tag:blogger.com,1999:blog-60380902008-08-25T16:56:53.038-07:00AI DevelopmentDennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comBlogger128125tag:blogger.com,1999:blog-6038090.post-54014657334626847342008-08-25T16:15:00.000-07:002008-08-25T16:56:53.059-07:00Narrow AI in PostJobFree.comI strongly believe that the best way to AGI (Artificial General Intelligence) is building narrow AI and then gradually extend it toward more and more General Intelligence.<br /><br />Finally, I implemented some of my AI techniques in real-life web site <a href="http://www.postjobfree.com">PostJobFree.com</a>.<br />Now PostJobFree.com intelligently calculates <a href="http://postjobfree.blogspot.com/2008/08/job-posting-limit.html">Daily Job Posting Limit</a>. The calculations are based on how many times recruiter's postings were viewed, and how many times these postings were reported as spam.<br />I cannot claim that this feature has "advanced intelligence", but it is intelligent nevertheless.<br /><br />Here are intelligent techniques we used to build that feature:<br /><br />1) Preprocessing data prior to using it in decision making.<br />Row data is coming in the form of "page views" and "spam report clicks".<br />Special process raw input into RecruiterRating and JobRating tables.<br /><br />2) <a href="http://www.dennisgorelik.com/ai/Forgetting.htm">Forgetting</a>.<br />The most recent data is usually more valuable for decision making.<br />That's why yet another PostJobFree process makes sure that old data is slowly losing it's value (and disappears if the value is too low).<br />We implemented it by simply decreasing values in some columns in RecruiterRating and JobRating tables by 1% every night.<br /><br /><br />Here's what I've learned from implementing my first real-life intelligent feature:<br />1) The best working formulas and algorithms are relatively simple.<br />2) Still it takes time to carefully propose, test, chose, and implement intelligent algorithm.<br />3) If the system is designed properly - performance is not an issue.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-70110517567060841202008-05-14T15:50:00.000-07:002008-05-14T15:53:05.343-07:00Artificial General Intelligence projectFunny quote from AGI mailing list:<br /><br />=======<br />Vladimir Nesov wrote:<br />> On Tue, Mar 11, 2008 at 7:20 AM, Linas Vepstas <linasvepstas@gmail.com> wrote:<br /><br />Linas Vepstas: How about joining effort with one of the existing AGI projects?<br /><br />Vladimir Nesov: "They are all hopeless, of course. That's what every AGI researcher<br /> will tell you... ;-)"<br /><br />Richard Loosemore: "Oh no: what every AGI researcher will tell you is that every project is hopeless EXCEPT one. ;-)"<br />=======Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-22660017947868750892008-02-09T16:10:00.000-08:002008-02-09T17:03:26.290-08:00How do we learnMark Gluck gives an interesting explanation about cognitive processes in human brain:<br /><a href="http://www.youtube.com/watch?v=2Ei6wFJ9kCc">The Cognitive and Computational Neuroscience...</a><br /><br />Mark explains that we learn both from observation and from experiment.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-48903224179940374012007-12-07T11:11:00.000-08:002007-12-07T17:02:00.250-08:00Reducing AGI complexity: copy only high level brain designIn my previous post <a href="http://www.dennisgorelik.com/ai/2007/12/complexity-and-incremental-agi-design.html">Complexity and incremental AGI design</a> I claim that complexity has very serious impact on AGI development.<br />If we want to improve our chances of successful AGI implementation, we need to cut complexity as much as possible.<br />In this post I want to touch the topic of copying human brain design while developing AGI.<br />Human brain structure is very complex it's almost impossible to describe in details how exactly brain works.<br />Richard Loosemore explains why this is the case:<blockquote>Imagine that we got a bunch of computers and connected them with a network that allowed each one to talk to (say) the ten nearest machines.<br /><br />Imagine that each one is running a very simple program: it keeps a handful of local parameters (U, V, W, X, Y) and it updates the values of its own parameters according to what the neighboring machines are doing with their parameters.<br /><br />How does it do the updating? Well, imagine some really messy and bizarre algorithm that involves looking at the neighbors' values, then using them to cross reference each other, and introduce delays and gradients and stuff.<br /><br />On the face of it, you might think that the result will be that the U V W X Y values just show a random sequence of fluctuations.<br /><br />Well, we know two things about such a system.<br /><br />1) Experience tells us that even though some systems like that are just random mush, there are some (a noticeably large number in fact) that have overall behavior that shows 'regularities'. For example, much to our surprise we might see waves in the U values. And every time two waves hit each other, a vortex is created for exactly 20 minutes, then it stops. I am making this up, but that is the kind of thing that could happen.<br /><br />2) The algorithm is so messy that we cannot do any math to analyze and predict the behavior of the system. All we can do is say that we have absolutely no techniques that will allow us to mathematical progress on the problem today, and we do not know if at ANY time in future history there will be a mathematics that will cope with this system.