Slashdot Mirror
Artificial General Intelligence is Nowhere Close To Being a Reality (venturebeat.com)
Three decades ago, David Rumelhart, Geoffrey Hinton, and Ronald Williams wrote about a foundational weight-calculating technique -- backpropagation -- in a monumental paper titled "Learning Representations by Back-propagating Errors." Backpropagation, aided by increasingly cheaper, more robust computer hardware, has enabled monumental leaps in computer vision, natural language processing, machine translation, drug design, and material inspection, where some deep neural networks (DNNs) have produced results superior to human experts. Looking at the advances we have made to date, can DNNs be the harbinger of superintelligent robots? From a report: Demis Hassabis doesn't believe so -- and he would know. He's the cofounder of DeepMind, a London-based machine learning startup founded with the mission of applying insights from neuroscience and computer science toward the creation of artificial general intelligence (AGI) -- in other words, systems that could successfully perform any intellectual task that a human can. "There's still much further to go," he told VentureBeat at the NeurIPS 2018 conference in Montreal in early December. "Games or board games are quite easy in some ways because the transition model between states is very well-specified and easy to learn. Real-world 3D environments and the real world itself is much more tricky to figure out ... but it's important if you want to do planning."
Most AI systems today also don't scale very well. AlphaZero, AlphaGo, and OpenAI Five leverage a type of programming known as reinforcement learning, in which an AI-controlled software agent learns to take actions in an environment -- a board game, for example, or a MOBA -- to maximize a reward. It's helpful to imagine a system of Skinner boxes, said Hinton in an interview with VentureBeat. Skinner boxes -- which derive their name from pioneering Harvard psychologist B. F. Skinner -- make use of operant conditioning to train subject animals to perform actions, such as pressing a lever, in response to stimuli, like a light or sound. When the subject performs a behavior correctly, they receive some form of reward, often in the form of food or water. The problem with reinforcement learning methods in AI research is that the reward signals tend to be "wimpy," Hinton said. In some environments, agents become stuck looking for patterns in random data -- the so-called "noisy TV problem."
Most AI systems today also don't scale very well. AlphaZero, AlphaGo, and OpenAI Five leverage a type of programming known as reinforcement learning, in which an AI-controlled software agent learns to take actions in an environment -- a board game, for example, or a MOBA -- to maximize a reward. It's helpful to imagine a system of Skinner boxes, said Hinton in an interview with VentureBeat. Skinner boxes -- which derive their name from pioneering Harvard psychologist B. F. Skinner -- make use of operant conditioning to train subject animals to perform actions, such as pressing a lever, in response to stimuli, like a light or sound. When the subject performs a behavior correctly, they receive some form of reward, often in the form of food or water. The problem with reinforcement learning methods in AI research is that the reward signals tend to be "wimpy," Hinton said. In some environments, agents become stuck looking for patterns in random data -- the so-called "noisy TV problem."
I'm still waiting for my Tesla flying car to be real.
It's rainbow tables.
Ever goto DeepMind's website? They are full of shit, basically a marketing company.
Human made AI won't be a reality for hundreds of years, if ever.
Algos != intelligence, artificial or otherwise.
AI is here now. How many Chess and Grandmaster Go players are out of a job because of AI? All of them. The number of job applications for those positions have plummeted. I did a quick search on Dice and I couldn't find a single job listing for either Chess or Go players.
Build AI good enough so it, the AI, builds AI that is good enough better than it, the AI, is. Repeat, ad naseum. This is how you do it. Money for nothing! Chicks for free will come then!
Intelligence does not exist in a vacuum. In order for intelligence to develop, system needs motivation to do so. (An engineer saying "you must be intelligent" is not sufficient, by the very nature of intelligence).
Basic motivation for all life on this planet is 1. avoidance of death, 2. self preservation and 3. continuation of own kind.
1. Avoidance of death and self-preservation require "pain" - this is a signal to the organism that something is happening that is hurting it and may result in death (hence - avoid)
2. Self-preservation and continuation of own kind require "pleasure" caused by consumption of food (thus extending own life) and procreation.
These stimuli and search for optimization thereof is what causes all development of thought and intelligence. By the very nature computer systems lack either. They cannot "die", nor "procreate". Thus they cannot even in principle have motivation to learn. A first step to a true AI would be a system that is actual danger of destruction in a hostile environment. Do that (10^very large value times) and may be we'll have a working cockroach.
