AGI until you reach crit cap.

If you define AGI as having the ability to think like a human then the main challenge is the set of algorithms we use.

The human brain is fundamentally different from a computer. Machine learning neural networks are only very simple models of the human brain. If we gain more understanding of the brain and develop algorithms that better model it, we might be able to develop a practical AGI.

If you take a course in design and analysis of algorithm, you will realize computing power cannot overcome certain problems. There are mathematical proofs that will show you the limitations of current computers. Quantum/photonic computing may change things up, but ultimately the right set of algorithms must still be developed. Even with the current architecture, if you develop the right algorithms on parallel supercomputers you might get a practical AGI.

Notice I say practical AGI, because it's currently impossible to develop an AGI that perfectly models the human brain. We just don't fully understand the human brain and cognition yet, let alone build a perfect model of it.

The other major challenge is money. I work in AI and it's not as well funded as we hope. We need more investments. I need investment lol.

They're basically completely separate. Google and DeepMind are both subsidiaries of Alphabet. DeepMind was its own company aimed at achieving AGI. When it got acquired by Alphabet, very little seems to have changed, except that they obviously got a lot more resources. Google Brain is a team within Google Research that focuses on machine learning projects "across different time horizons and levels of risk".

According to this 2017 AMA DeepMind and GBrain are quite similar. And although they're separate, there are quite a few collaborations. My impression is that DeepMind has more of a focus on AGI though, whereas Google Brain is (also) more focused on contemporary applications.

I'm curious what you're referring to with GBrain's "whole new paradigm". My impression is that DeepMind is more innovative. Sure, they're mostly using deep learning and RL, but it's not like RL was already working very well before DeepMind started working on it. And maybe I'm also expecting Shane Legg to (less publicly) continue his research on universal AI (with a lot of the former students DeepMind plucked from Marcus Hutter's lab).

IX. Financial markets are the most powerful information aggregators produced by the universe (so far)

I mean this is machineworld ultima shit, fuck me, using the EMH to determine the sigil in the stars and entrails that indicates the time of the coming of the basilisk

This is the best find on the sub in just the absolute longest fucking time