<br /><br />What this means is that the waves and vortices we observed cannot be "explained" in the normal way. We see them happening, but we do not know why they do. The bizarre algorithm is the "low level mechanism" and the waves and vortices are the "high level behavior", and when I say there is a "Global-Local Disconnect" in this system, all I mean is that we are completely stuck when it comes to explaining the high level in terms of the low level.<br /><br />Believe me, it is childishly easy to write down equations/algorithms for a system like this that are so profoundly intractable that no mathematician would even think of touching them. You have to trust me on this. Call your local Math department at Harvard or somewhere, and check with them if you like.<br /><br />As soon as the equations involve funky little dependencies such as:<br /><br />"Pick two neighbors at random, then pick two parameters at random from each of these, and for the next day try to make one of my parameters (chosen at random, again) follow the average of those two as they were exactly 20 minutes ago, EXCEPT when neighbors 5 and 7 both show the same value of the V parameter, in which case drop this algorithm for the rest of the day and instead follow the substitute algorithm B...."<br /><br />Now, this set of computers would be a wicked example of a complex system, even while the biggest supercomputer in the world, following a nice, well behaved algorithm, would not be complex at all.<br /><br />The summary of this is as follows: there are some systems in which the interaction of the components are such that we must effectively declare that NO THEORY exists that would enable us to predict certain global regularities observed in these systems.<br /></blockquote><br /><br />So, if low level brain design is incredibly complex - how do we copy it?<br /><br />The answer is: "we don't copy low level brain design".<br />Low level design is not critical for AGI. Instead we observe high level brain patterns and try to implement them on top of our own, more understandable, low level design.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-89933166105984494282007-12-07T10:42:00.001-08:002007-12-07T11:08:54.281-08:00Complexity and incremental AGI designWhy is it so hard to build Artificial General Intelligence (AGI)?<br />It seems we have almost everything we need: great hardware, mature software development industry, Internet, Google, lots of successful narrow AI project ... but AGI is still to hard to crack.<br /><br />The major reason is -- overall complexity of building AGI.<br /><br />Richard Loosemore is writing about it: <blockquote>Do we suspect that complexity is involved in intelligence? I could present lots of reasoning here, but instead I will resort to quoting Ben Goertzel: "There is no doubt that complexity, in the sense typically used in dynamical-systems-theory, presents a major issue for AGI systems"<br />Can I take it as understood that this is accepted, and move on?<br />So, yes, there is evidence that complexity is involved.</blockquote><br /><br />Richard also explains, how exactly complexity affects system development: <blockquote>when you examine the way that complexity has an effect on systems, you find that it can have very quiet, subtle effects that do not jump right out at you and say "HERE I AM!", but they just lurk in the background and make it quietly impossible for you to get the system up above a certain level of functioning. To be more specific: when you really allow the symbol-building mechanisms, and the learning mechanisms, and the inference-control mechanisms to do their thing in a full scale system, the effects of tiny bits of complexity in the underlying design CAN have a huge impact. One particular design choice, for example, could mean the difference between a system that looks like it ought to work, but when you set it running autonomously it gradually drifts into imbecility without there being any clear reason. </blockquote><br /><br />The is a good technique of dealing with complex system -- increase complexity gradually and carefully test every step.<br />That's why I think it's so important to build testable narrow AI systems prior to building AGI.<br />We have many Narrow Artificial Intelligent Systems already, but we need more. And we need them to become more advanced up to the point when they become AGI.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-88683155016277688392007-05-01T19:14:00.000-07:002007-05-02T15:33:32.307-07:00Self-emergence of intelligence in humans and artificial systemsHuman brain is self-emergent on many levels. Here's simplified sequence of human brain self emergence:<br />1) Human genes build "Brain Builder". Brain Builder consists of:<br />- Neurons Factory – neurons with reproductive ability. <br />- Brain Structure Manager – hormones and other mechanisms that define brain structure.<br /><br />2) Brain builder builds "Empty Brain" --- fully assembled, but mostly empty brain: <a href="http://www.dennisgorelik.com/ai/SuperGoal.htm">super goals</a> are defined, but there is no external knowledge yet, no sub-goals defined yet.<br /><br />3) By experimenting and learning Empty Brain evolves into Brain with Mind (fully working intelligent system, with lots of external knowledge and <a href="http://www.dennisgorelik.com/ai/SubGoal.htm">sub goals</a>).<br /><br />Every step in this sequence means self-emergence.<br /><br />What do you think, when we build artificial intelligent system, what system should we build: Genes, Brain Builder, Empty Brain, or Brain with Mind?<br /><br />I believe that building Empty Brain is our best option.<br />Below are my reasons.<br /><br /><h4>Why not build Brain with Mind?</h4>In order to build Brain with Mind we have to build Empty Brain anyway, but our task will be considerably more complex, because fully loaded mind is at least 10 times more complex than Empty Brain. It's like complexity of empty computer in comparison with complexity of all software which is loaded into regular "in use" computer.