On Thu, Aug 20, 2020, 1:09 AM Alan Grimes via AGI <[email protected]> wrote:
> Ok, I haz quite a bit of hardware atm, > > so here's a question: How actually do you propose to go about developing > an AGI? Considering that it will, inherently, be a child-mind at first > that needs stimulus and a playspace, how do you propose to do that? > We can skip the child rearing phase. The complexity of the human mind is 10^9 bits, half from DNA and half stored in long term memory. We aren't restricted to using the evolved model with all it's limitations. We already have machines that can do more than humans. The Turing test would be a big step backwards. The obvious applications of AGI are automating labor and uploading. Both require models of human behavior, the ability to predict your actions. First, to save you time by giving you what you want without having to ask. Second, to have a robot that looks like you carry out your predicted actions in real time. I estimate the value of automating the global economy to be world GDP divided by interest rates, or $1 quadrillion. Uploading all humans is valued at world GDP times life expectancy, or $5 quadrillion. I think these can be achieved with advances in nanotechnology (transistors require too much power) and a willingness to surrender privacy. Both trends are promising. Personally I can only make a tiny contribution to this effort in the form of language modeling research evaluated by text compression. That's all any of us can do. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T49f9cbeb186483fb-Mca244c9d5f5ab34afec7e2d0 Delivery options: https://agi.topicbox.com/groups/agi/subscription
