Re: [agi] Non-evolutionary models of recursive self-improvement
On Sunday 14 September 2008, Dimitry Volfson wrote: Well, then I don't understand what you're looking for. Brain chemistry is part of the model. Check out one of the sentences: The thalamus in the limbic system ('leopard brain') converts the physical need into an urge within the cortex. So if I shoot a physical need at a thalamus sitting in my lab, it'll pop out an urge ? You're just talking about the output of the neurons, not the concept of urge that most people talk about from Webster's etc -- which is of the mind, not the brain. I'm not saying that the mind is separate from the brain, I'm just saying that people are confused and probably wrong when they talk about the mind. They most often are .. having no background in neuroscience, etc. If you look on the page, you see some implementation details like - Wants and needs have to struggle against one another in a priority list for action now or later or not at all. The strength of the urge is thus important, with strong urges leading to needs that jump the queue, demanding immediate action. I'm a programmer, I know what a list and queue look like, show it to me. Nobody has yet shown neurons doing math, much less a list object. - Bryan http://heybryan.org/ Engineers: http://heybryan.org/exp.html irc.freenode.net #hplusroadmap --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Non-evolutionary models of recursive self-improvement
Actually, I remember reading something about scientists finding a list structure in the brain of a bird singing a song (a moving pointer to the next item in a list sort of thing). But whatever. It's not a very low level model, but the lower level activation is implied. When you imagine a goal-state, the relationship is represented in the brain somehow (in the neurons of course). And when evidence of the actualization of that goal-state comes in through the senses, the brains sends an opiate reward, which might make the person want to do whatever that was again in the correct context. Motivation circuits - familiar with the concept? If a motivation circuit gets over-energized then a person gets locked into doing the same thing over and over again (and not getting the goal-state), rather than having enough resources left to think about doing something different and what that different thing should be. Does someone need to know exactly how a motivation circuit becomes over-energized at the neuronal level in order to model it in an AI? I don't think so. Many things like this are known. And people don't need to understand such at the individual-neuron level to model what happens. Bryan Bishop wrote: On Sunday 14 September 2008, Dimitry Volfson wrote: Well, then I don't understand what you're looking for. Brain chemistry is part of the model. Check out one of the sentences: The thalamus in the limbic system ('leopard brain') converts the physical need into an urge within the cortex. So if I shoot a physical need at a thalamus sitting in my lab, it'll pop out an urge ? You're just talking about the output of the neurons, not the concept of urge that most people talk about from Webster's etc -- which is of the mind, not the brain. I'm not saying that the mind is separate from the brain, I'm just saying that people are confused and probably wrong when they talk about the mind. They most often are .. having no background in neuroscience, etc. If you look on the page, you see some implementation details like - Wants and needs have to struggle against one another in a priority list for action now or later or not at all. The strength of the urge is thus important, with strong urges leading to needs that jump the queue, demanding immediate action. I'm a programmer, I know what a list and queue look like, show it to me. Nobody has yet shown neurons doing math, much less a list object. - Bryan http://heybryan.org/ Engineers: http://heybryan.org/exp.html irc.freenode.net #hplusroadmap --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?; Powered by Listbox: http://www.listbox.com Click to book your dream cruise. http://thirdpartyoffers.juno.com/TGL2141/fc/Ioyw6i3nL6Yyjtb9De1g2zOYT4IGQyJIHWTsd9hNE8UMIeJrcJo6ST/ --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Non-evolutionary models of recursive self-improvement