<br />Bottom line: there is no point to ai developers to pre-load mind into strong AI, when Empty Brain system can do it itself.<br /><br /><br /><h4>Why not build Brain Builder?</h4>Complexity of Brain Builder is probably comparable with complexity of Empty Brain. But from engineering perspective developing Brain Builder is considerably more complex.<br />1) Let assume that we didn’t have designed Empty Brain yet. In this case we have no clue what the output of our Brain Builder should be. That means that we cannot test or debug Brain Builder. There are no checkpoints to verify that our development is on the right track.<br />Inability to test and debug complex system makes development of such system virtually impossible. <br />The only working approach in this situation would be to try to tweak some Brain Builder’s settings and then run full test: build Empty Brain and wait for several years to check if it evolves into Brain with Mind.<br />Mother Nature was quite efficient in this approach. It took just few billions years to develop proper Brain with Mind. I doubt that human researchers applying such approach would accomplish the task considerably faster.<br /><br />2) Let assume that we already designed working model of Empty Brain. In this case what’s the point to design Brain Builder? Our industry can easily reproduce any working model in mass quantity.<br /><br /><br /><h4>Why not build Genes?</h4>Building Genes which would build Brain Builder is even more complex than building Brain Builder itself.<br />The reasons are the same as in "Why not build Brain Builder?"<br />If we don’t have working model of Brain Builder yet – then we effectively cannot test & debug genes.<br />If we have working model of Brain Builder – then why bother with Genes?<br /><br /><br /><h4>Parallels with existing systems</h4>1) CYC is trying to build Brain with Mind system. Actually even worse – they are trying to build Mind without Brain --- no self-learning ability, no super-goals.<br />That road leads nowhere.<br /><br />2) <a href="www.google.com">Google</a> is Brain with Mind which was developed as Empty Brain. Google's Empty Brain has working crawler and other self-learning mechanisms. This approach proved to be very efficient, and eventually Google's Empty Brain emerged into Brain with Mind – very smart search system.<br /><br />3) It seems that there are no famous Brain Builder projects. But I’m sure that some researchers do attempts to build "Brain Builder". So far – no success at all for the reasons I explained above.<br /><br /><h4>Conclusion</h4>Building Empty Brain capable of self-emerging into fully capable Brain with Mind -- is the most feasible engineering approach in <a href="http://www.dennisgorelik.com/ai/Overview.htm">strong AI development</a>.<br /><br /><br />---<br />This post is a result of discussion with <a href="http://www.2mtheory.com/">David Ashley</a>. He is a proponent of "Brain Builder" approach.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-59957235447667154382007-04-15T11:50:00.000-07:002007-04-15T11:59:34.676-07:00Intelligence: inherited through genes or gained from environment?Human Intelligence is acquired from environment, not encoded genes.<br />Genes provide framework, which allow to learn from environment. This framework is critical for intelligence, but does not provide intelligence by itself.<br /><br />===== By Richard Loosemore (2007 April 05) in AGIRI forum =====<br />If we were aliens, trying to understand a bunch of chess-playing IBM supercomputers that we had just discovered on an expedition to Earth, we might start by noticing that they all had very similar gross wiring patterns, where "gross wiring" just means the power cables, bundles of wires inside each rack, and wires laid down as tracks on circuit boards. <br />But nothing inside the chips themselves, and none of the "soft" wiring that exists in code or memory.<br /><br />Having mapped this stuff, we might be impressed by how very similar the <br />gross wiring pattern was between the different supercomputers that we discovered, and so we might conclude that our discovery represented a significant advance in our understanding of how the machines worked.<br /><br />.....<br /><br />That last bit -- the [powerful algorithms that interact with the environment] bit -- is what makes the difference between a baby that sits there drooling and probing for its mother's nipple, and an adult human being who can understand the complexities of the human cognitive system.<br /><br />Anyone who thinks that that last bit is also encoded in the human genome has got a heck of a lot of work to do ...<br />=====Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-6147963129235453942007-02-20T21:47:00.000-08:002007-02-20T21:59:40.736-08:00Larry Page talks about AI=====<br /><a href="http://news.com.com/2100-11395_3-6160372.html">Google's Page urges scientists to market themselves</a><br />Google co-founder Larry Page has a theory: your DNA is about 600 megabytes compressed, making it smaller than any modern operating system like Linux or Windows. <br />.....<br />"We have some people at Google (who) are really trying to build artificial intelligence and to do it on a large scale," Page said to a packed Hilton ballroom of scientists. "It's not as far off as people think."<br />=====<br /><br />I agree with Larry Page: human's DNA has relatively small size.<br />Besides, not all human DNA is in charge of the brain. I guess that something like 10% of the whole DNA is related to brain development.