I actually appreciate your skepticism, because it helps me analyze and improve my position.
I will indeed go for the Turing competition eventually. But you should read what I said more carefully. I did not claim to be running everything integrated and fully functional today. This is an enormously complex system, and the pieces to be integrated are not conventional code, they are theory implemented in NNs. That is a lot of work. But I have significant pieces right now that do what no one else currently can do, and chunks planned but not implemented from the underlying theory yet, and years more to go within my development. But enough done that it's clear it's all on target. I said a time frame for human-like AI lies within Kurzweil's prediction. And I expect to have major pieces pulled together within two and certainly five years.
Also, it seems you're thinking conventionally. You call the AI a program. That shows you don't have the right model. Although some people run AI routines in software, full strong AI is nothing like conventional software. My AI is not an AI program, a piece of code. It is AI emerging from neural nets, and some new concepts, under the right machine architecture. That's nowhere the same thing as a program in C++ or LISP. Very different. My current software runs low level code that merely runs NNs on a Von Neumann platform, which is simply not optimal for MPP. The intelligence comes from the way the NNs connect, how they learn, what they learn, and how they control the dynamics of their organization. It is a layer above the code and is nothing like conventional programming, and I admit hard to understand.
Hahhaha. Sorry. Don't be too alarmed. I envision strong AI as a tool for humanity. Smart systems can help and advise the average person about things in their life just as a human friend would. I've mapped out implementing personal advisors for mundane things like cooking, health, personal relationships, learning or teaching, child-raising. Tools that can help writers create stories, movies. Intelligent characters in games. I envision recreating the personalities of dead actors and bringing them back as AIs. Groucho Marx is one of my target projects; I have AI that understands humor. If enough people ask, I'll sketch out how it works. I guarantee I'm not unhinged, nor so naive I'm overestimating what's achievable. Amazing stuff to come, within sight.
Go read any of my comments on Slashdot for the last five years. If I were crazy, stupid, a conman, or a grandiose phony lying egotist, it would have shown before this. Read my January 20 AI comment; I explained then some of my work. It's real; I didn't just make it up for today's thread.
As for Skynet.. hmm. Great idea! Killer robots destroying humanity? I...I didn't think of that! Cough.
I fully understand your skepticism. I've been working on this since 1983 though, and it's all full grounded in strong theory. And I have some things I haven't mentioned yet because you wouldn't believe I can do them, without seeing them. if you think the preceding was grandiose claims, you ought to see what I have in the magic basket to come. But I haven't published because I didn't want to give away the show, or have to go back and correct early work as being in error, hurting credibility. I don't want to just hand 24 years of work to Microsoft for free. Once I have enough essential elements patented, I'll open things, as a coherent and viable alternative to current directions in the field. When I bring this all to market, it will open a whole new area of software. But it is not Open Source. If someone wants to build AIs, let them spend 24 years of their time, or license what I've spent a huge chunk of life on.
There are people of questionable validity in the field. For instance, John LaMuth has a patent on implementing emotion in software. His patent is full of handwaving - in one part he points to an expert system and says in effect "oh, the logic is all in that black box." But never explains anything about what is inside. I on the other hand have a legitimate and working alternative to both LaMuth and the more widely accepted OCC model of emotion in computing systems. OCC too has holes. My model not only works more rigorously than OCC, it passes the Occam's Razor test - it is clean and understandable and computationally efficient. It makes it possible for an AI to understand human emotion, and have emotions like a human.
Ask me some questions about people and psychology, I may discuss some details of my approach if they are not key pieces whose explanation would hurt my business model. Again, my speaking up was to counter the people who do not realize how close we really are to serious breakthroughs. I want people to understand that we are very close to a change of paradigms in both computing, and society.
One not only needs many, many processors, but the system architecture must facilitate massive connection between processors. We cannot build a machine with billions of point to point connections like the brain does. Which is one of the keys to how the brain outperforms any machine. It is not just massive parallelism, it is the ability to have many simultaneous connections active at the same time. You simply cannot do that in a Von Neuman, and crossbars and switches don't work. Switch fabrics are not quite what is needed either. So brains don't map at all well into conventional sequential processors nor the buses they use nor the I/O they use. A company I worked for had a good intermediate solution but it is proprietary so I can't discuss it here. I have an innovative solution for this problem, though, also a part of the big plan.
