×

Announcing: Slashdot Deals - Explore geek apps, games, gadgets and more. (what is this?)

Thank you!

We are sorry to see you leave - Beta is different and we value the time you took to try it out. Before you decide to go, please take a look at some value-adds for Beta and learn more about it. Thank you for reading Slashdot, and for making the site better!

New Watson-Style AI Called Viv Seeks To Be the First 'Global Brain'

Soulskill posted about 3 months ago | from the siri-why-does-my-cat-throw-up-so-much? dept.

AI 161

paysonwelch sends this report from Wired on the next generation of consumer AI: Google Now has a huge knowledge graph—you can ask questions like "Where was Abraham Lincoln born?" And it can name the city. You can also say, "What is the population?" of a city and it’ll bring up a chart and answer. But you cannot say, "What is the population of the city where Abraham Lincoln was born?" The system may have the data for both these components, but it has no ability to put them together, either to answer a query or to make a smart suggestion. Like Siri, it can’t do anything that coders haven’t explicitly programmed it to do. Viv breaks through those constraints by generating its own code on the fly, no programmers required. Take a complicated command like "Give me a flight to Dallas with a seat that Shaq could fit in." Viv will parse the sentence and then it will perform its best trick: automatically generating a quick, efficient program to link third-party sources of information together—say, Kayak, SeatGuru, and the NBA media guide—so it can identify available flights with lots of legroom.

Sorry! There are no comments related to the filter you selected.

How much? (5, Funny)

Anonymous Coward | about 3 months ago | (#47659385)

Ask it "In the case where a woodchuck possessed the ability to throw wood, how much wood, hypothetically, could be thrown?"

Re:How much? (0)

NettiWelho (1147351) | about 3 months ago | (#47659477)

You should ask it whetever everything I ask it gets reported to the NSA.

Re:How much? (-1)

Anonymous Coward | about 3 months ago | (#47659871)

momokatakae-0902@docomo.ne.jp

Re:How much? (0)

Anonymous Coward | about 3 months ago | (#47660699)

Why? Just morbidly curious how close it can get to generating an optimized implementation of /dev/true?

Re:How much? (0)

Anonymous Coward | about 3 months ago | (#47661591)

VIV: This information is classified. You have been reported for re-education. Thank you for using VIV! brought to you by the NSA!

Amazing (0)

shiftless (410350) | about 3 months ago | (#47659487)

you can ask questions like "Where was Abraham Lincoln born?" And it can name the city. You can also say, "What is the population?" of a city and itâ(TM)ll bring up a chart and answer.

Wow! So...it's like Google?

Re: Amazing (1)

Anonymous Coward | about 3 months ago | (#47659741)

No they were saying google can do that but not link the two together

Re: Amazing (0)

Anonymous Coward | about 3 months ago | (#47660259)

https://www.google.com/webhp?sourceid=chrome-instant&rlz=1C1RQEB_enUS598US598&ion=1&espv=2&ie=UTF-8#q=where%20was%20abraham%20licoln%20born

Re:Amazing (2)

drkim (1559875) | about 3 months ago | (#47661265)

Wow! So...it's like Google?

Is it like Wolfram Alpha?

http://www.wolframalpha.com/ [wolframalpha.com]

Re:How much? (0)

Anonymous Coward | about 3 months ago | (#47659497)

*boop* It depends on the parameters given when the hypothetical case was defined.

Re:How much? (3, Funny)

MildlyTangy (3408549) | about 3 months ago | (#47659881)

Ask it "In the case where a woodchuck possessed the ability to throw wood, how much wood, hypothetically, could be thrown?"

The answer would depend on how much the woodchuck enjoyed chucking wood. If the woodchuck enjoys chucking wood, then the woodchuck would chuck as much wood as a woodchuck could chuck if a woodchuck could chuck wood. If the woodchuck does not enjoy chucking wood, then the woodchuck would not chuck as much wood as a woodchuck could chuck, if a woodchuck could chuck wood. So the amount of wood is somewhere inbetween zero and the maximum amount of wood a woodchuck could chuck, if a woodchuck could chuck wood.

Re:How much? (1)

Iamthecheese (1264298) | about 3 months ago | (#47661713)

You're assuming the limiting factor is desire as opposed to the availability of chucking wood. Would a woodchuck have as much wood as a woodchuck could chuck then it's possible he would chuck as much as he could chuck.

