Logan: have you ever programmed a robot? You have to measure the distance to the wall so you don't walk into it
Logan, Think carefully about your assumptions here. You’re assuming that a robot must be programmed as robots have always been. And if a normal robot is programmed to walk to a given goal, the programmer may indeed measure or plot the distance and route to the goal. That is the normal practice. And reasonable practice. IF you want to keep producing NARROW AI robots. You are actually basing everything on **narrow AI** assumptions (just as Ben’s & Jim’s concurrent thread is based). But we want an AGI ROBOT that can conduct activities like animals and humans – that can walk down a field or street just as YOU do - something that no robot has ever done before. Now consider how you actually walk down a new field or a new street. Do you first “measure the distance to the end of the field/street”? **Before** you walk down the field? That’s physically impossible isn’t it? (In a normal situation). And in a sense it’s physically impossible for a narrow AI robot too. It wasn’t actually the robot that measured the distance to the wall or goal – it was the PROGRAMMER. AGI is about creating courses of action – new courses of action - walking down a new field of whatever description – physical or metaphorical - that *can’t* be measured or plotted in advance. And for that maths/measurement simply doesn’t apply - at least not in any necessary way. Programs without maths are not only possible, they are essential here. Any program here can only, essentially, tell the robot to head for the goal, put one foot in front of the other, and hope for the best. Because you can’t know for sure what lies ahead in a new field – let alone measure it or the steps that must be taken. Narrow AI is about getting machines to take old journeys in old fields, that the programmer has already taken on behalf the machine – before it moves a metal muscle - and that the programmer knows exactly how to take. AGI is about getting machines to take new journeys in new fields, that robot and programmer alike *haven’t** already taken – *and don’t know exactly how to take.** (or measure). Nobody in AGI gets the distinction. From: Logan Streondj Sent: Tuesday, January 01, 2013 7:45 PM To: AGI Subject: Re: [agi] Why Logic & Maths Have Sweet FA to do with Real world reasoning On Tue, Jan 1, 2013 at 5:15 AM, Mike Tintner <[email protected]> wrote: Logan:I simply said that math was necessary for programming to work Really? You are saying that a robot can’t take steps to a goal – walk across a room or field – without some kind of counting or numbers being involved? certainly! have you ever programmed a robot? You have to measure the distance to the wall so you don't walk into it. Also assuming it has legs has to calculate step length so it doesn't exceed the amount of space available. Sure when you walk, you don't explicitly count it in mm or w/e, but you do implicitly based on measuring the amount of visible space, much of course is done by lower brain regions which are out of the way of conscious thinking. That – wh. is more or less what David talks about - a robot “taking steps to a goal” – is a good v. general way to think about both the final function of programming and AGI. Why do those steps have to involve maths? even version increments involve counting. I use a hexadecimal increment system in my roadmap. GIT uses sha hashes for versioning, which is a more complicated numbering system, that uses more advanced math functionality. (There does have to be some sense of quantities – for example, of putting more or less effort into those steps – but again why does that quantitative sense have to be precisely mathematical rather than crudely emotional? Emotions are for making the actual decisions, whereas math helps quantify the options, allowing for smarter decisions, which may lead to more positive emotions. When you do pressups, do you think your system is performing mathematical calculations of effort – or is your sense of pain rather something very crudely and imperfectly fluidly quantitative? After all, your system doesn’t actually know its precise limits – how can they be quantified?) sure they can be quantified, with kg's and things like that. A healthy vertabrate can on average safely lift and carry about 25% of their body weight for prolonged periods of time. Though potentially a 100% or more for short intervals. If getting groceries from the store, I often at least make rough calculations of how many kg I'm getting, as I carry the food in my backpack, and if I'm walking it could be half an hour of carrying or more. It can be very grueling to carry too much, so I like to be able to estimate with kg and know how much is safe. AGI | Archives | Modify Your Subscription ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
