If we've learned anything from science, it is that opposing sides often turn out to belong to the same coin. Heisenberg and Einstein spring to mind. So does classical science and quantum mechanics, eventually joining hands in M-theory, and in doing so, allowing Newton to come sit among the stars. Whatever the factors for decoherence might be, we should be reminded that we only observe what seems apparent to us. Thus, paradigms do shift and convergence remains a fact of technology. How about a quantum leap then?
In such a spirit of inclusivity then, let those who have to perform the "classical" Turing Test do so for its progressive purposes, but at the same time, why not design an appropriate AGI test that already assumes that a single coin with different sides already existed from the outset? Sorry, but even if I understand the point of it all, I see no real AGI future in allowing machines to battle it out, with a human being as the arbitrator to decide the suitability of the questions. It is reminiscent of robot wars. >From my understanding that is not how the visual test is supposed to be >conducted. The machine is supposed to identify rich meaning as defined by >humans, i.e., from a photographic context presented to it, e.g., 2 persons >walking in a busy, city street holding hands and apparently talking. It is >only if their hands were obscured in the photograph by their bodies, and the >question was asked: "What are they carrying?" that the arbitrator would rule" >"Fowl!" in favour of the machine. I would probably vote the same in favour of >a human participant too. But still, would the machine fail the imitation test >if it answered simply: "I don't know."? I'm concerned that at the rate the "classical" test is progressing, we'll even be missing Turing's imagined deadline. Society has already achieved another prediction of Turing's, which was that one would be able to talk about machine intelligence in public without encountering, general denial of such a possibility. In other words, society probably is ready now for pure AGI. My contention is that the world needs AGI machines that could make a constructive difference at large scale and rather sooner than later. For example, how about independently changing saline water into drinking water (with added test parameters for intelligence off course)? A useful test. Or perhaps, using adaptive, electromagnetic frequency to successfully treat a basic set of machine-diagnosed illnesses with? Another useful test. Less subjective too. Such tests, if successful, should prove the machine as a new species of machine, as pure AGI, not an attempt at replicating human functionality alone. Let's apply Toffler's principle of; Technology feeding upon itself. Thus, let the Turing-Test bar for AGI be set in such a manner as to allow quantum-based technology to have a 7-course meal. Perhaps, as a suggestion, as a start, why not just get on with establishing an adapted philosophical basis for a futuristic Turing Test? Why not start by extending the notion of Turing's "imitate". Suppose we agreed Turing's non-classical version of "imitate" to mean; AGI imitation at a quantum level, in order to solve problems by learning from environment, matter, biology and humans and be applied to environment, matter, biology and humans in an adaptive manner, as an independent, non-human-cell entity, as a purely intelligent machine, to the objective satisfaction of humans? Other than designing and emerging such a machine, remove all human control - and thus contest - from the actual test. The fundamental difference to the "classical" test would be the inherent assumptions that: 1) Machines could be AGI enabled; 2) AGI machines are not in a contest with what makes humans human. Set the bar so high the following generations of scientists and hackers and informed laymen and women may dream. Rob Date: Fri, 13 Mar 2015 10:14:50 -0700 Subject: Re: [agi] visual turing test From: a...@listbox.com To: a...@listbox.com The point of this is just that it's just upping the bar for machine learning contests, because the last challenges have been met. They are taking the previous narrow AI challenge and getting a little bolder, broadening it a bit to something closer to AGI but still potentially achievable by narrow AI methods. On Fri, Mar 13, 2015 at 7:30 AM, Benjamin Kapp via AGI <a...@listbox.com> wrote: This isn't a turing test because you aren't playing the imitation game. The computer isn't trying to convince a human it is a human. The test has a computer asking the questions to another computer. The humans role is to simply classify questions as "unanswerable" or not. The paper is just a piece which seeks to exploit the prestige of the researchers universities to raise the bar for image recognition in the field of image recognition NOT AGI. An AGI test would NOT say make a better domain specific algorithm. An AGI test would ask for a cross domain algorithm. E.g. create a program that can both beat someone at chess and write a poem using the same general purpose algorithm(s). Does this make sense? On Fri, Mar 13, 2015 at 2:34 AM, Nanograte Knowledge Technologies via AGI <a...@listbox.com> wrote: 65 years on, and we're still trying to prove the unprovable A<->B. Turing never suggested a visual test. This proposal is merely an interpretation of the original Turing Test. It promotes the development of human intelligence, not machine intelligence. Therefore, my contention would be that such a test be viewed as valid in its intent, but not reliable in its AGI philosophy. Seems, we're still missing the point of the original test, which is also referred to as a philosophical underpining of AI. Is there any other AGI philosophy? Any philosophical extensions to Turing? If not, we'll probably remain stuck there then. <I think this is what Ben's been on about for ages now.> For AGI purposes, let's then rather revisit the philosophy of AI, as offered by Turing, and extend it into the now-emerging future. What would such a test look like in quantum-mechanical terms? I think, that would be appropriate for raising the AGI bar. Rob > Date: Thu, 12 Mar 2015 11:35:25 -0700 > Subject: Re: [agi] visual turing test > From: a...@listbox.com > To: a...@listbox.com > > I skimmed over the article. It sounds like it pretty much IS a Turing > test. They are just asking more detailed questions about what is in a > picture to check to see if the machine "understands." Their > motivation is apparently that the visual testing is inadequate. > (Maybe I missed something) > > On 3/12/15, Nanograte Knowledge Technologies via AGI <a...@listbox.com> wrote: > > It does not matter how sophisticated the test is. Until we turn Turing on > > it's head, the test would still return a value of 1. The notion that a > > machine = human is outdated. Why try and prove it? Therefore, the "machine" > > has become but a catalyst for human development. I still contend that Turing > > had a different message for the world and that we may be missing it. > > > > From: a...@listbox.com > > To: a...@listbox.com > > Subject: [agi] visual turing test > > Date: Wed, 11 Mar 2015 19:34:34 +0100 > > > > http://machineslikeus.com/news/researchers-develop-visual-turing-test > > > > > > > > > > > > AGI | Archives > > > > | Modify > > Your Subscription > > > > > > > > > > > > > > > > > > > > ------------------------------------------- > > AGI > > Archives: https://www.listbox.com/member/archive/303/=now > > RSS Feed: https://www.listbox.com/member/archive/rss/303/11943661-d9279dae > > 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/303/26941503-0abb15dc > Modify Your Subscription: https://www.listbox.com/member/?&; > Powered by Listbox: http://www.listbox.com AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription 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