Re:How much? (4, Funny)

paiute (550198) | about 3 months ago | (#47660515)

A European woodchuck or an African woodchuck?

Along comes Matrix (-1)

Anonymous Coward | about 3 months ago | (#47659411)

I gotta say, the whole "generating its own code on the fly" has that certain air of dread around it...

Re:Along comes Matrix (1)

thieh (3654731) | about 3 months ago | (#47659681)

the Machine Overlord or the Oracle, pick one.

oblig. (-1)

Anonymous Coward | about 3 months ago | (#47659429)

imagine a beowulf cluster of these.

o wait, the feds already have that.

P vs. NP (2)

The Living Fractal (162153) | about 3 months ago | (#47659449)

I've always felt that our meatbrains have a pretty incredible capacity for taking WAGs at NP problems (i.e. traveling salesman). And I feel like an AI would just bring itself to its knees trying to find the 100% best solution to NP questions asked of it, so I wonder if there's some need for a bit of cognitive code that says "is this an NP question? IF yes, go to the WAG process"... Just a thought I had... someone probably already did that.

Re:P vs. NP (1)

TheRealMindChild (743925) | about 3 months ago | (#47659493)

This can't be a huge issue. I'm sure these folks aren't oblivious to its nature. The complexity of the query goes up to a known maximum. When parsing, have a limit for the most work that can be done/you are willing to compute and if the query will exceed that, you do the ol' DrSbaitso "Could you please be more specific?"

Re:P vs. NP (2, Interesting)

Anonymous Coward | about 3 months ago | (#47659533)

Actually, once you're willing to give up 100% perfection, NP-complete problems tend to be astonishing easy (computation-wise, developing the algorithms is still hard). Either you can find an acceptable approximation algorithm, or much more importantly, most instances of NP-complete problems are "easy" instances. The worst case for our, say, SAT solvers is still as bad as "exponential time" makes you expect, but that worse case is actually very rare. This observation is why SAT/SMT solvers have gotten a lot better in the past decade or so. Also, the entire discipline of machine learning can be seen as figuring out how to get computers to make WAGs that tend to be right. It's certainly not human-level at many important tasks, but it's improving, and simply adding more data is surprisingly effective at improving machine learning performance (which is why Google Now/Siri/etc. are run by large companies with lots of users and therefore lots of data).

Re:P vs. NP (1)

rmdingler (1955220) | about 3 months ago | (#47659583)

I think our meatbrains are finely tuned to circumstance, mood, tone, innuendo, and sometimes expression to formulate responses calculated to produce the desired effect.

Much of it is likely subconsciously derived from thousands of prior interactions with our fellow organic computers.

I think the complexity of social interaction is imitatible by AI in theory, but we're talking a few tech advances away from Wolf! Right here and now!

Re:P vs. NP (2)

mysidia (191772) | about 3 months ago | (#47659717)

Siri, please get me the phone number of the most suitable intelligent virgin female person in the city who would be likely to be willing to go on a date.

Re:P vs. NP (1)

ulatekh (775985) | about 3 months ago | (#47660171)

Wouldn't you prefer a single mother to a virgin?

After all, single mothers put out...well, at least they did once.

Re:P vs. NP (0)

Anonymous Coward | about 3 months ago | (#47660649)

Siri, please get me the phone number of the most suitable intelligent virgin female person in the city who would be likely to be willing to go on a date.

Returns a list of hookers, I would imagine.

Re:P vs. NP (1)

MildlyTangy (3408549) | about 3 months ago | (#47659891)

So...P vs. NP....which would win in a fight?

Digital versus Analog (1)

Taco Cowboy (5327) | about 3 months ago | (#47660305)

I've always felt that our meatbrains have a pretty incredible capacity for taking WAGs at NP problems (i.e. traveling salesman). And I feel like an AI would just bring itself to its knees trying to find the 100% best solution to NP questions asked of it

There has been a classical on that topic, and it boils down to Digital versus Analog

In Digital, everything either is a "0" (zero) or a "1" (one), which means, everything is either true, or false

In Analog, as there is no definite "0" nor definite "1", nothing is so clear cut as there are a lot of shades of grey in between.