XMOS's FPGA technology is good for certain types of manufacturers but in the end I really need custom silicon because I need a lot of them at lowest cost. When high-density nanotechnology carbon tube switches are usable, that'll be the way to build eventually.
By the way, on the software side, Ericson's Erlang is actually a very good candidate for AI connectionist processing in that it supports thousands of simultaneous processes and is very robust. Worth looking at if you do AI.
My architecture is a compromise, running connectionist machines on multithreaded multicore conventional processors for now. Several vendors in the Silicon Valley area, where I live, make processors whose current designs could be modified to handle NNs better. One needs to be able to manage efficiently the ability to swap the connection matrices and weights in and out of running threads, each running a small recognizer (NN). I worked with a chip design team doing something like this, a year ago. I envision 1024-core processors, each multithreaded. Eight threads is a practical limit. So a 16,000 processor machine of these (16K packages each with 1024x8 threads) might be at the lower bound of managing millions of neurons cost-effectively and fast enough to handle real-world problems. As I said before, I really need massively parallel. Danny Hillis, where art thou and thy Connection machine:) and I am still working out a good theory of knowledge-swapping in virtual-NN based systems. Patent when I have rigorous theory and a right approach. So no real details here.
Consider the 2008 election campaign. As each candidate makes statements and postures, we expand our knowledge about them and the political situation. Our belief collection not only modifies but grows as we add new data. Now consider anything else in our lives: technology, new software, new movies, books, personal events. With each, we are likely to add new beliefs about some things. So, a 1-year old has a belief set and knowledge, and 60-year old has accumulated a larger belief set and more knowledge over time.
"what amounts to 18 years of accumulation"
Exactly why I said amounts to. If I can record and swap out and copy all the knowledge of a matured AI, I can rapidly clone it into new ones, thereby not taking 18 years to train them. I just have to have an architecture that supports reading in weights and setting connections and configuring billions of neurons. Howevr, there is a downside in this because it limits flexibility of each new intelligence. If I want to make laborers but also make Hemingways, I have to support significantly different knowledge bases. Since in my architecture, knowledge maps into recognizers, each 'personality' will have different microarchitectures.
To explain brain function during sleep, I'd have to explain in depth my theory of states of consciousness. That takes up a chunk of my Vol 3 and can't be quickly done here. I posit multiple levels of states of consciousness ranging from unconscious (non functional), unconscious (dreaming), semiconscious (hallucinating), normal (literal), normal (imagining), normal (intuitive reaction), normal (meditating), and metaconscious (somewhat like autistic savant). I believe that during sleep when dreaming, retained emotional states attached to events of the day (concerns about goal success or failure) plus physiological stimuli (cold room, or hand in bowl of warm water etc) cause the brain to recreate types of events matching the emotion. For example, daytime worry about pressures (money, job, relationship) will reflect into the dream state where we will create a story around the quality of these pressures. I think we then try to resolve the emotion by a solution within the dream story. Thus in the above case one might dream of running away, or flight, or even think up a real solution to the money or people problem. (Flight is not a practical solution, but I didn't say the brain always finds a perfect solution, it just goes exploring.) At the end, the brain tries to lay the emotional drive to rest, and stores any relevant knowledge away. On academic material learned during the day, that does not have emotional tags attached, and is handled in a different drier way. It just gets stored depending on how strongly it was reinforced during the day. However in the case of Kekule dreaming of a snake with its tail in its mouth, his unconscious processes sought and found a solution to the carbon structure for the benzene ring, driven by a strong desire to solve the puzzle, an emotional drive. An AI might dream of a tough problem and creatively, not logically, seek solutions where a perfect logic machine might stall on something not in the prescribed world frame.
Patents are necessary when you are competing against Microsoft and Google, because a businessman cannot allow someone else to use the legal system against him on his own inventions, even take them away. Google has claimed AI within 20 years, and I believe they are closer than MS. However, neither side shows signs of discovering some key pieces I have. Protection is good, and I very likely will partner with Google if the opportunity arises. But I have to own my work first.