<br /><br />I wrote about that over 3 years ago:<br />-----<br /><a href="http://dennisgorelik.com/ai/TheTimeHasCome.htm">The time has come The time has come to develop Strong Artificial Intelligence System</a><br />Strong AI project is quite complex software project. However even more complex systems were implemented in the past. Many software projects are more complex than human DNA (note that human DNA contains way more than just genocode for intelligence).<br />-----Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1168164407646118082007-01-07T01:54:00.000-08:002007-01-07T02:06:47.660-08:00Should Strong AI have its own goals?<strong>Short answer: </strong>Yes and No.<br /><strong>Long answer: </strong>Strong AI can add and modify millions of softcoded goals. At the same time <a href="http://www.dennisgorelik.com/ai/StrongAI.htm">Strong AI</a> shouldn't be able to change its own super goals. <br />Why? <br /><br />Here are the reasons:<br /><br />1) In its normal working cycle strong AI modifies <a href="http://www.dennisgorelik.com/ai/SoftcodedGoals.htm">softcoded goals</a> in complience with embedded <a href="http://www.dennisgorelik.com/ai/SuperGoal.htm">super goals</a>. If strong AI has ability to modify super goals then strong AI will modify (or terminate) super goals instead of achieving these goals. <br /><em><strong>Example:</strong><br />Without ability to modify super goal "survive", computer will try to protect itself, will think about power supply, safety and so on.<br />With ability to modify super goals computer would simply terminate goal "survive" and create goal "do nothing" instead just because it's the easiest goal to achieve. Such "do-nothing" goal would result in the death of this computer. </em><br /><br />2) If Strong AI can change its super goals then Strong AI would work for itself instead of working for its creator. Strong AI's behavior would eventually become uncontrollable by AI creator / operator.<br /><br />3) Ability to reprogram its own super goals makes computer behave like a drug addict. <br /><em><strong>Example:</strong><br />Computer can create new super goal for itself: "listen to music" or "roll the dices" or "calculate PI number" or "do nothing". It would result in Strong AI doing useless stuff or simply doing nothing. Final point: uselessness for society and death.</em>Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1154763498947757712006-08-05T00:09:00.000-07:002006-12-13T17:01:16.136-08:00Massive words/phrases database publishes by GoogleGoogle research publishes their massive words/phrases database:<br />===<br /><a href="http://googleresearch.blogspot.com/2006/08/all-our-n-gram-are-belong-to-you.html">All Our N-gram are Belong to You</a><br />We processed 1,011,582,453,213 words of running text and are publishing the counts for all 1,146,580,664 five-word sequences that appear at least 40 times. There are 13,653,070 unique words, after discarding words that appear less than 200 times. <br />Watch for an announcement at the LDC, who will be distributing it soon, and then order your set of 6 DVDs.<br />===<br />This team can be contacted at: ngrams@google.comDennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1149904265665880412006-06-09T18:49:00.000-07:002006-06-09T18:51:05.690-07:00Motivational system1) I agree that direct reward has to be in-built<br />(into brain / AI system).<br />2) I don't see why direct reward cannot be used for rewarding mental<br />achievements. I think that this "direct rewarding mechanism" is<br />preprogrammed in genes and cannot be used directly by mind.<br />This mechanism probably can be cheated to the certain extend by the<br />mind. For example mind can claim that there is mental achievement when<br />actually there is none.<br />That possibility of cheating with rewards is definitely a problem.<br />I think this problem is solved (in human brain) by using only small<br />dozes of "mental rewards".<br />For example, you can get small positive mental rewards by cheating your<br />mind to like finding solutions to "1+1=2" problem.<br />However, if you do it too often you'll eventually get hungry and would<br />get huge negative reward. This negative reward would not just stop you<br />doing "1+1=2" operation over and over, it would also re-setup your<br />judgement mechanism, so you will not consider "1+1=2" problem as an<br />achievement anymore.<br /><br />Also, we all familiar with what "boring" is.<br />When you solve a problem once - it's boring to solve it again.<br />I guess that that is another genetically programmed mechanism with<br />prevents cheating with mental rewards.<br /><br />3) Indirect rewarding mechanisms definitely work too, but they are not<br />sufficient for bootstrapping strong-AI capable system.<br />Consider a baby. She doesn't know why it's good to play (alone or with<br />others). Indirect reward from "childhood playing" will come years later<br />from professional success. <br />Baby cannot understand human language yet, so she cannot envision this<br />success.<br />AI system would face the same problem.<br /><br />My conclusion: indirect reward mechanisms (as you described them) would not be<br />able to bootstrap strong-AI capable system.<br /><br />Back to real baby: typically nobody explains to baby that it's good to play.<br />But somehow babies/children like to play.<br />My conclusion: there are direct reward mechanisms in humans even for<br />things which are not directly beneficial to the system (like mental<br />achievements, speech, physical activity).<br /><br />(from AGI email list).Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1149867765528796702006-06-09T08:40:00.000-07:002006-06-09T08:42:45.543-07:00Richard Loosemore - RewardRichard Loosemore (rpwl at lightlink.com):<br />All thinking systems do have a motivation system of some sort (what you <br />were talking about below as "rewards"), but people's ideas about the <br />design of that motivational system vary widely from the implicit and <br />confused to the detailed and convoluted (but not necessarily less <br />confused).<br />===<br /><br /><a href="http://www.dennisgorelik.com/ai/Reward.htm">Reward</a>Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1134752335355629612005-12-16T08:58:00.000-08:002005-12-16T08:58:55.400-08:00Colloquium on the Law of Transhuman Persons<a href="http://www.imminst.org/forum/index.php?s=&act=ST&f=69&t=7868">Colloquium on the Law of Transhuman Persons</a>