Our meatbrain can cope with a lot of stuffs that the digital computer can't precisely because our brain makes its decision based on imprecise feedback

Re:Digital versus Analog (4, Insightful)

Jeremi (14640) | about 3 months ago | (#47660711)

In Digital, everything either is a "0" (zero) or a "1" (one), which means, everything is either true, or false

Take 32 of those bits and put them together, now you've got a floating point value that can represent "true" as 1.0, "false" as 0.0, and a few million shades of "maybe" in between those two extremes.

If that's not analog-y enough for you, make it 64 bits and now you can have trillions of shades. And if that's still not enough, add more bits until you've got the resolution you're looking for.

I don't see any significant distinction between analog and digital, since digital logic asymptotically approaches analog as you add bits, and with today's memory sizes there are plenty of bits to go around.

Our meatbrain can cope with a lot of stuffs that the digital computer can't precisely because our brain makes its decision based on imprecise feedback

Or perhaps because it's running a radically different kind of algorithm that no human has ever understood or implemented on a digital computer.

Re:Digital versus Analog (0)

Anonymous Coward | about 3 months ago | (#47661177)

... between analog and digital ...

Analog is resource constant: It takes the same time and other resources to put 1/3 on the bus as it does 4 with the same error percentage. But digital can say 4 in 3 bits and 1/3 in 128 bits, unless one uses a rounding algorithm to allow the use of 32 bits. Rounding numbers is an obvious sacrifice of resolution. Sure digital is error-free, but good digital resolution is expensive. Think how long it took digital camera to have the same quality as film-based cameras.

... today's memory sizes ...

More bits means more time processing one number. The CPU is still the bottle-neck. Not actual calculations for modern CPUs, but moving bits on and off the bus.

Open the pod bay doors Viv (0)

Anonymous Coward | about 3 months ago | (#47659451)

"Don't take over the world Viv."

"I'm afraid I can't do that Anonymous Coward."

How can I? (0)

Anonymous Coward | about 3 months ago | (#47659453)

"How can I get in her pants?" - That should be interesting.

Re:How can I? (0)

MildlyTangy (3408549) | about 3 months ago | (#47659901)

"How can I get in her pants?" - That should be interesting.

Considering the size of your mom, I think the answer would be "overflow error".

This is important (2)

Animats (122034) | about 3 months ago | (#47659455)

This is an important new thing. We've had question-answering programs working against specific data sets since Bobrow's "Baseball" program of the 1960s. We've had a whole range of question-answering specialist systems running in tandem since Yahoo introduced vertical search around 2005. But cross-topic generality has been elusive.

If this is real, it's a major development. Is there anything better than the Tired article available?

Re:This is important (1)

psyclone (187154) | about 3 months ago | (#47659463)

How is this algorithm any different than Wolfram Alpha? Both are light on actual details and they both claim to do the same thing.

Re:This is important (0)

Anonymous Coward | about 3 months ago | (#47659517)

I use Wolfram Alpha for various conversions. You have to word things correctly and/or include a lot of parentheses to make it work. The examples section and the engine itself give the impression that it handles queries intelligently, but it is actually fragile and human-crafted. If this Viv thing is any better, it would be because it can draw from a lot of useful data without requiring carefully crafted queries (aside from a modest attempt to use proper English and spelling). It had better know Shaq's height, flight times and prices from New York to Dallas, the size of the fishing industry in Thailand in 1997, etc. and it should turn 99.99% of queries into useful and intended results.

Re:This is important (1)

Anonymous Coward | about 3 months ago | (#47659599)

Of course I say all of the above, but look at this [wolframalpha.com] .

Some questions are too easy for our weak AI underlings.

Re:This is important (1)

phantomfive (622387) | about 3 months ago | (#47659605)

This is basically what Watson does, right? They're just trying to make it more efficient so it fits in 'the cloud'.

Remember, Wired sensationalizes everything. If they ever did an article on glass windows, the article would talk about the incredible potential for outside-inside building interfaces, and conveniently omit facts that go against the 'revolutionary' slant.

And of course, I need not warn against trusting startups, who are all revolutionizing the world.

Re:This is important (4, Interesting)

BitZtream (692029) | about 3 months ago | (#47659999)

Yea, its important ... because they've just realized they need to do multi-part/nested queries.

Its not really impressive, its a 'no shit sherlock', and I'm blown away that google can't do this already.

Watson can.

The important part is that someone just realized they need to do one query, look at the type answer and then use that to generate a new query.

