Anastasios, ANASTASIOS> First, we are looking for the minimal complexity that can bootstrap itself. SERGIO> That's a causal set with code that can minimize the action functional, and deal with I/O. I/O is always causal sets coming directly from sensors or going directly to actuators.
Humanity, efficient architectures, objectives (why "four"?), will come in as input once the bootstrap is operational and ready to digest them. Shortcuts of all kinds will be critical in the algorithm that minimizes the functional. As long as the functional is minimized, it doesn't matter how. This is an implementation detail. Implications will be much easier to understand when we begin to get results from the machine, rather than just speculating about them. ANASTASIOS> Or, it would be nice if someone had an idea about those 3 dimensions! SERGIO> Last month I watched a movie recommended by Matt, called "The story of 1." The movie expands on the cause-effect relationships over many centuries that led to the discovery of number 1. Merchants counted sheep by making marks on a stick. Centuries later they started grouping the marks in groups of 5. Later, they noticed that they could count the groups. A hierarchy emerged from there, and the first number system was created. Arabic numbers emerged in India, travelled to the arab world, then crossed the Mediterranean and came in conflict with christian Europe, where a different system was used. It's a fantastic example of the role of causality in evolution. An understanding of causality is what is needed to create input for the AGI. The story of 1 is causality. By simply writing that story as a causal set, one would already have a first formal, very coarse-grained description of number 1. Enter it to the AGI, and let the AGI use it as it would, and also learn about it. This is the critical point: the AGI can LEARN! As it learns, it will progressively refine the description it has. Automatically, without human intervention. Humans only provide the information as input. This is just how you and me would do it. Suppose we want to understand better how merchants begun grouping their marks. What would we do? We would use what we know, search for some more on the Internet, in books, in Wikipedia. Pretty soon, we would be experts on the number 1. So also will the AGI. I wish somebody made a film about 3D. I don't know the story of 3D. When was 3D "discovered" by civilization? Euclide? Did cavemen know they were 3D? Do 2012 Kalahari bushmen know they are 3D? We don't need to know about 3D to have a functioning brain. If your approach takes you to a point where 3D becomes a requirement, then you are on the wrong path. Sergio -----Original Message----- From: Anastasios Tsiolakidis [mailto:[email protected]] Sent: Monday, September 10, 2012 5:17 PM To: AGI Subject: Re: [agi] Discovering physical dimensions On Mon, Sep 10, 2012 at 10:51 PM, Sergio Pissanetzky <[email protected]> wrote: > The general mistake here, is that everybody is thinking that "I, ME, MYSELF" > must do it. Discover physical dimensions, answer questions about vision, > interactions, etc. That's not AGI, that's more of I, me, myself. Me myself and I are concerned with the exact opposite of your depiction. First, we are looking for the minimal complexity that can bootstrap itself. Second, we are concerned with the "humanity" of the intelligence, from a human point of view it is not OK to answer problems after 10 million years, or by outputting 1trillion symbols, in whatever language. Third, we are looking for efficient architectures, even if we have answered the first problem it would be extremely nice to have all kinds of shortcuts in place, eg needing 100TB of storage instead of 100XB. Fourth, it is hard for us to remember four objectives, which is one of the many reasons we need intelligent machines. Certainly a half-AGI and half-spiritual interest of mine is to investigate the very implications of the human "hardwiring", whether it is about issues of free will, agency, qualia, socially distributed knowledge and skills etc. Or, it would be nice if someone had an idea about those 3 dimensions! AT ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/18883996-f0d58d57 Modify Your Subscription: https://www.listbox.com/member/?&; Powered by Listbox: http://www.listbox.com ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-c97d2393 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-2484a968 Powered by Listbox: http://www.listbox.com