<br />
<br />There are photos here how they disscussed law related to transhumans. Florida's beach pictures included :-)
<br />Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1134709397386797382005-12-15T20:58:00.000-08:002005-12-15T21:03:17.396-08:00How to prevent bad guys from using results of AI reserch?<em>David Sanders> I would like to see a section up on your site about the downsides of AIS and what preventative limits need to take place in research to ensure that AIS come out as the "good" part of humans and not the bad part. The military is already building robotic, self propelled and thinking vehicles with weapons.<br /></em><br />Recipe for "safe from bad guys research" is the same as recipe for any<br />research: openness.<br /><br />When ideas are available for society - many people (and later many<br />machines) would compete in implementation of these ideas. And society<br />(human society / machine society / or mixed society) - would setup<br />rules which would prevent major misuse of new technology.<br /><br /><br /><em>David Sanders> How long do we really have before an AIS, demented or otherwise) decides to eliminate its maker?<br /></em><br />Why would you care?<br />Some children kill their parents. Did our society collapsed because of<br />that?<br /><br />Some AISes would be bad. Bad not just toward humans, but toward other<br />AISes.<br />But as usual --- bad guys wouldn't be a majority.<br /><br /><em>David Sanders> As countless science fiction stories have told us, even the most innocent of actions by an AIS may spell disaster,<br /></em><br />1) These are <strong>fiction</strong> stories.<br />2) Some humans can cause disasters too, so what?<br /><br /><em>David Sanders> because like I said above the don't fundamentally understand us, and we don't understand them.<br /></em><br />Why wouldn't AISes understand humans?<br /><br /><em>David Sanders> We will be two completely different species, and they might not hold the same sanctity of life most of us are born with.<br /></em><br />Humans are not born with sanctity. Humans gain it (or not gain) while<br />they grow.<br />Same would apply to machines.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1134709033965115182005-12-15T20:44:00.000-08:002005-12-15T20:57:13.976-08:00Discussion about AIS weaknessesThis discussion inspired by web-page <a href="http://www.dennisgorelik.com/ai/WeaknessesOfAIS.htm">Weaknesses of AIS</a>.<br /><em></em><br /><em>David Sanders> AIS cannot exist (for now) without humans.<br /><br /></em><em></em>That’s not really a weakness, because time span of this weakness wouldbe pretty short. Right now strong AI systems exist only in our dreams. :-) Within ~20 years of creating strong AI, many AISes would be able to survive without humans. Please, note that AISes would not kill humans. There would be benefits of human-AISes collaboration for all sides. This is completely different topic though. :-)<br /><br /><em>David Sanders> If they fail to understand and appreciate the human world...<br /></em><br />If you don't understand and appreciate human world of Central Africa... would it harm you?<br />May be you mean "If AISes don't understand human world at all"? But in this case what would these AISes understand? And what would mean that these not-understanding systems intelligent?<br /><br /><em></em><em>David Sanders> [AISes] Not able to perceive like a human. They cannot hear, see, feel, taste or smell like a human.<br /></em><br />Not true. Only first and limited versions of AISes wouldn’t be able to perceive like a human. Sensor devices are not too hard to implement. The major problem is implementation of Main Mind for AIS.<br /><br /><em></em><em>David Sanders> They can only feel these things like they imagine they do. Again, this makes them fundamentally incongruous with humans and I don't believe its something you can "teach around." Try to explain what "blue" is to someone who never had sight.<br /></em><br />Have you ever seen "black hole", "conscience", or "electron"? Yet you know what they are, don't you? :-)<br />Blind person can understand what "blue" means: "sky is blue", "water is blue", ...<br /><br /><em>David Sanders> Until AIS have robot bodies / companions, they rely on humans for natural resources. However, once the singularity hits, that probably won't matter anymore. It is not inconceivable to think of a time in 200-500 years there are no more humans, just AIS.<br /></em><br />Humans would probably exist long after strong AI is created. Humans just would not be the most intelligent creatures anymore :-)<br /><br /><em>David Sanders> I disagree with AIS and natural selection. I think this will happen on its own by their very nature.<br /></em><br />AISes can be influenced by natural selection as much as all other living organisms. But humans had millions of years of natural selection. When would AISes have that much?<br /><br /><em></em><em>David Sanders> AIS will be more open about self modification as you point out. AIS will be able to make other AIS and will soon learn how to evolve themselves very quickly.