A week in Generative AI: Energy, Character.ai & 3D Assets
A week in Generative AI: Energy, Character.ai & 3D Assets
News for the week ending 7th July 2024
There hasn’t been as much news on the frontier of generative AI this week, but there have been some really interesting reports published around AI’s energy demand and new voice features released from both Character.ai and ElevenLabs. There has also been some really great work from Meta on generating 3D assets that will be a bigger deal than most people think. Oh, and a giant robot 🤖.
What’s going on with AI’s energy demand?
There has been a lot of great reporting recently on the increasing energy consumption of AI. We’ve had an article from Bloomberg on how AI is wreaking havoc on global power systems, a report from Goldman Sachs on how AI is poised to drive 160% increase in data centre power demand, and ArsTechnica have taken a closer look at AI’s supposed energy apocalypse.
To add to the list, we now have a Guardian article reporting that Google’s emissions have climbed nearly 50% in five years due to AI energy demand, and Bill Gates saying we shouldn’t worry about AI’s energy draw as ultimately AI will identify ways to help cut power consumption.
There are a lot of people, all saying a lot of things but fundamentally our demand for AI-technologies is increasing, which means as this continues (which it will), data centres are set to use c.4% of global energy by 2030. This sounds like a staggering amount (which it is), but it’s worth noting that refrigerators and air conditioners account for 17%-20% of global energy use. Who knew?! 🤯🥶🤯
Regardless, there are only really two solutions to the problem - we need to find better ways of generating high volumes of carbon-neutral energy (Sam Altman’s big bet is on nuclear fusion company Helion), and more importantly we need to transition to less power-hungry AI hardware. NVIDIA’s latest Blackwell chip claims a 25x improvement in energy consumption, but specialised AI processors such as Groq and Sohu take this further with claims of being 10x more efficient than Blackwell.
Hopefully we can find ways to improve the energy efficiency of AI hardware faster than demand increases for AI technologies.
Introducing Character Calls
We haven’t heard from Character.ai for a little while, but they’ve just released a new feature that allows you to have ‘phone calls’ with your favourite characters on the platform. This builds on the Character Voice feature that they launched a few months ago and there have been over a million voices created by users so far.
This is similar to the real-time conversations that OpenAI announced with GPT-4o (which are delayed) but instead of just having a few voices to choose from you can choose from millions of characters that already exist, or even create your own with their own unique personalities.
Character.ai is big amongst the under 25s, and there is a lot of excitement in the big tech companies to follow their lead. Meta has deployed something similar in their Chatbots and Google is reportedly now developing a similar concept as well.
Listen to your favourite books and articles voiced by iconic stars
This was bound to happen at some point, but I’m glad to see ElevenLabs doing this the right way by partnering with the estates of iconic stars. So far they’ve partnered to bring Judy Garland, James Dean, Burt Reynolds and Sir Laurence Olivier’s voices to their Reader App. This means that users can listen to any text (articles, PDFs, ePubs, newsletter etc.) in any of these actors voices.
This reminds me a little of when there was craze of celebrity-voiced sat navs from the likes of TomTom and Garmin, so I’m not sure if this will end up just being a gimmick or whether people will genuinely like being able to choose the voice of different celebrities to read to them.
Meta announces 3D Gen: fast test-to-3D asset creation
I think this is a bigger deal than it looks like on the surface (pun intended!).
If 3D assets become as easy to create as 2D images, we’re going to see an explosion of creativity in video, augmented reality (AR) and virtual reality (VR). 3D assets are already starting to replace 2D assets in a lot of advertising creative as they can be programmatically manipulated and changed more easily, leading to more variation for testing and optimisation.
I’m interested in how 3D asset generation and video generation start to come together in new creative tools over the coming years as I think it could be the catalyst for AR and VR to finally start taking off in the mainstream.
Japan introduces enormous humanoid robot to maintain train lines
The is the biggest humanoid robot I’ve seen, and a bit like if the Terminator hooked up with a fork-lift truck. It’s human-operated, but allows the engineers to perform tasks they just wouldn’t have been able to otherwise.
I think this is an under-appreciated area of robotics. A lot of the hype is around human-sized, humanoid robots that can perform tasks autonomously, but I think human operated robots that augment humans will be a much bigger deal in the short to medium term.
AI Ethics News
Google’s emissions climb nearly 50% in five years due to AI energy demand
Bill Gates says not to worry about AI gobbling up energy, tech will adapt
Anthropic looks to fund a new, more comprehensive generation of AI benchmarks
Apple Intelligence privacy sets a new standard, but it's not perfect
ChatGPT for macOS raises concerns for storing chats in plain text
OpenAI breach is a reminder that AI companies are treasure troves for hackers
News outlets are accusing Perplexity of plagiarism and unethical web scraping
Figma disables its AI design feature that appeared to be ripping off Apple's Weather app
Smudgy chins, weird hands, dodgy numbers: seven signs you’re watching a deepfake
Long Reads
One Useful Thing - Gradually, then Suddenly: Upon the Threshold
Six Colors - I’ll have my AI email your AI
MIT Technology Review - We analyzed 16,625 papers to figure out where AI is headed next
Sana - Jeanette Winterson on AI’s parallels with religion and the need for alternative intelligence
“The future is already here, it’s just not evenly distributed.“
William Gibson