It's okay to doubt it; until I demo pieces to people, they don't believe it either. I do have strong AI right now running on a Von Neuman class processor, but AI really needs a massively parallel architecture. Speed IS relevant. You can't just run big wide NNs simulated in a Von Neuman, it takes forever. The best architecture I find is where processing and recognition are the same as memory. That is, once the AI learns something, the pattern recognizer also serves as the memory and the belief logic. In a way, the NNs ARE the data storage, there is no separate RAM. And my AI is implemented in a connectionist architecture that embodies symbolic processing in a new way. The knowledge storage is synonymous with the logic. But this requires massive though simple logic units and parallelism. Further, it is not completely self-organizing from the ground up, I view that as a hopeless approach.
I define consciousness and awareness within a pre-determined architecture, not entirely self-organizing from scratch. The visual front-end in particular is very rigid, but I think it is okay because there is little need for an organism to self-evolve completely new architectures, but rather to be able to run cascaded pattern-recognizers. See the work of Biederman for examples of this. The VFE feeds deeper processing doing cognition, and there is feedback from that to the VFE for training. Just as a baby learns to see, and recognize shapes, and build up from that. The front end is trained as the cognitive end learns and grows too.
AI does not spontaneously rise alone from massive databases, either. I view that approach as useless and a false trail. However, human intelligence does depend on belief systems and knowledge, and those continually grow as we mature from infancy. But to create the equivalent of an 18 year old, you have to have what amounts to 18 years of accumulation of knowledge about the world, and draw upon that. And there is a key but proprietary subtlety about that I'm devoting an entire volume to, that is the key to humanlike AI. That volume is essentially a doctoral thesis about consciousness reworked for use by a design staff. As for funding, no yachts yet. But I'm real, sane, and not a charlatan, and have explained my technology to my patent attorney. I expect to be hiring staff within two years. I posted on Slashdot not for glory but to counter all the nay-sayers who haven't a clue what is achievable.
I'll go Ray one better. We will have this before 2029. My company is working to bring to market synthetic intelligences that among other things have feelings, emotion, and mood, and understand human emotion. One may say, why try to do this? Because in order to understand people, a synthetic intelligence must understand these things, which serve real functions in people, and real functions in AIs trying to operate in a people world.
Note that it is not necessary to build 'perfect robots'. People think, and yet they are not perfect. They make mistakes, yet navigate through life. So we do not have to make flawless logic brains. The way people work is that we try to find good if not optimal solutions to problems, but we do not always exhaustively search for the perfect solution. Thus many problems in life can be solved in different ways than you would expect. We do not have to build a machine that finds the optimal solution to a traveling salesman problem in order to make a system that can walk from the kitchen to the front door. It just has to be able to get there reasonably optimally. Also, we do not have to replicate the human brain in order to think much like a human, we merely have to come up with functional systems that can provide similar functions. For instance, the human brain has the amygdala, which can be likened to an interrupt controller for emotional responses. Well, that functionality can be done in a hardware-software system that reasons about priorities of tasks and goals depending on their current 'value' of urgency to the 'brain'.
Many current researchers in many cases are missing the mark. For example, as good as it is, the widely-used AI textbook by Stuart Russell and Peter Norvig (who heads research at Google) has major omissions. It does not dwell on key things needed to bridge between AI and human psychology. Other things like the OCC model of emotion used in AI is incomplete and incorrect in parts. A new approach has been needed, and one I've been developing for decades in stealth mode. I'm writing a 5-volume book set on it. I want it to be the Knuth set of AI.
I'm in the process of patenting the mechanizations of my underlying technologies, and trying to cut deals with companies making multicore processors so their architectures support the thread swapping needed to make virtual neural nets practical. Once we get a 1024 simplified-core processor that supports virtual NNs, it'll be a lot easier to build a machine with many of these that does for NNs what disk swapping does for OSs, than to build a billion-neural processor hardwired machine. And easier to do visual perception systems properly too. So Ray is right. If I can drive certain companies to build the right silicon, we can get there by or before 2029. My current software does what I said, but it's too slow on current hardware. Needs new processors and new system architectures, and it will take 20 years to get the infrastructure all built up. But not a lot more.
Dear DOE; please accept my research grant proposal for a project involving energy, and pornography. Results might be available in as little as 30 seconds.
There's considerable utility in developing AIs that can understand drives, feelings, and emotions. Because one cannot understand people unless it understands these things in people. However, this understanding does not spring right out of any AI, it has to be engineered into the architecture and come from both mechanisms and data in knowledge bases. Further, it's clear from human disorders that some people do not understand other people's emotions much if at all, and it hampers them in interacting with people. And you have only to consider teenage dating to see examples of this. Until a kid comes to understand other people's feelings and needs and drives, he isn't going to do well.