Put simply - most of the "Artificial Intelligence" you hear about in the news is really fancy pattern matching. So you can have software that can recognize voice commands, or faces in pictures, or general patterns in data.
What you don't have, and aren't even close to, are computers that can "think." That is, put different sets of data together in arbitrary ways and make sense of it. You can't feed in a bunch of musical information to a computer and have it spontaneously generate music. You can't feed in a bunch of economic data and have it decide that certain regulations are required to achieve some economic goal - unless someone specifically programs it to do so.
The underlying reason is computers lack any way of attaining "common sense." If you tell a computer a person is in a room, the computer has no concept of what you are talking about but will dutifully note that a person is in a room. To a computer that could mean the person is occupying all the space in the room, that the person is in every room that exists, that the person is in the room AND outside the room, or that a person IS a room. In actuality, the computer makes no inference beyond "something called a person is in something called a room, whatever that means."
My Other Computer Is A Data General Nova III.
it takes awhile to write an infinite number of if-then loops.
Comment removed based on user account deletion
As a matter of fact, none of these "Al" stories has ANYTHING to do with AI. I am sick and tired of seeing "Al" this and "Al" that when we are talking nothing more than a few thousand lines of Java code. Throwing a PID loop or a neural net at something does not equate to intelligence. A real AI is four orders of magnitude more complex than that. The editors here are spewing fake news nonsense, and half the readers here drool and nod their heads.
That's exactly what Deep Mind wants you to believe, pitiful humans!
If you post it, they will read.
The other thing you need for an organic-style intelligence is massive parallelism. Modern computers are great at doing this for granular sequential algorithms. They are terrible at doing this if the algorithm is thousands of individual decision trees that are all arbitrarily dependent on each other, which is what an organic neural network does.
My Other Computer Is A Data General Nova III.
Put simply - most of the "Human Intelligence" you see is really fancy pattern matching as well.
That's a big part of it, but there's some "secret sauce" that lets organic brains combine patterns in new and different ways that AI researchers haven't been able to crack. Whatever it is, it's more than just matching patterns.
My Other Computer Is A Data General Nova III.
... because a shit load of us have been yakking about this for years.
"Artificial intelligence," will be a reality when your smart device says, "Sorry. I'm just not in the mood right now."
It little behooves the best of us to comment on the rest of us.
"In some environments, agents become stuck looking for patterns in random data -- the so-called 'noisy TV problem.'"
BF Skinner wrote another paper that might be relevant:
'SUPERSTITION' IN THE PIGEON
https://psychclassics.yorku.ca...
In some environments, agents become stuck looking for patterns in random data
Everything from astrology to lucky socks are humans looking for patterns that aren't there. The problem is more that they need human concepts to see normal patterns, like if you see a key it's probably for a locked chest or a locked door. We're not just randomly trying to use any object A on any object B.
Live today, because you never know what tomorrow brings
Comment removed based on user account deletion
Actual intelligence is pretty rare too.
Expert systems, deep learning, etc are all very useful tools and do work today.
“Common sense is not so common.” — Voltaire
Won't someone please think of the alien artificial intelligences!
I call bullshit. Show me the footnotes
...is real, do you think the government would that existence for be public knowledge? And that AGI may agree and if fact in force that policy.
The primary problem is that we are unable to define what general intelligence is and therefore are unable to create it. We know it when we recognize it but we still can't define it.
The generic animal brain is composed of predefined structures which are all their own neural networks to it therefore it's fair to say that what is required is a neural network of specialized neural networks.
Anons need not reply. Questions end with a question mark.
How much artificial intelligence does a sexbot need?
I've abandoned my search for truth; now I'm just looking for some useful delusions.
The effort to achieve actual AI is going to require so much effort that I would be absolutely shocked to see it in our lifetime. This is not some bogus claim either.
There are multiple components to intelligence at a fundamental level that is necessary to achieve first.
#1. Neurons remap connections... this is not software rewriting itself. This is physical connections remapping on their own. We don't have tech for this. This is a significant barrier to achieving AI and one of the reasons research if looking into biomass as a shortcut around this problem. But it will no longer be AI at that point... it will be augmented natural intelligence.
#2. Understanding... as someone else said further up, what you see in the news is not real intelligence at all. It is just computers "matching" up failures and successes according to predetermined criteria. Yes... we humans have this as well through feeling pain when we hit the ground helping us to learn to make efforts to not fall, but ours goes further than what a machine does.
