Slashdot Mirror
Alpha Go Takes the Match, 3-0 (i-programmer.info)
mikejuk writes: Google's AlphaGo has won the Deep Mind Challenge, by winning the third match in a row of five against the 18-time world champion Lee Se-dol. AlphaGo is now the number three Go player in the world and this is an event that will be remembered for a long time. Most AI experts thought that it would take decades to achieve but now we know that we have been on the right track since the 1980s or earlier. AlphaGo makes use of nothing dramatically new — it learned to play Go using a deep neural network and reinforcement learning, both developments on classical AI techniques. We know now that we don't need any big new breakthroughs to get to true AI. The results of the final two games are going to be interesting but as far as AI is concerned the match really is all over.
Is this AI software written in Rust?
Every time a computer beats a human at a "smart" game, we hear the same thing. And every time, when all is said and done, all we have is a program that can play a game well (and maybe a really aggressive marketing campaign to sell consulting services, Watson).
Look, we barely understand what intelligence is let alone know what it means to have a computer replicate it. We can have computers perform tasks that we ascribe to smart people and call it intelligence, but that's about it right now.
And, with deep learning and neural nets, we haven't gained any real insights into intelligence. We just have a black box mathematical function that can play a game.
Although in a sense it is "nothing new" (neural networks and monte carlo statistical techniques), the combination is one of the most convincing demonstrations of short-circuiting huge branching factors to arrive at what a human player would call "intuition".
Chess has a branching factor of about 35. This is small enough that if you prune the most dismal lines, you can brute force the rest to many ply and arrive at a very good result, but this is not a well generalizable technique.
Go has a branching factor of about 250. This cannot be brute forced, even with aggressive pruning. The result of a NN evaluator function plus Monte Carlo has been astonishing: it was not predicted for computer Go to reach this strength for decades yet, but here we are.
The implications of this combination of techniques to other kinds of problems requiring "intuition" will be interesting to watch.
They've already stated that their next goal is to do it again without the human database, but rather through iteration. And the big final advancement *this time* was made by it playing a ton of games with itself. Studying the human games was not enough to get to this level.
And humans do the same thing. They spend their lives studying the important games that came before. So the point is it did it pretty much the same way humans do. And it has already played a move that no strong human has ever played (Game 2, move 37). At first it (not surprisingly) appeared to be blunder, until its strength became clear. Humans will now learn from the computer and their level of play will rise. It happened in chess and checkers, and in a very big way in backgammon. Any strong human backgammon player today would trounce the World Campion of 20 years ago.
Until this week you could have made a similar statement about Go.
No it's magic.... really!
love is just extroverted narcissism
We know now that we don't need any big new breakthroughs to get to true AI.
Grossly exaggerated claim. The following article worth reading on this subject by no one else than two of authorities in the field, one did work on the backgammon game in the 90s and the other one on the IBM Deep Blue program that win over the world chess champion Garry Kasparov in 1997. http://www.ibm.com/blogs/think... In particular:
"However, research in such “clean” game domains didn’t really address most real-life tasks that have a “messy” nature. By “messy,” we mean that, unlike board games, it may be infeasible to write down an exact specification of what happens when actions are taken, or indeed what exactly is the objective of the task. Real-world tasks typically pose additional challenges, such as ambiguous, hidden or missing data, and “non-stationarity,” meaning that the task can change unexpectedly over time. Moreover, they generally require human-level cognitive faculties, such as fluency in natural languages, common-sense reasoning and knowledge understanding, and developing a theory of the motives and thought processes of other humans."
Achille Talon
Hop!
Life could be a really complex search algorithm we just can't begin to comprehend. Life could be a non-linear approximation of 42...
Not much has changed in AI, the fundamentals on these new systems are still the same as before. The difference is we have more computing power (and CS work hours) to put into problems we thought were more amazingly complex than they were. It's likely that Go is actually as difficult as we thought it was so our progress on this search problem is not a result of a huge leap but merely a realization by some of us that the problem wasn't as big as we thought. It is akin to rating password security simply by the number of combinations and ignoring randomization: your 8 letter password might seem complex enough but in actuality it is weak because you picked a word from the dictionary. The prediction of your decision (randomness) aspect is a critical aspect.