For AIs to come, in order for them to work on human problems, not just spreadsheet calculations, they have to have an innate understanding of human psychology. If I need a surrogate agent to do some task for me, it may need to know my values in order to act as I would. I'm an AI researcher and see this as a necessity, and one of the things that differentiates powerful synthetic intelligences from the trivial and narrow ones.
I'm not talking about the multitude of honest workers. All it takes is one bad apple to do a lot of damage, to mix a metaphor. If a gang member steals enough data for the gang to make a lot of money using even just 500 employees' personal data, it can be far more than a low-wage job pays, and the economics are attractive to the gang. Besides, few thieves would be there long enough to collect retirement benefits. The idea in a lot of low-end crime is work as little as you can, make some easy money, then split. Many do not or cannot think far ahead to consequences. So (risk+effort)/reward ratio is not always a deterrent for them. And not all IT professionals are sane and straight arrow. I've known quite a few, thank you, who are messed up people. One kept child porn on the company server. I guess he thought that being sys admin kept him safe. Another had a drug habit. Costly, I expect, and led to illegal activity to raise money. Another peddled company data to unscrupulous competitors. I've seen enough cases that I do not believe we can trust everyone to be deterred by the same factors that keep the rest of us straight and narrow.
Indeed good grief. I saw an article some time ago noting that some Southern California gangs were infiltrating girlfriends into various financial processing institutions to steal credit card information, banking info, and so on. Even into the DMV. So there's certainly low-level activity. At one company I worked at, a crook got a job in the accounting department and somehow stole all the HR data, and some of that was used to get credit cards. How long before serious organized crime runs multiple active efforts for this? And how many Web commerce sites do criminal background checks on IT personnel?
Dear Scientists; please store your sensitive nuclear data with Google. We promise not to give it to the Chinese. Our company motto: Do No Evil You Can Get Caught Profiting From.
As if we didn't have enough trouble with drunken diapered astronauts, now NASA's come up with a way to have martinis in space! They should have stuck with Jello Shots in a Tube, TangDrivers, and secretly fermenting raisins from their Space Lunches. Not to mention huffing escaping gas from the air conditioning system. Yes, these plain-vanilla pilots and scientists have a wild side. The dewy-eyed novices on all-male flights awarded their first "Member of 50-Mile High Club" patch. The ones with a secret tattoo of Richard Simmons on their lower back saying "Your Space Buddy!" The "NASA Says Save Water in Space, Shower With Your Co-Pilot" ecology program. Oh, the horror. Cover your eyes, children.
Wal-Mart, Target, etc are still viable sales points for games. I wouldn't buy there, but millions of consumers are too large a number to be ignored to marketers.
If they can do it with a greeting card, I imagine they could create product packaging that plays sounds. How about a PC game box that plays the sounds from the game, for example American McGee's Alice which had interesting tracks?
No no. See, I knew the American lawyers would advise client to take the better deal. So it is only $10,000. By the way, my wife is available for only $5000 dollars, and she doesn't bitch, moan, or nitpick anywhere as much as the last 10 posters. She has few flaws, and does not require another sex change operation, as she has already had that one. She is perfect for Slashdotters. Act fast.
I am sad to note that raisins and grapes apparently are toxic to dogs and can cause acute renal failure even in limited amounts. Dogs are also lactose intolerant, and chocolate is not good for them either. Rat poison attracts dogs and tastes good to them, but has terrible effects. Unfortunately these awful substances are completely legal to possess, even if mixed with foods such as dog treats. I also note that in college, we used to use surgical-tubing slingshots to hurl things from a distance. Not that that's relevant in any way.
Slashdot Mirror
I will indeed go for the Turing competition eventually. But you should read what I said more carefully. I did not claim to be running everything integrated and fully functional today. This is an enormously complex system, and the pieces to be integrated are not conventional code, they are theory implemented in NNs. That is a lot of work. But I have significant pieces right now that do what no one else currently can do, and chunks planned but not implemented from the underlying theory yet, and years more to go within my development. But enough done that it's clear it's all on target. I said a time frame for human-like AI lies within Kurzweil's prediction. And I expect to have major pieces pulled together within two and certainly five years.