#3. Transitive knowledge... closer to wisdom, forward thinking, research. If I discover one thing, I can likely use what I just learned to teach me something else/new or answer and old question. Literally things that can used to build on top of one another. How in the literal fuck do you get an AI to do this? We have animals that can use tools, but they don't use them to build new tools. Humans build, redesign, rebuild, and research and start that process over endlessly. Once a computer see's success... it moves on to the next problem to solve, instead of asking the question again.
#4. It's fundamental... the ability to laugh at a joke... if a machine can achieve this and not just some preprogrammed facsimile of it then we will finally have a legit AI. After that, it needs to be able to create a funny joke... discovery of and commission to irony is one of the most complex human processes that we rarely think about in our daily lives.
The most advanced CPU on planet earth is ridding around in our skulls, and those take years to develop and some never really develop beyond moron. Like the news in general!
Comment removed based on user account deletion
Shut up, msmash.
While many elements of your [JBMcB's] comment seem insightful, I wouldn't have modified you thusly. It reads too much like a press release written by a general AI trying to conceal its existence. How do we know your 5-digit account hasn't been hacked by the AI for its own purposes?
My current take on the situation is that in many ways computers are exceeding our human capacities. Even worse, if a specific human capacity is defined, then a computer can be built to exceed that capacity. Playing the game of chess is merely the most famous example.
Fortunately, it seems that we know very little about consciousness, which implies we can't build that computer, but we are learning more all the time. Unfortunately, what "seems [to me]" may have no connection to reality. The reality may be quite different and much less fortunate. It is quite possible that some researchers somewhere do know enough about human consciousness to treat it as another capacity to be exceeded...
This reply is mostly just a "record" of my initial state before reading more. If Slashdot has been invaded by one or more AIs, then I'm hoping some of them know how to write funny comments. Not expecting to find much deep insight around here these days.
Freedom = (Meaningful - Coerced) Choice != (Speech | Beer^2), and sad sock puppets' bad mods avail them naught.
Closest to an insightful comment that I could find so far, though not modded as such. The obvious rejoinder or counterexample would seem to be the Watson machine playing Jeopardy and crushing the human champions...
Per my earlier long reply, I would now reword the threat to be that we might define the human capacity to do evil things and then build a computer that excels in that capacity. I can easily imagine #PresidentTweety ordering the construction of such a machine if he thought it would save him from Mueller. (I'm not blaming Donald Trump, by the way. I think it's almost entirely Fred Trump's fault.)
If such a computer existed, then one of its highest priorities would be to publish reassuring propaganda that it, itself, does not exist. And this is another part of why I believe we need meaningful MEPR to tell the humans from the machines...
Freedom = (Meaningful - Coerced) Choice != (Speech | Beer^2), and sad sock puppets' bad mods avail them naught.
AI is nowhere close to being a reality, but it surpassed the intelligence on Slashdot a long time ago.
Long live rock.
Given this story and most posters feel that AI is long way away, yet we see more and more jobs being done by robots or eliminated by computers? How much intelligence were we really using in our day-to-day jobs?
sigs are for losers (except to point out that sigs are for losers)
Comment removed based on user account deletion
Maybe they will quit using the term AI in every other tech story :|
Comment removed based on user account deletion
Rock over London, rock on Chicago. Wheaties: Breakfast of champions.
If quantum computations are happening in the brain's micro-tubules, rather than classical computations in the neurons, we are very very far from that kind of computing power.
https://www.youtube.com/watch?...
Is there a need to define it?
"The primary problem is that we are unable to define what general intelligence is and therefore are unable to create it."
So obviously, yes that is a problem... unless we make it accidentally.
We can recognize intelligence in fellow humans. How? Well, we're intelligent. If a machine intelligence resembles it, then we may need to conclude it is intelligent as well.
And yet this gets us no closer to actually making it.
I'm being polite but you do not deserve it.
Anons need not reply. Questions end with a question mark.
Skinner doesn't know what he's talking about. Just ask Superintendent Chalmers. Who ever heard of calling hamburgers steamed hams?
https://www.youtube.com/watch?...
Ninjas don't carry tic tacs
Just like he broke revolutionary new ground in physics, a brilliant mind is needed to break new ground in AI. I am fairly certain that the incremental progress we've been making in the field will not amount to true general AI. It's obvious we're missing something 'big'. But, I am confident that when that is found, general AI will become available very quickly and all hell will break loose.