<br /></em><br />"Evolving themselves" is part of artificial selection, not natural selection.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1133207832367331962005-11-28T11:57:00.000-08:002005-11-29T10:54:46.393-08:00Matt Bamberger - Matt Bamberger<a href="http://www.mattbamberger.com/">Matt Bamberger - Matt Bamberger</a><br /><br />Matt worked for Microsoft, tried to retire ... unsuccessfully, so he works again and has extensive software development experience. Matt is interested in AGI (Artificial General Intelligence) and Singularity.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1129740932464815682005-10-19T09:55:00.000-07:002005-10-19T10:07:32.063-07:00An Integrated Self-Aware Cognitive ArchitectureThat looks like a very interesting project in a Strong AI field.<br />Though I (Dennis) personally disagree with couple of basic ideas here.<br />1) It seems that Alexei Samsonovich pays a lot of attention to self-awareness.<br />For me it's not clear why self-awareness is more important than awareness about surrounding world in general.<br />2) Another questionable thing is about AI being autonomous.<br />As far as I know, there is no intelligent system which is autonomous from the society. Human's baby would never become intelligent without society.<br />In order to make AI system intelligent, Alexei Samsonovich would have to connect the system to <a href="http://www.dennisgorelik.com/ai/AISAndSociety.htm">society</a> somehow. For example through the Internet. <br /><br />Anyway, the following looks like great AI project.<br />You may want to try to take part in it.<br /><br />From: Alexei V Samsonovich <table><tr><td>samsonovich</td><td>@</td><td>cox.net</td></tr></table><br />Date: Tue, 18 Oct 2005 06:02:46 -0400<br />Subject: GRA positions available<br /><br />Dear Colleague:<br /><br />As a part of a research team at KIAS (GMU, Fairfax, VA), I am searching <br />for graduate students who are interested in working during one year, <br />starting immediately, on a very ambitious project supported by our <br />recently funded DARPA grant. The title is "An Integrated Self-Aware <br />Cognitive Architecture". The grant may be extended for the following <br />years. The objective is to create a self-aware, conscious entity in a <br />computer. This entity is expected to be capable of autonomous cognitive <br />growth, basic human-like behavior, and the key human abilities including <br />learning, imagery, social interactions and emotions. The agent should be <br />able to learn autonomously in a broad range of real-world paradigms. <br />During the first year, the official goal is to design the architecture, <br />but we are planning implementation experiments as well.<br /><br />We are currently looking for several students. The available positions <br />must be filled as soon as possible, but no later than by the beginning <br />of the Spring 2006 semester. Specifically, we are looking for a student <br />to work on the symbolic part of the project and a student to work on the <br />neuromorphic part, as explained below.<br /><br />A symbolic student must have a strong background in computer science, <br />plus a strong interest and an ambition toward creating a model of the <br />human mind. The task will be to design and to implement the core <br />architecture, while testing its conceptual framework on selected <br />practically interesting paradigms, and to integrate it with the <br />neuromorphic component. Specific background and experience in one of the <br />following areas is desirable: (1) cognitive architectures / intelligent <br />agent design; (2) computational linguistics / natural language <br />understanding; (3) hacking / phishing / network intrusion detection; (4) <br />advanced robotics / computer-human interface.<br /><br />A neuromorphic candidate is expected to have a minimal background in one <br />of the following three fields. (1) Modern cognitive neuropsychology, <br />including, in particular, episodic and semantic memory, theory-of-mind, <br />the self and emotion studies, familiarity with functional neuroanatomy, <br />functional brain imaging data, cognitive-psychological models of memory <br />and attention. (2) Behavioral / system-level / computational <br />neuroscience. (3) Attractor neural network theory and computational <br />modeling. With a background in one of the fields, the student must be <br />willing to learn the other two fields, as the task will be to put them <br />together in a neuromorphic hybrid architecture design (that will also <br />include the symbolic core) and to map the result onto the human brain.<br /><br />Not to mention that all candidates are expected to be interested in the <br />modern problem of consciousness, willing to learn new paradigms of <br />research, and committed to success of the team. Given the circumstances, <br />however, we do not expect all conditions listed above to be met. Our <br />minimal criterion is the excitement and the desire of an applicant to <br />build an artificial mind. I should add that this bold and seemingly <br />risky project provides a unique in the world opportunity to engage with <br />emergent, revolutionary activity that may change our lives.<br /><br />Cordially,<br />Alexei Samsonovich<br /><br />-- <br />Alexei V Samsonovich, Ph.D.<br />George Mason University at Fairfax VA<br />703-993-4385 (o), 703-447-8032 (c)<br /><a href="http://mason.gmu.edu/~asamsono/">Alexei V Samsonovich web site</a>Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1127437205451179602005-09-22T18:00:00.000-07:002005-09-22T18:00:05.483-07:00Lies, Damned Lies, Statistics, and Probability of Abiogenesis Calculations<a href="http://www.talkorigins.org/faqs/abioprob/abioprob.htm">Abiogenesis</a> - how the life self-formed.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1123881811650787972005-08-12T14:23:00.000-07:002005-08-12T14:23:31.716-07:00Wired 13.08: The Birth of Google<a href="http://www.wired.com/wired/archive/13.08/battelle.html?tw=wn_tophead_4">Wired 13.08: The Birth of Google</a>