Also, it seems you're thinking conventionally. You call the AI a program. That shows you don't have the right model. Although some people run AI routines in software, full strong AI is nothing like conventional software. My AI is not an AI program, a piece of code. It is AI emerging from neural nets, and some new concepts, under the right machine architecture. That's nowhere the same thing as a program in C++ or LISP. Very different. My current software runs low level code that merely runs NNs on a Von Neumann platform, which is simply not optimal for MPP. The intelligence comes from the way the NNs connect, how they learn, what they learn, and how they control the dynamics of their organization. It is a layer above the code and is nothing like conventional programming, and I admit hard to understand.
Go read any of my comments on Slashdot for the last five years. If I were crazy, stupid, a conman, or a grandiose phony lying egotist, it would have shown before this. Read my January 20 AI comment; I explained then some of my work. It's real; I didn't just make it up for today's thread.
As for Skynet .. hmm. Great idea! Killer robots destroying humanity? I...I didn't think of that! Cough.
There are people of questionable validity in the field. For instance, John LaMuth has a patent on implementing emotion in software. His patent is full of handwaving - in one part he points to an expert system and says in effect "oh, the logic is all in that black box." But never explains anything about what is inside. I on the other hand have a legitimate and working alternative to both LaMuth and the more widely accepted OCC model of emotion in computing systems. OCC too has holes. My model not only works more rigorously than OCC, it passes the Occam's Razor test - it is clean and understandable and computationally efficient. It makes it possible for an AI to understand human emotion, and have emotions like a human.
Ask me some questions about people and psychology, I may discuss some details of my approach if they are not key pieces whose explanation would hurt my business model. Again, my speaking up was to counter the people who do not realize how close we really are to serious breakthroughs. I want people to understand that we are very close to a change of paradigms in both computing, and society.
XMOS's FPGA technology is good for certain types of manufacturers but in the end I really need custom silicon because I need a lot of them at lowest cost. When high-density nanotechnology carbon tube switches are usable, that'll be the way to build eventually.
By the way, on the software side, Ericson's Erlang is actually a very good candidate for AI connectionist processing in that it supports thousands of simultaneous processes and is very robust. Worth looking at if you do AI.
My architecture is a compromise, running connectionist machines on multithreaded multicore conventional processors for now. Several vendors in the Silicon Valley area, where I live, make processors whose current designs could be modified to handle NNs better. One needs to be able to manage efficiently the ability to swap the connection matrices and weights in and out of running threads, each running a small recognizer (NN). I worked with a chip design team doing something like this, a year ago. I envision 1024-core processors, each multithreaded. Eight threads is a practical limit. So a 16,000 processor machine of these (16K packages each with 1024x8 threads) might be at the lower bound of managing millions of neurons cost-effectively and fast enough to handle real-world problems. As I said before, I really need massively parallel. Danny Hillis, where art thou and thy Connection machine :) and I am still working out a good theory of knowledge-swapping in virtual-NN based systems. Patent when I have rigorous theory and a right approach. So no real details here.
Consider the 2008 election campaign. As each candidate makes statements and postures, we expand our knowledge about them and the political situation. Our belief collection not only modifies but grows as we add new data. Now consider anything else in our lives: technology, new software, new movies, books, personal events. With each, we are likely to add new beliefs about some things. So, a 1-year old has a belief set and knowledge, and 60-year old has accumulated a larger belief set and more knowledge over time.
"what amounts to 18 years of accumulation"
Exactly why I said amounts to. If I can record and swap out and copy all the knowledge of a matured AI, I can rapidly clone it into new ones, thereby not taking 18 years to train them. I just have to have an architecture that supports reading in weights and setting connections and configuring billions of neurons. Howevr, there is a downside in this because it limits flexibility of each new intelligence. If I want to make laborers but also make Hemingways, I have to support significantly different knowledge bases. Since in my architecture, knowledge maps into recognizers, each 'personality' will have different microarchitectures.
