Oh, I saw a very interesting news story about Jed McCaleb. The founder of Ripple and Stellar is putting serious money on the table for AGI research. We're talking about 1 billion dollars through the Astera Institute, plus 600 million focused on neuroscience. This comes from a cryptocurrency fortune valued at around 3.9 billion.

What catches attention is the different approach that Jed McCaleb is betting on. Instead of following the path of the Transformers that dominate now, the institute is investigating how the human brain actually works. They are starting with rats, mapping neural activity while performing tasks through brain-machine interfaces, then planning to expand to monkeys and eventually humans. The idea is to turn these discoveries into new AI architectures.

The project is led by Dileep George, a former executive at DeepMind, and they plan to scale up to 30 researchers this year. McCaleb has a very clear critique of the status quo: the current Transformer architecture only makes predictions. Crucial elements like planning, decision-making, and motivation are missing. According to him, an AI based on the principles of the human brain would have a better chance of being understood and controlled by humans.

Interestingly, Yann LeCun, former head of AI at Meta, recently created the AMI lab and secured 1 billion in seed funding to research "world models" in the same direction. But Sam Altman from OpenAI disagrees, arguing that AGI will require multiple medium-scale discoveries, not a single new direction.

So basically, Jed McCaleb is betting heavily that the solution lies in better understanding how the brain works, not in scaling up current models further. It will be interesting to see how this evolves.