There is lots of code in our DNA yes (eye reflexes, semantics, etc). We do have a big hardware intensive brain. We do collect lots of data. But look at how close DALL-E etc is to AGI. Surely, there is not that many things we need in our DNA to build AGI if GPT-3 has many of our mental mechanisms and results. And my AGI Guide really makes you feel like everything is covered, we are really damn close, worst case we are 50% close, all those little things getting it to work etc are really small and come fast once we get the bigger picture and BAMN got it, ex. it's all prediction/patterns, these patterns help us get the rest of AGI faster.
As for our hardware brain, yes we are kinda far from bamn got a super computer in our own house, but, not only will an ASI run on your computer (assuming it only eats/ generates a small amount of data, anyone in AI knows this), but we do have supercomputers that will run teams of AGIs, just look at DALL-E it does wonders and runs on your PC, and can clone just reached ASIs once made which takes less time than teaching each, and eating data for them is fast. So the compute needed and memory is pretty small really, these things run already, you'll quickly get a adult ASI then quickly clone it, they will run in parallel and update each other. Brute Force Search is most costly, ASI won't need so much time or data examples, it will quickly see the answers. This thing will run efficiently while take a trillion less time to find cures. Our human brains do the same, we find answers with little memory/speed resources, faster than if were ANIs with biggg resources or if were BFS. Yes I know our brain is so "big" but in the end a lot of it is extra headroom and has connections already built that it doesn't necessarily need. And the data amount they need, we have loads of data on the internet and can make more. It will generate its own too in its brain. They also eat data faster than us, and can clone their brain. Again just look at how accurate and diverse DALL-E is. It's clear we are closing in on the needed code, hardware, and data trained on. We should get just over AGI by 2030-2040. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T54594b98b5b98f83-M77de69a3a5d2e37606a448b9 Delivery options: https://agi.topicbox.com/groups/agi/subscription