What I'm trying to say is that the massive branching factor of Go in terms decision trees is like having long password length (like 56 letters long vs chess being 8) but the INTUITION of good decisions is akin to approaching a true random unpredictable password and it is NOT as difficult as we thought it was! Therefore, human professional intuition is not as miraculous or brilliant as previously thought. That is what this "leap" in Go is demonstrating. This is development humbling and/or frightful and should make one ponder our human hubris.
Democracy Now! - uncensored, anti-establishment news
We know now that we don't need any big new breakthroughs to get to true AI.
But, this isn't 'AI', it's just another 'expert system'.
Are YOU using the TOOL, or is the TOOL using YOU? Think about it!
But, this isn't 'AI', it's just another 'expert system'.
No. Alpha-Go is pretty much the opposite of an "expert system". Expert systems encode expert human knowledge in a series of explicit rules and if-then tables. Alpha-Go is based on neural nets and self-learning. There is no list of explicit rules.
By your definition it would be really hard to do AI :D
Program an AI! But without algorithms! No... stupid programmer, not even using self learning algorithms!
Not an easy task for a programmer to create a soul, even if such a thing exists.
I suspect that if we were to slightly tweak the rules that AlphaGo would become useless and our master player would easily adapt.
AlphaGo did most of its learning through self-reinforcement: playing against itself. So if the rules were changed, it could learn the new rules quickly through self-reinforcement, while the human player would have to relearn a lifetime of habits. Lee Sedol has been training daily since he was four years old. AlphaGo surpassed him after only a few hundred hours of training. The results of a rule change would likely be the opposite of what you predict. AlphaGo would adapt far quicker than a human.
We know now that we don't need any big new breakthroughs to get to true AI.
Err, no. Just... no.
systemd is Roko's Basilisk.
There's a well known phenomenon where every time some AI research produces a successful result, someone comes along and says "That's not true A.I" "It's just a computer program that has to be told what to do."
(This is the "no true Scotsman" argument.)
So let's see the list of such "non-AI" technologies:
- Natural-Language translation (getting pretty usable now)
- Speech recognition combined with ability to answer fairly arbitrary questions quite well on average.
(talking to Google via my Android phone)
- Self-driving cars (getting pretty close - will be better drivers than people on average pretty soon)
- Chess
- Jeopardy
- Go
- Detection of suspicious speech and patterns of communication (no doubt used by NSA on most Internet and phone traffic)
- Recognition of particular writer from all writers on Internet by analysis of their writing style
- Person identification by face picture recognition
- Object type and locaton type recognition from pictures
- Walking, box-stacking robot "Atlas 2"
Just algorithms.
Does it actually matter what you personally choose to call this kind of technology? It is what it is, and it's advancing quickly.
"It's not true AI" sounds like the desperate retreat cry of a person in a very defensive stance, afraid of losing a sense of human uniqueness.
Where are we going and why are we in a handbasket?
Yes. Amazingly it IS really hard to do AI. In fact it is so hard we arent even close and may never achieve it.
We may or may not ever have true AI, but a sufficiently advanced expert system would be able to fulfill most of the things people imagine they'd 'need' from an actual AI. (And I mean a very, very advanced expert system, probably a couple of decades away from where we are now. Throw a few hundred million dollars at the problem and I bet we'd make some serious progress towards it.)
But as for a true AI, I suspect it will happen eventually...the trick will be in recognizing and/or determining that it is truly "self-aware" (whatever that actually means).
Simple Turing tests may not suffice. Even though some of the current chatbot-type systems can converse passably for a little while, none can hold a genuinely sensible discussion on any abstract topic without stumbling and giving itself away rather quickly. I bet almost most people here could suss one out in fairly short order.
Wait another decade or two, however, and I suspect we'll see some expert systems that will be difficult to distinguish from a human operator, and in some fields, far more competent.
Just cruising through this digital world at 33 1/3 rpm...
"We know now that we don't need any big new breakthroughs to get to true AI."
This is so wrong that it's hard to know where to start.
Just cruising through this digital world at 33 1/3 rpm...
I don't count any of those things as AI, although they are components of AI since they are all pieces of (or combinations of) observing and interacting with the environment.
To me, "true AI" is something that can decide to do something other than that for which it was constructed. Can AlphaGo do anything other than play Go? If you tell it to play Go, can it decide, "No thanks, I'd rather cure cancer, it's a more rewarding problem"?
While AlphaGo and the like are very fantastic achievements, I don't think they are intelligence - they are "merely" very effective specialized problem-solving machines.