To explain brain function during sleep, I'd have to explain in depth my theory of states of consciousness. That takes up a chunk of my Vol 3 and can't be quickly done here. I posit multiple levels of states of consciousness ranging from unconscious (non functional), unconscious (dreaming), semiconscious (hallucinating), normal (literal), normal (imagining), normal (intuitive reaction), normal (meditating), and metaconscious (somewhat like autistic savant). I believe that during sleep when dreaming, retained emotional states attached to events of the day (concerns about goal success or failure) plus physiological stimuli (cold room, or hand in bowl of warm water etc) cause the brain to recreate types of events matching the emotion. For example, daytime worry about pressures (money, job, relationship) will reflect into the dream state where we will create a story around the quality of these pressures. I think we then try to resolve the emotion by a solution within the dream story. Thus in the above case one might dream of running away, or flight, or even think up a real solution to the money or people problem. (Flight is not a practical solution, but I didn't say the brain always finds a perfect solution, it just goes exploring.) At the end, the brain tries to lay the emotional drive to rest, and stores any relevant knowledge away. On academic material learned during the day, that does not have emotional tags attached, and is handled in a different drier way. It just gets stored depending on how strongly it was reinforced during the day. However in the case of Kekule dreaming of a snake with its tail in its mouth, his unconscious processes sought and found a solution to the carbon structure for the benzene ring, driven by a strong desire to solve the puzzle, an emotional drive. An AI might dream of a tough problem and creatively, not logically, seek solutions where a perfect logic machine might stall on something not in the prescribed world frame.
Patents are necessary when you are competing against Microsoft and Google, because a businessman cannot allow someone else to use the legal system against him on his own inventions, even take them away. Google has claimed AI within 20 years, and I believe they are closer than MS. However, neither side shows signs of discovering some key pieces I have. Protection is good, and I very likely will partner with Google if the opportunity arises. But I have to own my work first.
I define consciousness and awareness within a pre-determined architecture, not entirely self-organizing from scratch. The visual front-end in particular is very rigid, but I think it is okay because there is little need for an organism to self-evolve completely new architectures, but rather to be able to run cascaded pattern-recognizers. See the work of Biederman for examples of this. The VFE feeds deeper processing doing cognition, and there is feedback from that to the VFE for training. Just as a baby learns to see, and recognize shapes, and build up from that. The front end is trained as the cognitive end learns and grows too.
AI does not spontaneously rise alone from massive databases, either. I view that approach as useless and a false trail. However, human intelligence does depend on belief systems and knowledge, and those continually grow as we mature from infancy. But to create the equivalent of an 18 year old, you have to have what amounts to 18 years of accumulation of knowledge about the world, and draw upon that. And there is a key but proprietary subtlety about that I'm devoting an entire volume to, that is the key to humanlike AI. That volume is essentially a doctoral thesis about consciousness reworked for use by a design staff. As for funding, no yachts yet. But I'm real, sane, and not a charlatan, and have explained my technology to my patent attorney. I expect to be hiring staff within two years. I posted on Slashdot not for glory but to counter all the nay-sayers who haven't a clue what is achievable.
Note that it is not necessary to build 'perfect robots'. People think, and yet they are not perfect. They make mistakes, yet navigate through life. So we do not have to make flawless logic brains. The way people work is that we try to find good if not optimal solutions to problems, but we do not always exhaustively search for the perfect solution. Thus many problems in life can be solved in different ways than you would expect. We do not have to build a machine that finds the optimal solution to a traveling salesman problem in order to make a system that can walk from the kitchen to the front door. It just has to be able to get there reasonably optimally. Also, we do not have to replicate the human brain in order to think much like a human, we merely have to come up with functional systems that can provide similar functions. For instance, the human brain has the amygdala, which can be likened to an interrupt controller for emotional responses. Well, that functionality can be done in a hardware-software system that reasons about priorities of tasks and goals depending on their current 'value' of urgency to the 'brain'.
Many current researchers in many cases are missing the mark. For example, as good as it is, the widely-used AI textbook by Stuart Russell and Peter Norvig (who heads research at Google) has major omissions. It does not dwell on key things needed to bridge between AI and human psychology. Other things like the OCC model of emotion used in AI is incomplete and incorrect in parts. A new approach has been needed, and one I've been developing for decades in stealth mode. I'm writing a 5-volume book set on it. I want it to be the Knuth set of AI.