On Sunday 14 September 2008, Dimitry Volfson wrote: Actually, I remember reading something about scientists finding a list structure in the brain of a bird singing a song (a moving pointer to the next item in a list sort of thing). But whatever. That does sound interesting, yes, I'd like to find a citation on it. Do you know where I might find that? Was it a magazine, journal, etc.? It's not a very low level model, but the lower level activation is implied. How could it be used if it's left unspecified and hand-wavy? When you imagine a goal-state, the relationship is represented in the brain somehow (in the neurons of course). And when evidence of the Of course. But how? actualization of that goal-state comes in through the senses, the brains sends an opiate reward, which might make the person want to do whatever that was again in the correct context. How is it that only one class of molecules correlates to goalism? This seems suspect considering the complex infrastructure of the brain. Motivation circuits - familiar with the concept? Yes, but only from psychology, not from stuff we can actually build or understand. If a motivation circuit gets over-energized then a person gets locked into doing the same thing over and over again (and not getting the goal-state), rather than having enough resources left to think about Perseveration occurs for many other reasons than 'over-energized' though .. doing something different and what that different thing should be. Does someone need to know exactly how a motivation circuit becomes over-energized at the neuronal level in order to model it in an AI? I don't think so. Another illustration of the problem with this line of hypothesis that I have is that you're trying to make intelligence, a vague concept in the first place, with a foundation made out of motivation, another somewhat vague psychology concept. I don't care how many times the mouse hits the button, etc. Also, I recently cited this: http://www.ece.utexas.edu/~werner/siren_call.pdf It's an elaboration of a few of my points here. Many things like this are known. And people don't need to understand such at the individual-neuron level to model what happens. No, you miss my point. It's not that I'm saying there's some scale of microscopism that we have to climb down (brain, region, tissue slice, neuron, axon, subcellular mechanism, ..) to understand things; that's not it at all. What I'm talking about is actually considering the neurons as the physical components that make up the 'brain' which is the physical location of, supposedly, these 'goal-states'. These biological systems (brains) are the real things that can be experimentally tested or perhaps manipulated, but on the other hand the semantic space of goals, meaning, motivation is hardly meaningfully manipulated, even with the WordNet or Cyc relations. I can randomly generate new designs for experiments using WordNet or Cyc relations or something, where we observe mice subjected to a battery of different psychochemical compounds. Then, using WordNet, we can pull out random labels for each of the behaviors, maybe it's a goal or maybe it's a who knows what that the creature supposedly intrinsically has, and then what? You'd plot the data sets in some multidimensional manner, maybe a State Vector Machine, I'd have to ask some mathematicians, and then there's this strong likelihood of statistical irrelevance of assigning these labels to the different phenotypes observed in experiments. These same phenotypes are the things of folk psychology as well. The trick is that instead of observing rats, you're observing people. Given that scenario, what would I care if it's subatomic or neuron-level or whole brain level? So, no, our disagreement is about something else. - Bryan http://heybryan.org/ Engineers: http://heybryan.org/exp.html irc.freenode.net #hplusroadmap --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Non-evolutionary models of recursive self-improvement
Bryan Bishop wrote: Secondly, I'm still wondering about the representations of goals in the brain. So far, there has been no study showing the neurobiological basis of 'goal' in the human brain. As far as we know, it's folk psychology anyway, and it might not be 'true', since there's no hard physical evidence of the existence of goals. I'm talking about bottom-up existence, not top-down (top being us, humans and our social contexts and such). Look at The Brain's Urge System at ChangingMinds.org http://changingminds.org/explanations/brain/urge_system.htm: . Notice that the stimulus can be pure thought. Meaning that a mental image of a goal-state can form the basis of urge-desire-action. - Dimitry Looking for insurance? Click to compare and save big. http://thirdpartyoffers.juno.com/TGL2141/fc/Ioyw6i3m275cwCIqqfzaFFrlqKhcwvvYMpvWAlmDHTzZKHdEudmmKD/ --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Non-evolutionary models of recursive self-improvement