"There are a dozen opinions on a matter until you know the truth. Then there is only one." - CS Lewis (paraprhase)
If it's learning via playing sample games versus preprogrammed behavior, that is a step closer to true artificial intelligence. I like to see a better explanation of exactly how it's designed, and how many games it took to train it.
I've abandoned my search for truth; now I'm just looking for some useful delusions.
Aren't there, in fact, different rulesets for Go involving things like methods of scoring?
"but now we know that we have been on the right track"
Devalue the human race is the right track? I think many of you will come to agree with me when you're faced with an AI taking your job.
"If any question why we died, Tell them because our fathers lied."
So far, every post I have read that makes the same claim as yours lacks a critical piece: a clear description of what would qualify as AI.
Often, when I state that question, I get a long, rambling, disorganized list of random things humans do, and no indicator that making a computer do them would yet qualify as true AI. That is why I keep emphasizing the word "clear." Make it clear or you are just being religious.
So, exactly where are those goal-posts?
Despite the name "neural network", there is nothing "very similar" between the way AlphaGo works and brains work.
That seems correct. AlphaGo is playing go at a level beyond that of humans. The take home point seems to be that brains aren't really competitive and are probably a dead-end technology.
Game 2, move 37 was amazing. The commentator had already pointed out the issues with the two stones trapped to the lower/middle left and the loose group to the lower right. This move linked everything together in a light way. This was an "ear-reddening move."
There are, but they are inconsequential in a majority of games. Only in relatively strange situations do they affect the final score.
I think you're right as far as the cynics are concerned. People are worried that AI will somehow take over humanity. However, I think the examples you listed do not show (at least what I consider) true AI. But rather than argue whether a task falls into "true AI", let's see what tasks existing AI cannot do well, but we expect a reasonably intelligent human to do:
There are still major hurdles before these types of tasks can be done by a computer. The first requires extracting a basic principle (connecting), and reapplying it to similar problems. The second require knowledge (the AI needs to know what "grass", "growing" or "exciting" is), and the third requires constructing object representations of real-life phenomena (auto mechanic, car, oil change) and reasoning about their behavior (closing time, wanting to go home, working quickly). You'll note that this third one is very similar to what we call programming. So if this can be done, then we will have already achieved technical singularity.
I think all of the advances you listed are important and relevant, but they are still within the realms of pattern recognition, tree search, and classical programming (if (light.isRed() || objectAhead.isPedestrian()) { car.SlowToStop(); }). I strongly suspect mere recombinations of these techniques will not get us AI's that can solve more advanced problems. We need something new. And until we do, we do not have strong AI.
To me, "true AI" is something that can decide to do something other than that for which it was constructed
Many people can't even decide to stop eating.
The goal posts are simple. "True AI" = Consciousness.
It is the difference between a robot that can do a single activity better than any human, and a robot that can perform *all* the activities of human (which of course, means fitting into the same physical space as a human).
Is this a ridiculously high bar? Absolutely. But that's the snake-oil that's being sold in the popular press, so I figure it's fair game.
The goal posts are simple. "True AI" = Consciousness.
It's not a simple goal post if you can't provide a practical definition of consciousness.
and a robot that can perform *all* the activities of human
There are many people that can't perform *all* the activities of a human. Stephen Hawking can't even catch a ball.
I was watching 15 minute summary of match #1 yesterday by "Michael Redmond 9 dan professional and Chris Garlock" ... they were placing the stones on the grid and talking about where the position is weak and strong etc. ... what is not clear to me - did they memorize the whole game? I did not see them to use any notes or record of the game ...
Something that could choose its own problem domains and work on domains that help preserve its physical existence would be interesting.
The goal posts are simple. "True AI" = Consciousness.
It's not a simple goal post if you can't provide a practical definition of consciousness.
and a robot that can perform *all* the activities of human
There are many people that can't perform *all* the activities of a human. Stephen Hawking can't even catch a ball.
First, let's get this straight. "True AI" is a marketing term and it's bandied about by the media and people looking to get funding from the unsuspecting. It promises essentially something that is essentially indistinguishable from a human being.
You want a goal post - here's one. You put 3 AIs and 3 humans with no intellectual, language or cultural barriers in remote communication with each other constantly for 3 months in both personal and a professional manner. When the humans are unable to determine which are the other humans and which are the AIs, I think that's "True AI" or as I call it, "Hollywood AI".