I'm in the process of patenting the mechanizations of my underlying technologies, and trying to cut deals with companies making multicore processors so their architectures support the thread swapping needed to make virtual neural nets practical. Once we get a 1024 simplified-core processor that supports virtual NNs, it'll be a lot easier to build a machine with many of these that does for NNs what disk swapping does for OSs, than to build a billion-neural processor hardwired machine. And easier to do visual perception systems properly too. So Ray is right. If I can drive certain companies to build the right silicon, we can get there by or before 2029. My current software does what I said, but it's too slow on current hardware. Needs new processors and new system architectures, and it will take 20 years to get the infrastructure all built up. But not a lot more.
"Can you do bagels?"
"I'm sorry, Dave. But do you want to hear me sing "Daisy, Daisy?"
"Spare change?"
"You lookin' at ME?"
"Mister, buy me a drink?"
"Are you sure you're not Richard Stallman?"
"I think my nuts are loose. Check 'em for me?"
Thank you
For AIs to come, in order for them to work on human problems, not just spreadsheet calculations, they have to have an innate understanding of human psychology. If I need a surrogate agent to do some task for me, it may need to know my values in order to act as I would. I'm an AI researcher and see this as a necessity, and one of the things that differentiates powerful synthetic intelligences from the trivial and narrow ones.
I'm not talking about the multitude of honest workers. All it takes is one bad apple to do a lot of damage, to mix a metaphor. If a gang member steals enough data for the gang to make a lot of money using even just 500 employees' personal data, it can be far more than a low-wage job pays, and the economics are attractive to the gang. Besides, few thieves would be there long enough to collect retirement benefits. The idea in a lot of low-end crime is work as little as you can, make some easy money, then split. Many do not or cannot think far ahead to consequences. So (risk+effort)/reward ratio is not always a deterrent for them. And not all IT professionals are sane and straight arrow. I've known quite a few, thank you, who are messed up people. One kept child porn on the company server. I guess he thought that being sys admin kept him safe. Another had a drug habit. Costly, I expect, and led to illegal activity to raise money. Another peddled company data to unscrupulous competitors. I've seen enough cases that I do not believe we can trust everyone to be deterred by the same factors that keep the rest of us straight and narrow.
Indeed good grief. I saw an article some time ago noting that some Southern California gangs were infiltrating girlfriends into various financial processing institutions to steal credit card information, banking info, and so on. Even into the DMV. So there's certainly low-level activity. At one company I worked at, a crook got a job in the accounting department and somehow stole all the HR data, and some of that was used to get credit cards. How long before serious organized crime runs multiple active efforts for this? And how many Web commerce sites do criminal background checks on IT personnel?
Dear Scientists; please store your sensitive nuclear data with Google. We promise not to give it to the Chinese. Our company motto: Do No Evil You Can Get Caught Profiting From.
Well, if they combined this with a RealDoll, they might have something useful.
As if we didn't have enough trouble with drunken diapered astronauts, now NASA's come up with a way to have martinis in space! They should have stuck with Jello Shots in a Tube, TangDrivers, and secretly fermenting raisins from their Space Lunches. Not to mention huffing escaping gas from the air conditioning system. Yes, these plain-vanilla pilots and scientists have a wild side. The dewy-eyed novices on all-male flights awarded their first "Member of 50-Mile High Club" patch. The ones with a secret tattoo of Richard Simmons on their lower back saying "Your Space Buddy!" The "NASA Says Save Water in Space, Shower With Your Co-Pilot" ecology program. Oh, the horror. Cover your eyes, children.
Ah. Welcome to the $10,000 0.5 mw HeNe laser tube.
In related news, by an odd coincidence, a huge asteroid landed on NASA headquarters and totally obliterated it. Martians everywhere cheered.
Wal-Mart, Target, etc are still viable sales points for games. I wouldn't buy there, but millions of consumers are too large a number to be ignored to marketers.
If they can do it with a greeting card, I imagine they could create product packaging that plays sounds. How about a PC game box that plays the sounds from the game, for example American McGee's Alice which had interesting tracks?
Actually, after you read it, we de-created New York. Then we created Slashdot, to distract you.
--- Bogo Mugubwai, Nigeria
I am sad to note that raisins and grapes apparently are toxic to dogs and can cause acute renal failure even in limited amounts. Dogs are also lactose intolerant, and chocolate is not good for them either. Rat poison attracts dogs and tastes good to them, but has terrible effects. Unfortunately these awful substances are completely legal to possess, even if mixed with foods such as dog treats. I also note that in college, we used to use surgical-tubing slingshots to hurl things from a distance. Not that that's relevant in any way.