On Saturday 13 September 2008, Dimitry Volfson wrote: Look at The Brain's Urge System at ChangingMinds.org http://changingminds.org/explanations/brain/urge_system.htm: . Notice that the stimulus can be pure thought. Meaning that a mental image of a goal-state can form the basis of urge-desire-action. No, that's the fictional version of the 'mind', nothing about the actual brain. - Bryan http://heybryan.org/ Engineers: http://heybryan.org/exp.html irc.freenode.net #hplusroadmap --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Non-evolutionary models of recursive self-improvement (was: Ability to improve ones own efficiency as a measure of intelligence)
On Wednesday 10 September 2008, Matt Mahoney wrote: I have asked this list as well as the singularity and SL4 lists whether there are any non-evolutionary models (mathematical, software, physical, or biological) for recursive self improvement (RSI), i.e. where the parent and not the environment decides what the goal is and measures progress toward it. But as far as I know, the answer is no. Have considered resource constraint situations where parents kill their young? The runt of the litter or, sometimes, others - like when a lion takes over a pride. Mostly in the non-human, non-Chinese portions of the animal kingdom. (I refer to current events re: China's population constraints on female offspring, of course.) Secondly, I'm still wondering about the representations of goals in the brain. So far, there has been no study showing the neurobiological basis of 'goal' in the human brain. As far as we know, it's folk psychology anyway, and it might not be 'true', since there's no hard physical evidence of the existence of goals. I'm talking about bottom-up existence, not top-down (top being us, humans and our social contexts and such). Does RSI have to be measured with respect to goals? Can you prove to me that there exists no non-goal oriented improvement methodology? Keeping some possibilities open, as you can guess. I suspect that a non-goal oriented improvement function could fit into your thoughts in the same way that you might hope the goal variation of RSI would. - Bryan http://heybryan.org/ Engineers: http://heybryan.org/exp.html irc.freenode.net #hplusroadmap --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com
Re: [agi] Non-evolutionary models of recursive self-improvement (was: Ability to improve ones own efficiency as a measure of intelligence)
--- On Fri, 9/12/08, Bryan Bishop [EMAIL PROTECTED] wrote: On Wednesday 10 September 2008, Matt Mahoney wrote: I have asked this list as well as the singularity and SL4 lists whether there are any non-evolutionary models (mathematical, software, physical, or biological) for recursive self improvement (RSI), i.e. where the parent and not the environment decides what the goal is and measures progress toward it. But as far as I know, the answer is no. Have considered resource constraint situations where parents kill their young? The runt of the litter or, sometimes, others - like when a lion takes over a pride. Mostly in the non-human, non-Chinese portions of the animal kingdom. (I refer to current events re: China's population constraints on female offspring, of course.) There are two problems with this approach. First, if your child is smarter than you, how would you know? Second, this approach favors parents who don't kill their children. How do you prevent this trait from evolving? Secondly, I'm still wondering about the representations of goals in the brain. So far, there has been no study showing the neurobiological basis of 'goal' in the human brain. As far as we know, it's folk psychology anyway, and it might not be 'true', since there's no hard physical evidence of the existence of goals. I'm talking about bottom-up existence, not top-down (top being us, humans and our social contexts and such). You can define an algorithm as goal-oriented if it can be described as having a utility function U(x): X - R (any input, real-valued output) and an iterative search over x in X such that U(x) increases over time. Whether a program has a goal depends on how you describe it. For example, linear regression has the goal of finding m and b such the straight line equation (y = mx + b) minimizes RMS error given a set of (x,y) points, but only if you solve it by iteratively adjusting m and b and evaluating the error, rather than use the conventional closed form solution. The human brain is easiest to describe as having a utility function described by Maslow's hierarchy of needs. Or you could describe it as a state table with 2^(10^15) inputs. Does RSI have to be measured with respect to goals? Can you prove to me that there exists no non-goal oriented improvement methodology? No, it is a philosophical question. What do you mean by improvement? -- Matt Mahoney, [EMAIL PROTECTED] --- agi Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/ Modify Your Subscription: https://www.listbox.com/member/?member_id=8660244id_secret=114414975-3c8e69 Powered by Listbox: http://www.listbox.com