There are many people that can't perform *all* the activities of a human.
Indeed, I'm certain "True AI" will reach the state of a live human in a vegetative state any time now...
Come on, we both know what I meant. No need to get pedantic.
In both cases, what's being sold is that computers will be able to do things that are nearly impossible for computers to do, ignoring the things that computers do exceptionally well, all in the name of chasing funding and/or headlines. I find it as irritating as someone trying to sell an oil rig as a ballet dancer. It utterly demeans the impressive accomplishments in AI in favour of the nearly impossible.
Come on, we both know what I meant.
Unfortunately, in the this debate, there are plenty of groups of people that don't understand what the other means.
A minah bird or a parrot may learn to repeat hundreds of human speech patterns which it has learned by listening.
Does the bird understand any of the individual words? Does it understand the meaning of the words as group? Can it rearrange the words into new coherent speech? Is the bird intelligent?
Once researchers decide to agree on a definition of what AI is, only then can we decide if that goal is reached by a particular project. Until then its just turtles all the way down.
"You must try to forget all you have learned. You must begin to dream." -- Sherwood Anderson
I think you're wrong. The problem is that current computers are vastly weaker than the brain at problems that a neural network is well adapted to learning. A secondary problem is deficiency in sensors and manipulators.
But what's really missing in the current AIs is a deep motivational stack. They are currently operating with a very shallow heap of motivations. E.g., if you were to ask Alpha Go why it bothered to play go, and it could even understand the question, it wouldn't be able to tell you. True, ask any go player, and the answer you get is generally a false statement due to lack of insight, but Alpha Go couldn't reach to that level.
I think we've pushed this "anyone can grow up to be president" thing too far.
How would this same AI model (i.e. not a retrained model) do in Chess?
Why can't it be retrained? A human Go player with no experience in chess wouldn't know the first move, but after observing some games to deduce the rules, then playing enough games to practice, they'd probably be competent.
Guess what; that's exactly how this system was trained. In fact, an earlier model taught itself to play dozens of old Atari games in the same way.
Could it hold a basic conversation? Identify a picture of a cat?
Actually, Google's context-aware voice recognition & response system is largely driven by a similar layered neutral network, as is it's visual search that can indeed identify a picture of a cat. And with a robot arm attached I wouldn't bet against the burger flipping either.
Strong AI it isn't, but neither is it a classical pre-programmed computer. "Intuitive" computing is the best description I've seen of modern neural networks - it weighs a large number of factors gained from experience and makes a decision.
Why would anyone engrave "Elbereth"?
We know now that we don't need any big new breakthroughs to get to true AI.
But, this isn't 'AI', it's just another 'expert system'.
It is a combination. It is a neural network guiding an expert system, and being guided by a go board state evaluator.
No. I'm sorry but i *don't* know what you mean. You didn't define consciousness. By my definition the Atlas robot showed consciousness.
What the current robots all lack is a deep motivational structure. Also the computers they run on are underpowered compared to human brains for the kind of task they are performing. This may be addressed by the "neural computers" that people keep talking about building.
P.S.: Consciousness is the ability to asses your own state and compare it with the external physical world, and react in such a way as to achieve your goals. (Please note that by this definition a thermostat connected to a furnace and an air conditioner has a minimal amount of consciousness though no intelligence to speak of. It probably has one unit of whatever we eventually decide to measure consciousness in, corresponding to one bit in information theory.)
I think we've pushed this "anyone can grow up to be president" thing too far.
I deny that humans are capable of general problem solving. They are quite capable of solving a finite set of problems, some with more difficulty than others. (An in what category do you place the proof of the four-color theorum, where every step could be done by a trained mathematician, but no mathematician could do the whole thing, because they couldn't hold the entire proof in memory? The proof was done by a human-computer cooperation. I think it's an edge case.)
I think we've pushed this "anyone can grow up to be president" thing too far.
Of your first example rudimentary forms have been exhibited this year. (Computers that learned to play various computer games by first watching someone else play, and then playing themselves until they succeeded.)
The second is one that people fail at all the time. (Your example was pretty clear, but there are still people who would miss it. And POEs law.)
The third example is even worse. It depends on specialist knowledge. At some shops the mechanics won't change your oil faster, they'll just keep your car overnight so they can finish the paperwork and go home on time. So here what is needed is learning by experience. Neural nets already handle this one, too, to an extent. And it's another one where people often fail.