<br />It began with an argument. When he first met Larry Page in the summer of 1995, Sergey Brin was a second-year grad student in the computer science department at Stanford University.....Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1122223014777759202005-07-24T08:22:00.000-07:002005-07-24T09:36:54.826-07:00Supergoals<h3>Anti-goals</h3><i>I cannot find it now on your site, but, it seems your system has or will have the opposites to goals (was it goals with negative desirability?)</i><br /><br /><b>Answer:</b>In general, same <a href="http://www.dennisgorelik.com/ai/SuperGoal.htm">supergoal</a> works in both negative and positive directions.<br />Super goal can give both positive and negative reward to the same concept.<br /><b><i>For example</i></b>, supergoal "Want more money" could give negative reward to "Buy Google stock" concept, responsible for investment money into Google stock, because it caused money spending. One year later same "Want more money" supergoal may give positive reward to the same "Buy Google stock" concept, because this investment made the system richer.<br /><br /><h3>Supergoal: "can act" or "state only"? </h3>Supergoals can act. Supergoal actions are about modification of <a href="http://www.dennisgorelik.com/ai/SoftcodedGoals.htm">softcoded goals</a>.<br />Usually Supergoal has state. Typically supergoal state keeps information about <a href="http://www.dennisgorelik.com/ai/SatisfactionLevel.htm">supergoal satisfaction level</a> is at this moment. Supergoal may be stateless too.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1122013621894559022005-07-21T23:19:00.000-07:002005-07-21T23:27:01.900-07:00Glue for the system<i>it seems to me, that you use cause-effect relations as a glue to put concepts together, so they form a connected knowledge; is it the only glue your system has?</i><br /><br />Yes, correct: <a href="http://www.dennisgorelik.com/ai/CauseEffectRelation.htm">cause-effect relations</a> are the only glue to put <a href="http://www.dennisgorelik.com/ai/Concept.htm">concepts</a> together.<br />I decided to have one type of glue instead of many types of glue.<br />It's easier to work with one type of glue.<br /><br />At the same time I have something else that you may <br />consider a glue for the whole system:<br />1) <a href="http://www.dennisgorelik.com/ai/DesirabilityAttribute.htm">Desirability attributes</a> (<a href="http://www.dennisgorelik.com/ai/SoftcodedGoals.htm">softcoded goals</a>)- keep information about system's priorities.<br />2) <a href="http://www.dennisgorelik.com/ai/HardcodedUnits.htm">Hardcoded units</a> - connect concepts to the <a href="http://www.dennisgorelik.com/ai/World.htm">real world</a>. <a href="http://www.dennisgorelik.com/ai/SuperGoal.htm">Super goals</a> are the special subset of these hardcoded units.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1121736930067810272005-07-18T18:24:00.000-07:002005-11-10T10:04:00.223-08:00<h2>What AI ideas has Google introduced?</h2>Google not introduced, but practically demonstrated the following ideas:<br /><br />1) Words are the smallest units of intelligent information. <a href="http://www.dennisgorelik.com/ai/Word.htm">Word</a> alone has meaning. Letter alone - doesn't. Google searches for words as a whole. Not for letters of substrings.<br /><br />2) <a href="http://www.dennisgorelik.com/ai/Phrase.htm">Phrases</a> are important units of information too. Google underlines importance of phrases by supporting search in quotes, like "test phrase".<br /><br />3) Natural language (plain text) is the best way to share knowledge between <a href="http://www.dennisgorelik.com/ai/IntelligentSystem.htm">intelligent systems </a>(people and computers).<br /><br />4) Programming languages that are the best for mainstream programming - the same languages are the best for intelligent system development. LISP, Prolog, and other artificial programming languages are less efficient in intelligence development than mainstream languages like C/C++/C#/VB/: (Google proved this idea by using plain C as a core language for "advanced text manipulation project".<br /><br />5) Huge knowledge base does matter for intelligence. Google underlines importance of huge knowledge base.<br /><br />6) Simplicity of knowledge base structure does matter. In comparison with CYC's model, Google's model is relatively simple. Obviously Google is more efficient/intelligent than dead CYC.<br /><br />7) Intelligent system must collect data automatically (by itself, like in Google's crawler). Intelligent system should not expect to be manually fed by developers (like in CYC).<br /><br />8) To improve information quality, intelligent system should collect information from different types of sources. Google collects web pages from web, but also it collects information from Google toolbar - about what web pages are popular among users.<br /><br />9) Constant updates and <a href="http://www.dennisgorelik.com/ai/Forgetting.htm">forgetting</a> keeps intelligent system sane (Google constantly crawls the Web, adds new and deletes dead web-pages from its memory).<br /><br />10) Links (<a href="http://www.dennisgorelik.com/ai/Relation.htm">relations</a>) add intelligence to a knowledge base (Search engines made the Web mode intelligent);<br />Good links convert knowledge base into intelligent system (Google's index with web work as a very wise adviser (read: intelligent system)).