Google Brain

So I'm sure everyone here has heard of the transformer and sort of self-attention.

This was sort of the moment where Researchers at Google Brain came up with a new architecture that actually allowed scaling language model training.

It was way more performance on existing hardware, so could actually run these workloads on GPUs.

And compared to previous neural net architectures like recurral neural nets or LSTMs, they were able to employ this technique called attention, which basically led to way better performance in next token prediction.

So, yeah, the biggest mistake is to not be thinking about it.

or to think as some very, very well-informed people who are much smarter than me, like Andrew Ng, who was one of the co-founders of Google Brain, and Yann LeCun, who was one of the three godfathers of AI.

He was one of the people who ushered in deep learning.

They think that we will not get effectively a new species on this planet which is smarter than us.

Our first story looks at AI pioneer Andrew Ng and his call for stronger action against the misuse of artificial intelligence, particularly the creation of non-consensual deepfake intimate images.

Speaking on The Economist's Inside Tech podcast, the Google Brain co-founder highlighted the need for strict regulations and meaningful penalties to prevent people from using AI to harm others.

His concern is that increasingly powerful tools can make it easier to create realistic, explicit images of people without their consent.

But Ng draws a clear distinction between the dangers of AI misuse and the idea that artificial intelligence itself could threaten humanity's existence.

And, you know, up to this point, Google has had pretty much the run of the table on AI talent.

They have this division Google Brain that is considered sort of the all-stars of AI.

And then they make this acquisition.

And this is pretty...

I talked to Andrew Ng recently, and he's the one who's been in the space for more than 20 years.

He started Google Brain, he co-founded Coursera, and he has one of the best courses on machine learning.

He's very positive, and I tend to trust him because he's seen the market, he's seen what was going on, but he's also practically using it, and He thinks that, you know, when we talk about airplanes, for example, yes, there is a small chance it's going to crash, right? But the amount of good it brought to the world with, you know, travel, me being here in the US and my family being in Thailand, isn't that amazing that I can just hop in the plane and see them? Same with AI.

All the hopefully amazing things it's going to be able to do when it comes to health care and creating drugs that don't hurt you but only help you because they're made for you.

In contrast, Sundar Pichai, CEO of Alphabet, has championed a more optimistic timeline earlier in 2025.

Speaking at the World Government Summit in Dubai, Pichai predicted that practically useful quantum computers could emerge within five to ten years, drawing comparison to the early 2010s development of AI when Google Brain was just starting.

Bottom line is, the road to a cryptographically relevant quantum machine is paved with formidable hurdles.

The first and arguably the biggest hurdle is the qubit stability.