I think we've pushed this "anyone can grow up to be president" thing too far.
I hope to see this AI take on other expert Go players, perhaps several at a time or a team of them.
This is a really exciting development, one I didn't expect to see in my lifetime and I really want to see how far it can be stretched.
But all isn't lost for human players as Fan Hui 2p helped improve the AI's playing after his 5-0 loss to it some months back and he says he now has new insight into the game and has improved his global standing from 633 to somewhere in the 300s since becoming familiar with AlphaGo's play.
Pain is merely failure leaving the body
Didn't think i'd see this happening for a long time. Wonder whats next now...
Sorry, but no human can perform all the activities a human can. Can you do quantum physics, compose symphonies, play Go, manage a hedge fund, and write software to do all of those on your own? Hell, most humans still can't understand algebra, some of whom even live in the first world and think they know enough about economics to vote. And some of those humans can't fit into the same physical space as a human.
So yes it is a ridiculously high bar. Furthermore, no human can do any of those things without being trained for decades but there is an expectation (not necessarily coming from you) that "true AI" is supposed to be able to create new things ex nihilo.
The point is that AI is ARTIFICIAL intelligence. Just like an artificial leg, you would not say "that artificial leg is not a true artificial leg because it can't do all the things a leg does". We are trying to create AIs, not Is. We have not created true consciousness that is true. But leaving aside questions on what is consciousness and if it even exists, what does consciousness have anything to do with intelligence? People who are conscious have committed crimes stupidly and get caught. Are they more intelligent than a computer that can play Go?
Those who do not learn from commit history are doomed to regress it.
If he told you where the goal-posts were, you wouldn't be able to know how fast they're moving.
You're probably thinking of that one guy who used machine learning on a bunch of NES games. Frankly, there's not much intelligence in that. Games like Super Mario can be won by a carefully scripted series of moves executed at the exact right time. You don't even need to be looking at the screen. In other words, it's a search problem that can be brute forced (or solved with something similar to A* search).
People do fail at detecting sarcasm. But that's not the point. The point here is that people understand watching grass grow is boring even if they never tried it, and software doesn't. Likewise, the fact that you can argue about bringing in cars towards the end of the day means you have a mental model of how an auto shop works. You can see possibilities that I didn't see because your model is different from mine. The model can answer other questions too. For example, do auto shops have car jacks? Do they have WD-40? Do they use accounting software? Now how did you come up with the answers? Did someone tell you or did you infer it somehow?
The interesting thing is, even with just a few experiences with an auto shop, people can create this mental model and make reasonably accurate predictions about how it works. Computers don't do that.
The future is here: "Bacon. Bacon! It's Bacon! Bacon Bacon Bacon! IT'S BACON!!!"
Love sees no species.
How does AlphaGo feel about it's victory? i bet it's ecstatic.
While I can't argue with your exact statements here, to me it sounds more like collected lifetime experience than intelligence.
OTOH, that does bring up another point. People have deep analogy detectors that work in general cases. I'm not certain that these AI programs do. And unlike the lack of deep motivational structures, that DOES seem to me to be a part of intelligence.
I think we've pushed this "anyone can grow up to be president" thing too far.
I'm with the other guy: If I can't hold a believable conversation with it, then it's not 'artificial intelligence' at all, whatever that's supposed to mean. What I want to see is more like 'artificial sentience'. When they come up with a man-made machine that is indistinguishable or at least on a par with a human being, then I'll be impressed. So far they haven't even been able to come up with a reasonable facsimile of how a dog behaves.
Are YOU using the TOOL, or is the TOOL using YOU? Think about it!
Yes. Amazingly it IS really hard to do AI. In fact it is so hard we arent even close and may never achieve it.
Ah, so it's not intelligence at all unless it's human intelligence? It can never be sentient unless it engages in conversations exactly like we do, gives its own purpose just like we do, and only does tasks which humans can do? That doesn't sound very intelligent to me at all. "Oh sure, it's self aware and everything - but only if it does exactly this, behaves exactly like we do, and is completely like humans in every way shape and form." You seem to not like algorithms, even completely self-modifying, because they follow a set path. But aren't you setting it exactly one route, whose only purpose is to imitate us?
