Intelligent powered Siri here.
Developers get access to Apple Intelligence for the first time
This week Apple released the first beta of iOS 18.1, which is the version of iOS 18 that will contain many of the Apple Intelligence features that Apple announced at WWDC in June. Apple Intelligence is currently only available in the US, but there are still ways it can be accessed from other markets.
I haven’t had a chance to try and of the new features yet, but initial reports have been good. There’s a revamped Siri experience, new writing tools, and call transcription and summary features.
It was also confirmed on Apple’s earnings call this week that the rollout of all of Apple Intelligence’s features will be staggered across the rest of this year. Here’s how Tim Cook described it:
The rollout, as we mentioned in June, we’ve actually started with developers this week. We started with some features of Apple Intelligence, not the complete suite. There are other features like languages beyond U.S. English that will happen over the course of the year, and there are other features that will happen over the course of the year, and ChatGPT is integrated by the end of the calendar year, and so, yes, it is a staggered launch.
I’m very excited for Apple Intelligence as I think it will take generative AI beyond chatbots and copilots into a much more product-led and user-friendly experience. I have high hopes that Apple will address the product-market fit problem that we’ve had with generative AI so far, allowing more users to access the technology and get more value from it.
Microsoft says OpenAI is now a competitor in AI and search
Following the giving up of their observer role on OpenAI’s board last month, Microsoft now says in its annual report that it considers OpenAI as a competitor in AI and search. Microsoft is the biggest investor in OpenAI, owning c.49% of the company having invested $13bn over the last few years.
For all intents and purposes, this doesn’t really change their relationship and is probably more to ward off some of the antitrust concerns in the US and similar regulatory scrutiny in the EU. When their partnership was first formed both companies acknowledged the potential for competition, and this is just that.
Watch a robot peel vegetables with human-like dexterity
This is a great demo of a robot peeling vegetables - a task that requires a lot of dexterity and ability to handle delicate objects, which is a very hard robotics task. This will hugely broaden the number of tasks robots can perform, especially in the manufacturing of delicate objects.
Figure AI teases Figure 02 Robotics platform
Looks pretty slick! Looking forward to seeing more and with their partnership with OpenAI, I expect this to be powered by GPT-4o with the new voice mode, so I’m sure it will have a very human-like voice and you’ll be able to interact with it very naturally.
NVIDIA discovers way to scale robotics data
This is a bit geeky and technical, but incredibly impressive. It’s worth watching the video - NVIDIA have demonstrated a way to scale robotics data in simulation, allowing us to train robots on much larger data sets, improving their ability to adapt to more situations. This will really accelerate the tasks humanoid robots will be able to perform.
AI Ethics News
OpenAI pledges to give U.S. AI Safety Institute early access to its next model
UK antitrust body probes Google’s ties with AI rival Anthropic
Anger mounts over environmental cost of Google data centre in Uruguay
Elon Musk’s X under pressure from regulators over data harvesting for Grok AI
Perplexity is cutting checks to publishers following plagiarism accusations
OpenAI’s Sam Altman is becoming one of the most powerful people on Earth. We should be very afraid
Long Reads
SIGGRAPH 2024: AI and The Next Computing Platforms With Jensen Huang and Mark Zuckerberg
The Guardian: Will OpenAI’s $5bn gamble on chatbots pay off? Only if you use them
One Useful Thing: On speaking to AI
“The future is already here, it’s just not evenly distributed.“
William Gibson