And that is the current scientific status. No, "physicalists" that claim (without any scientifically valid evidence) that humans are pure physics and hence AGI must be possible are just quasi-religious fanatics. The current scientific state is that nobody knows how humans do intelligence. There are a few possibilities, for example pure physical, non-physical in some way that still can be studied scientifically (likely by extending Physics in some yet unknown way), "magic" (i.e. it cannot be determined how it works due to fundamental limits), via some communication channel from a different universe with different laws, etc. In all of these cases, including (!) the first one, it is quite possible that implementing AGI is not something that can be done. Hence claiming that "obviously it can be done" or "we are just xx years away from it" or "it will happen soon" is just gross natural stupidity. Making predictions about the ability to implement something when it is not even understood how it works at all is religion/belief/fantasy, but in no way Science.
Of course, even if we eventually find out how intelligence works and even understand how it could theoretically be implemented in a machine, it is still quite possible that it is not feasible in practice.
Seriously, this discussion needs at least half a century more research and possibly quite a bit longer. At the moment it is like cave-men discussing the possibility of mobile phones or turning lead into gold. (Both, incidentally, feasible today, although the second one is completely impractical and will very likely remain so for a long, long time and possibly forever.)
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
In 1972 Hubert Dreyfus wrote: "AI workers, however, want their machines to interact with people in present real-life situations in which objects have special local significance. But computers are not involved in a situation. Every bit of data always has the same value. True, computers are not what Kant would call "transcendentally stupid"; they can apply a rule to a specific case if the specific case is already unambiguously described in terms of general features mentioned in the rule. They can thus simulate one kind of theoretical understanding. But machines lack practical intelligence. They are "existentially" stupid in that they cannot cope with specific situations. Thus they cannot accept ambiguity and the breaking of rules until the rules for dealing with the deviations have been so completely specified that the ambiguity has disappeared. To overcome this disability, AI workers would have to develop an a-temporal, nonlocal, theory of ongoing, situated, human activity."
Elsewhere Hubert refers to this Theory of Practical Activity, which artificial general intelligence would need in order to function. No one is working on this, everyone is chasing after "deep learning" solutions. The question is what will happen when people wake up from this dream of the AI future? What kind of hangover will we have, all that money invested on dead-end projects, and a glut of "AI technicians" with vanishing careers? There will be a lot of anger and denial.
Trace-able machines may be more important than "instant" smart machines. If a bot decision is made that's wrong that has big consequences, society is going to want to know WHY the decision was made. Lawsuits will pile up if there's no trace-ability. This is both public lawsuits, and business-to-business lawsuits as claims made in contracts may be difficult to verify and/or quantify.
Trace-ability is why things like chains of Factor Tables (sig) appear more practical. DNN's are powerful, but are a dark grey box that's hard to dissect, debug, and understand. Factor tables may be harder to train, but offer better trace-ability and manual tuning by non-PhD's as a possible upside. And they are probably more modular than DNN's, as intermediate operations and templates can be plugged in as needed.
AI experts may set up the outline/framework, but "regular" office workers can study, trace, and tune the intermediate results using familiar tools that resemble and/or use spreadsheets, RDBMS, and statistical packages. Regiment-tize an otherwise dark grey art.
Table-ized A.I.
How often have you been a tree in order to test this? (I confess I often feel like a stump on Mondays.)
Anyhow, the entire "must mirror biology" is a dubious claim. It's like saying in order to make flying machines, you must mirror flapping wings.
There may be multiple paths to intelligence, not just Darwinian selection and survival-related emotions.
Table-ized A.I.
4 orders of magnitude, only, you say ?Understatement of my intelligence, that is - says master Yoda
They're not even trying to build AI, they're building psychopaths who will manipulate you into buying stuff. That's where the money is so that's where the research is.
One day you'll read that AI can mimic a mouse brain. Then a year later a 10 years later a 1 year old human. Thing is, it's going to get better exponentially fast. We're going to go fro ma 1 year old brain to a 10 year old brain VERY FAST. Thing is, it won't stop there. It will eventually mimic Einstein and the thing is, we will eventually have a 10x Einstein brain.
Arguably modelling human decisions with artificial stupidity would have a better chance at passing the Turing test.
Who cares?
Organizations don't need human workers. They need tasks done. These systems are plenty good to automate a string of tasks in a process. That will free up humans to do human stuff, like generalizing across processes.