Perhaps a little analogy might help. You have a passing familiarity with bees and ants, yes? You are probably aware that there is one queen in the colony and that every other ant does everything they can to protect the queen. Would you consider those drones intelligent? Probably no, because they would obey orders without question and only do one thing. And yet, while not up to our level, you cannot deny that ants possess at least some level of understanding. They know who their queen is, they recognize each other. They understand their place in a society. They know that their life is worthless compared to their queen, and will lay down their lives to protect it. So why should the computer have to do everything that we do, exactly like we do, and only like we do? The computer may not socialize, and that's perfectly fine. But could you have done well at Go? I don't mean to play the game, but could you as a person physically evaluate all those moves? Can you grasp the meaning behind each one, the subtle shifts and changes?
The English word "intelligence" is a derivative of a Latin word "intellegere", which means "to understand". Could you understand all of that? I don't think you could. Sure, you could learn the rules, you could explain to me some basic plays, and after many years you might even beat a decent opponent. But I highly doubt you could ever understand the game at a level as this, and truly grasp the meaning behind each play. In other words, I don't think you could understand it, no matter how long you played, and indeed most people probably couldn't. We accept a common notion that some people are born with skills, that they have prodigious gifts, and that they see things the way we never will. Albert Einstein and Steven Hawking perceive the world much different than you or I do. This machine will never make friends, it will never betray someone for their own goals, and it will never grasp at a higher meaning in terms of existence. BUT, it DOES seek a higher meaning in this game - to evaluate it, to play it, to understand each and every little nuance. It evaluates possible moves, it could apparently learn from them, and it appeared to adapt well to its opponent during the fight. If you don't define seeking a higher understanding and self improvement as intelligent, then what, praytell, do you? Morals and ethics? Many psychopaths live in this world and are considered intelligent without them.
Look, I'm not saying this machine is going to take over the world. I'm not saying it has the ability to wonder why it exists. But you have a very strict and narrow minded view of what "AI" should be, and you don't appear to have an open mind at all on this subject. If your machine seeks only to imitate our thoughts and behaviors, doesn't that make it less intelligent then this machine right here? This one at least does seek a higher purpose and is (partially) self aware - to improve itself, it had to learn where it went wrong and what mistakes it made. It didn't copy behaviors of humans, or absorb the mindset of the people who programmed it: it looked at millions of games and judged for itself what was a good play and what was a bad one, and in what situations to use each.
And that, sir, is pretty damn intelligent.
"Set a man a fire, he'll be warm for the rest of the night. Set a man afire, he'll be warm for the rest of his life."
Again, I distinguish between the field of "AI", and what gets bandied around as "True AI", which is essentially something indistinguishable from a human being.
Its the difference between chess (~40 possible moves each turn), go (hundreds of possible moves) and reality (millions of possible outcomes).
We're a million miles away from that, and more to the point, it's ludicrous to spend resources trying to get there. It's like trying to make an internal combustion engine do ballet instead of concentrating upon what an internal combustion engine does well.
Even more pernicious is trying to sell the idea that "True AI" is sometime in the foreseeable future. It makes for good books (because really, we want a computer that we can empathize with), but it makes terrible reality, which is why it's mostly used to scam money from government organizations and gullible people or companies that don't understand AI.
We are trying to create AIs, not Is. We have not created true consciousness that is true. But leaving aside questions on what is consciousness and if it even exists, what does consciousness have anything to do with intelligence?
I think we're mostly in agreement. I have a lot of respect for AI.
However, the whole "True AI" business is hogwash because it promises (mostly to the uninformed) something that approximates all aspects of human intelligence (or god help me, the singularity where we upload ourselves).
It's bunk because if you're building something to move a person, you build a vehicle, not a replacement for a leg. There's no reason to try to build human beings, and it's probably 10^10 harder to build a general human-like processor than to do what we do with Go, etc.
I have no problem with "AI". It's been around for ages and we're making impressive advancesments. I have a huge problem with "True AI" or "Strong AI" or whatever the buzzword is for something indistinguishable from human thought processes. That we're probably centuries away from. (We're probably centuries away from a deep understanding of human intelligence. Every time I think we're getting closer, I read the literature to find that we're not even at "phlogiston" levels of understanding about the human brain. We're not about to reproduce human intelligence.
(At least if we're lucky. If we're not, we've just obsoleted every human being on the planet at the same time that we learn that it's a completely moral act to eliminate consciousnesses anytime its convenient to do so - not a good combination.)