Technology

➚ MOVE FAST: Energy. Amazon, Google, and Meta have signed a pledge to support tripling nuclear capacity by 2050 as they increase energy usage to develop AI products while trying to reduce their carbon footprint. Notably absent from the pledge were Microsoft and Apple.

➘ BREAK THINGS: Power. TSMC has pitched Nvidia, AMD, and Broadcom on a joint venture that would operate Intel’s foundry business, Reuters reported. The proposal comes on the heels of Trump reportedly calling for TSMC to help rebuild Intel, as the American tech giant’s future remains in flux.

Step aside agentic AI; angelic AI is having a transcendent moment. A Lutheran church in Finland last week held a service almost entirely prepared by artificial intelligence tools — complete with an original sermon, visual depictions of Jesus and Satan, and musical accompaniment, the Associated Press reported. Also portrayed were digital avatars of the church’s pastors and a former Finnish president (dead for four decades) who gave an Old Testament reading.

ChatGPT wrote most of the words; Massachusetts startup Suno played pop-inspired music; London-based Synthesia created the pastors’ avatars; and Silicon Valley’s Akool generated a video interaction between Jesus and Satan. Though not the first church to use the new technologies, the service represents the many ways AI is integrating itself into every aspect of daily life.

AI still has its limits. Attendees who spoke with the AP described the service as entertaining but distant, lacking the empathy and soul that comes with an in-person pastor. ChatGPT also wouldn’t give blessings or absolution, according to the church’s vicar (a good thing, we think). On a more basic level, AI could help religious leaders prepare sermons with its broad understanding of religious texts and current events.

As conversational AI improves, it will be interesting to watch whether religious individuals will be able to connect with chatbots in a meaningful way — receiving the kind of comfort that comes with attending church or praying, for example. If that happens, it may represent a new crossed threshold for the technology, but it could also present a new philosophical problem: ensuring a greater divide between God and science.

Google unveiled Gemini Robotics earlier this morning, a new family of AI models based on Gemini 2.0, which it released late last year. The accompanying research paper shows off some astonishing capabilities, like AI robots that can fold origami, offering a glimpse into the future of AI.

“Gemini 2.0’s embodied reasoning capabilities make it possible to control a robot without it ever having been trained with any robot action data,” the paper reads. “It can perform all the necessary steps, perception, state estimation, spatial reasoning, planning and control, out of the box. Whereas previous work needed to compose multiple models to this end.”

In other words, you build the physical robot, Google’s AI models will make it work.

This is an area we’ve been watching closely. We wrote last month about Apptronik, a humanoid robotics startup that partnered with Google DeepMind. Jeff Cardenas, the company’s CEO, told me he believed the final roadblock to building humanoid robots had been cleared — powerful, general-purpose AI models capable of understanding the physical world.

Apptronik made a bet that the trajectory of AI software capabilities would eventually enable robotics. So it decided to build the hardware platform — the custom-designed digits with fine motor skill ability, for instance, and a way for some company to come along with an AI model that can control its Apollo robots.

That is essentially what DeepMind unveiled today. Granted, it is still in the research phase, and it will be a while before we have Rosey the Robot. But the key here is that Google has built a foundation model that uses the real world to learn how to reason.

Large language models have shown some ability to “reason” when given text inputs, but it’s also very easy to show the limitations of text-based reasoning.

DeepMind’s leadership has long believed that the path to artificial general intelligence is multimodal. Put simply, if we want to have AI models that can reason like humans, they need to learn to use the same inputs humans use.

The human brain remains a marvel. Whatever our neurons are doing as our brains develop, the process is built on a foundation of patterns represented in the physical world, even before we learn language and can read text.

I recently asked Google DeepMind co-founder Demis Hassabis about whether he believed Gemini would evolve into the same model that powers robotics and self-driving cars and he said, essentially, yes.

The paper today shows pretty clearly how that will happen. It’s inevitable that Google’s robotaxi company, Waymo, will eventually replace its current technology with a single AI model, likely some future version of Gemini.

An important thing to note: Multimodal AI requires far more compute than text-based AI. There are more “tokens” in images than in text. This is partly why Google has been pushing the limits of AI “context” windows, or how much information models are able to hold onto before they forget what they were doing.

As companies like Google, Amazon, Meta, and Microsoft spend hundreds of billions of dollars on new data centers packed with powerful AI processors, we will see an acceleration of AI technology that uses a model of the world to reason. These models will enable robotics and autonomous vehicles, but they will be the same ones that write code and produce research reports.

If that seems odd, think about how well you’d be able to reason without ever having a sense of smell, taste, or touch.

The amount US consumers lost to scams in 2024, up $2.5 billion from the year before, according to new FTC data. Most of the scams involved fake digital investments and imposters through online or phone channels. Blockchain analysis firm Chainalysis found that crypto fraud likely hit a record high last year of nearly $10 billion globally, bolstered by “pig butchering” scams — which include a fake romantic relationship — and the growth of AI, which makes scamming easier.

Meta has begun testing its first in-house chip that will train its AI systems, a move intended to reduce soaring infrastructure costs and its dependence on suppliers like Nvidia, Reuters reported.

The company is reportedly working with Taiwanese semiconductor giant TSMC on a chip designed for AI-related tasks, making it more efficient than the typical GPUs often used for such work. Executives have publicly discussed the development of custom silicon, though previous test runs were challenging (Meta shuttered a planned 2022 rollout of an inference chip, resorting to Nvidia GPUs). But the latest news is a step forward toward Meta’s goal of using internal training chips by next year. It has reportedly finished the first “tape-out,” which involves sending the design to a chip foundry, signifying it can actually be manufactured.

Several big tech companies including Microsoft and Google are years ahead in their custom chip ambitions. OpenAI, on the other hand, is in a similar position as Meta, developing an in-house chip with hopes to launch next year. Such plans bring the future of Nvidia into question, especially on the heels of DeepSeek’s R1 model launch, which demonstrated that quality AI models don’t necessarily need a massive number of costly chips.

The CEO of the largest US electricity provider criticized the Trump administration’s decision to double down on gas-fired power to solve the looming power shortage in the US spurred by AI.

John Ketchum, CEO of NextEra Energy, told Semafor’s Tim McDonnell on the sidelines of the CERAWeek summit in Houston that the cost of gas turbines and the skilled labor to install them are both up three-fold from just two years ago. Renewables and batteries are the cheapest, fastest, and easiest way to meet the power demand from data centers, he added.