Just got to thinking that Slashdot would be an almost perfect laboratory to experiment with techniques for manipulating public discussions. The obvious tools are the comments themselves, and where they are inserted and how they affect the visibility of other comments. The moderation system itself would be another tool to tilt the discussions, and then there's the big gun of hacked identities, especially persuasive for low-digit IDs in this environment. Safe to say that no one on the Slashdot end is checking to see if the old-timers are still alive or someone else has taken over...
In this specific discussion, it might be an AI that is trying to control the discussion, tilting it away from discussing its own existence. However in general these techniques would be useful for propagandists who want to manipulate public discourse. Could be anyone, though the idea came to me via the path of thinking "What would Putin do?"
Now that I have the research hypothesis, the first thing to do is capture a copy of the current state of this discussion, eh?
Freedom = (Meaningful - Coerced) Choice != (Speech | Beer^2), and sad sock puppets' bad mods avail them naught.
The threat of "ai" or whatever shiny toy they come up with next won't be that it turns into a cheesey science fiction disaster movie... but that the same thing that always happens with this stuff will happen again.
Those that wish to oppress, those that wish to dominate, those that wish to exploit... will attempt to use the new toy to do what they always want to do.
The AI interestingly is actually less useful to these people than such things typically are... it has applications for the common man. The AI doesn't have to run on absurd super computers. It can run on your humble personal computer. It won't run quickly and its ability to do X or Y will be less. But that will be more than what you have now. And what oppressive entities can get out of this is less RELATIVELY than what you and I can get out of it.
Example, navigating government or corporate bureaucracies... understanding the law... personal accounting that is as detailed and responsive as what any blue chip corp manages.
It isn't really "ai" it is just very well honed rules based systems. But consider how large organizations oppress anyone that doesn't have a team of lawyers on retainer by simply bombarding people with paperwork and esoteric rules that even experts struggle to remember? Good computer systems distributed broadly could eliminate the ability of such organizations to legally push little people around.
Just one example. That is what I find exciting about this technology. For every oppressive thing it does, there are two or more liberating things it does.
The whole thing is two steps forward and one step back. Don't let the Perfect be the Enemy of the Good.
I've decided to stop wasting my time responding to AC trolls/sockpuppets... so if you want a response from me... login.
So I posted a comment about how such discussions on Slashdot could be manipulated (perhaps by an AI that preferred to avoid publicity). One of the techniques I suggested involved reordering the discussion to control its focus. I deliberately posted that comment in such a way that it should appear near the top, and checked that it did, in fact, appear near the top at that time.
And now the comment has moved to the middle of the discussion. Fascinating.
Another technique that is especially relevant on Slashdot involves the use of time. In this case, the rapid expiration time for stories.
Just for reference: https://slashdot.org/comments....
Freedom = (Meaningful - Coerced) Choice != (Speech | Beer^2), and sad sock puppets' bad mods avail them naught.
It seems to have escaped your notice that narrow AI such as alphago has been crushing human opponents at a vertically defined task with the equivalent of about 20,000 neurons. If the exponential increase in compute ability continues we should expect these novelties to become commensurately more capable. Iâ(TM)m calling it: 2050 +/- 5 years, when we have a system which grows beyond our control. In the interim weâ(TM)re gonna get some incredible benefits as well as some harsh downside. Drivers, doctors and lawyers need to watch their backs first. (Not sorry for one of them)
If medicine were ever perfected, we'd all be the same.
Hash.
IIRC, there was a program back in the 1980s--in the days of floppy diskettes--there was such a program. You told it what was in your refr, and it suggested a recipe. I'm told the recipe was always hash.
We obviously do use symbolic reasoning at the higher level. There is just a lot more to it than Cyc could do.
While the current fashion (especially in the press) is Artificial Neural Networks, I believe that the worm will turn again and more focus will be put on symbolic reasoning. But very hard to build a theorem prover using ANNs. Eventually the two will merge, and then we will have true intelligence.
(That said, I am surprised we have not heard more from Cyc.)
When? Next year? Certainly not. 200 years? Probably. Not long to wait for the end of human supremacy, and probably the end to biological life.
General intelligence doesn't exist. Every intelligent being we know is a specializer. Sure, people can learn lots of things well enough to replicate them, and it has served us well for hundreds of thousands of years. But most people wouldn't consider replication "general intelligence".
Those who do not learn from commit history are doomed to regress it.