<br /><br />11) Links must have weights (like in Google's Page rank). These weights must be taken into consideration in decision making.<br /><br />12) Couple of talented researchers can do far more than lots of money in wrong hands. Think about "'Serge Brin & Larry Page search' vs 'Microsoft's search'".<br /><br />13) Sharing ideas with public helps research project to come to production. Hiding ideas - kills the project in the cradle. Google is very open about its technology. And very successful.<br /><br />14) Targeting practical results helps research project a lot. Instead of having "abstract research about search", Google targeted "advanced web-search". Criteria of success of the project were clearly defined. As a result Google project quickly hit production and generated tremendous outcome in many ways.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1121664649020264792005-07-17T22:27:00.000-07:002005-07-17T22:30:49.020-07:00<h2>How does strong AI schedule super goals?</h2>Strong AI doesn't schedule <a href="http://www.dennisgorelik.com/ai/SuperGoal.htm">super goals</a> directly. Instead <a href="http://www.dennisgorelik.com/ai/StrongAI.htm">strong AI</a> schedules <a href="http://www.dennisgorelik.com/ai/SoftcodedGoals.htm">softcoded goals</a>. To be more exact, super goals schedule softcoded goals by making them more/less desirable (see <a href="http://www.dennisgorelik.com/ai/RewardDistributionRoutine.htm">Reward distribution routine</a>). The more desirable softcoded goal is – the higher probability is that this softcoded goal will be <a href="http://www.dennisgorelik.com/ai/Activate.htm">activated</a> and executed.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1121664438805725102005-07-17T22:23:00.000-07:002005-07-17T22:27:18.810-07:00<h2>How strong AI finds a way to satisfy super goal</h2><br />The idea is simple: whatever satisfies <a href="http://www.dennisgorelik.com/ai/SuperGoal.htm">super goal</a> now -- most probably would satisfy the super goal in the future. In order to apply this idea, super goals must be programmed in a certain way. Every super goal itself must be able to distinguish what is good and what is bad.<br />Such approach makes super goal kind of "advanced sensor".<br />Actually not only "advanced sensor", but also "desire enforcer".<br /><br /><b>Here's the example how it works</b>:<br />Super goal’s objective: to be rich.<br />Super goal sensor implementation: check strong AI’s bank account for amount of money on it.<br />Super goal enforce mechanism: mark every concept which causes increasing the bank account balance as "desirable". Mark every concept which causes decreasing the bank account balance as "not-desirable".<br /><br />Note: "mark concept as desirable/undesirable" doesn't really work in "black & white" mode. Subtle super goal enforcement mechanism either increases or decreases desirability of every <a href="http://www.dennisgorelik.com/ai/CauseConcept.htm">cause concept</a> affecting the bank account balance.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.comtag:blogger.com,1999:blog-6038090.post-1121629083167221602005-07-17T11:46:00.000-07:002005-07-17T18:32:59.530-07:00Concept type<i>Your concepts have types: word, phrase, simple concept and periheral device. What is a logic behind having these types? </i><br />In fact "peripheral device" is not just one type. There could be many <a href="http://www.dennisgorelik.com/ai/PeripheralDevices.htm">peripheral devices</a>. <br />Peripheral device is a subset of <a href="http://www.dennisgorelik.com/ai/HardcodedUnits.htm">hardcoded units</a><br />Concept can be of any hardcoded unit type.<br />Moreover, one hardcoded unit can be related to concepts of several types. <br />For example: text parser has direct relations with concept-words and concept-phrases. (Please don't confuse these "direct relations" with <a href="http://www.dennisgorelik.com/ai/Relation.htm">relations in the main memory</a>). <br />Ok, now we see that <a href="http://www.dennisgorelik.com/ai/StrongAI.htm">strong AI</a> has many concept types. How many? As many as AI software developer code in hardcoded units. 5-10 concept types is a good start for strong AI prototype. 100 concept types is probably good number for real life strong AI. 1000 concept types is probably too many.<br /><br />So, what is a "concept type"? Concept type is just a reference from concept to hardcoded unit. Concept type is a reference from <a href="http://www.dennisgorelik.com/ai/Concept.htm">concept</a> to <a href="http://www.dennisgorelik.com/ai/World.htm">real world</a> through a <a href="http://www.dennisgorelik.com/ai/HardcodedUnits.htm">hardcoded unit</a>.<br /><br /><i>What concept types shold be added to strong AI?</i><br />If AI developer feels that concept type XYZ is useful for strong AI...<br />and if the AI developer can code this XYZ concept type in hardcoded unit...<br />and if this functionality is not implemented in other hardcoded unit yet...<br />and the <a href="http://www.dennisgorelik.com/ai/MainMemory.htm">main memory</a> structure doesn't have to be modified to accomodate this new concept type...<br />then the developer may add this XYZ concept type to strong AI.<br /><br /><i>What concept types should not be added?</i><br />- I feel that such concept types as "verb" and "noun" should not be added, because there is no clear algorithm to distinguish between verbs and nouns.<br />- I feel that "property concept type" should not be used, because "property concept type" is already covered by "cause-effect relationships" and because implementation of property type concepts will make main memory structure more complex.Dennis Gorelikhttp://www.blogger.com/profile/17700219093521377626noreply@blogger.com