Well, this week was an absolutely crazy week in the world of AI. It started with a prediction from Anthropic CEO Dario Amodei that it is only two to three years until AI systems are better than humans at almost everything, then eventually better than all humans at everything.
Timelines to AGI are rarely predicted to be greater than five years these days. Then, completely unexpectedly, the Deepseek R1 model from China stole all the headlines, breaking barriers in so many areas. The model is achieving benchmark scores that are comparable to the ChatGPTo1 model, but at a tiny fraction of the price. And it’s open source.
It’s also changing the game by leaning heavily on reinforcement learning rather than huge training runs. It’s been astounding its users with the way that it ‘thinks’, and there are some fascinating examples. We’re 25 days into 2025 and the boundaries of what’s possible are already being redefined.
Then on Tuesday we had some massive news. Donald Trump announced an investment of around $500 billion into a US AI infrastructure project named Stargate. This is a collaborative effort between OpenAI, Microsoft, Nvidia, Oracle and more.
It’s designed to put America at the forefront of the great AI arms race. It will likely entail huge data centres in Texas and other locations that are unlike anything we’ve ever seen. OpenAI followed up with a statement about Project Stargate that concluded with the following, “all of us look forward to continuing to build and develop AI, and in particular AGI, for the benefit of all of humanity.
We believe that this new step is critical on the path and will enable creative people to figure out how to use AI to elevate humanity.” OpenAI followed up with several other big announcements this week. First it was the release of their ‘Operator’ tool.
Operator can take control of a browser and complete sequences of tasks. It’s built on top of the GPT4o model, but sadly it’s only currently available on the $200 Pro plan. An AI that runs tasks on your behalf isn’t exactly groundbreaking, but it feels like it’s the first properly integrated version of an AI agent into a big platform.
Perplexity followed suit with their own operator style update, but this one anyone can use for free. The Perplexity Assistant is now available on the Android app. It can be set as the native assistant on your device and it can call on various apps to perform a series of tasks across several different services.
It’s a very impressive update that’s gone a bit under the radar with all of the other big news going on. Following on from the operator announcement, Sam Altman then confirmed that the free tier of ChatGPT is going to get the o3-mini model soon. It seems like the Deepseek R1 model has put pressure on all the big players to start offering more.
In Generative AI, Kling announced that its new ‘Elements’ feature is now live. This allows users to add up to four reference images that are then combined to generate a video that will include all of those components. It’s an amazing new tool that gives far more control over video composition.
Pika Labs gave early access to some users to its 2.1 version and so far we’ve seen some fantastic results. It seems to look on par with VEO for cinematic feel and be producing dynamic motion similar to Ray2. Harvard says that AI tutors are better than Harvard professors.
Students who used an AI tutor in a Harvard physics class, instead of the professors teaching, learned more than twice as much and apparently they did it in less time and were more motivated and engaged. Learning, like most disciplines, is being positively disrupted by the use of AI.
Xanadu, a quantum computing company, announced the world’s first universal photonic quantum computer. It’s a 12 qubit machine that consists of 35 photonic chips and runs at room temperature. Christian Wiedbrook stated that, “Xanadu has now solved scalability.
We now could in principle scale up to thousands of server racks and millions of qubits.” If all of this is starting to make you feel anxious about AI taking over, don’t worry. I’m sure it’s all under control. I probably shouldn’t mention that at a UN Security Council meeting, top diplomats warned, “the fate of humanity must never be left to the black box of an algorithm.” The great AI wars of 2025 are well underway.
Hope you all enjoyed this week’s roundup. Let’s see if the coming week calms down a little or continues to ramp up.