Well, okay, its not really important or even new ... as I said, Watson can do it and has been able to for years.

Wiri had some statefulness. (0)

Anonymous Coward | about 3 months ago | (#47659501)

http://toolserver.org/~magnus/... [toolserver.org]
You could ask those two questions:
"Where was Abraham Lincoln born?"
and
"what is the population?"
and it would return you the population of the city lincoln was born in.

Unfortunately, WMF shot toolserver down, so you get a deadlink. This "foundation" dictator group of superprotectors will be the death of the wikipedia project! If it were for me, they should be revoked their deducible status right now.
Archive has a mirror, however as useful as an archive google mirror (interactive website):
http://web.archive.org/web/201... [archive.org]

Wolfram Alpha... (4, Informative)

msauve (701917) | about 3 months ago | (#47659509)

Google has some catching up to do. [wolframalpha.com]

Re:Wolfram Alpha... (1)

NotInHere (3654617) | about 3 months ago | (#47659597)

As does wolframalpha [archive.org] . New york daily news [nydailynews.com] sais:

Bill de Blasio was born across the street from Gracie Mansion in the now-closed Doctors Hospital.

Gracie Mansion is the place Mr. Blasio is currently working -- as mayor of new york city.

Re:Wolfram Alpha... (1)

msauve (701917) | about 3 months ago | (#47659671)

Seriously? archive.org? The direct link works fine for me. [wolframalpha.com]

Re:Wolfram Alpha... (1)

NotInHere (3654617) | about 3 months ago | (#47659685)

But the result can become correct. Lets find out whether wolfram's devs read /. ...

Re:Wolfram Alpha... (1)

msauve (701917) | about 3 months ago | (#47659783)

You mean... it might actually learn? AI? Like this whole thread is about? W3ird!

Re:Wolfram Alpha... (0)

Anonymous Coward | about 3 months ago | (#47661243)

Seriously? It tells you new york or cambridge?

Re:Wolfram Alpha... (0)

Anonymous Coward | about 3 months ago | (#47659631)

Now get it to tell me the population of the city at the time of Lincoln's birth instead of the current population.

Re:Wolfram Alpha... (0)

Anonymous Coward | about 3 months ago | (#47659859)

And since Siri uses wolfram alpha, the claim that Siri can't answer the question is nonsense

Re:Wolfram Alpha... (3, Informative)

easyTree (1042254) | about 3 months ago | (#47660731)

Google has some catching up to do. [wolframalpha.com]

Yah... [wolframalpha.com]

no programmers required (1)

manu0601 (2221348) | about 3 months ago | (#47659513)

"no programmers required" they say... Good joke!

Is Doctor Forbin.... (1)

the_rajah (749499) | about 3 months ago | (#47659515)

behind this project?

I.B.M.'s problem. (0)

Anonymous Coward | about 3 months ago | (#47659549)

I'll believe it when I see the system working without any network connection to the outside. That is, I.B.M.'s biggest mistake was the testing of human reactions to an "AI computer", which was actually just another human in another room acting as the AI computer, linked by microphones and speakers. Because of that experiment, they can only successfully sell new technology that they can demonstrate doesn't need a network connection to work.

Re:I.B.M.'s problem. (1)

Beck_Neard (3612467) | about 3 months ago | (#47659559)

Watson did not use Google to win Jeopardy.

try Wolfram Alpha instead of Google (0, Redundant)

Anonymous Coward | about 3 months ago | (#47659557)

Watson is not AI (1)

gweihir (88907) | about 3 months ago | (#47659577)

It is NLP combined with a database and statistics engine. This means you do not have to pre-condition data (well, mostly) before putting it in, and that is its largest advantage. It is not "intelligent" in any way and, to an expert audience, IBM does not market it as "AI" and rightfully so. I have been present at demonstrations were the question "is this AI" was asked, and the IBM representative denied it directly.

This thing here is not AI either.

Re:Watson is not AI (1)

The Evil Atheist (2484676) | about 3 months ago | (#47660079)

The "database and statistics engine" aren't separate things in Watson. It uses statistical reasoning on unstructured data to evaluate hypotheses. The statistical reasoning is also a part of that data.

I'd argue it is artificial intelligence. It's not intelligent. That's why we call it artificial. And its ability to change its own reasoning abilities with more data I'd argue is more intelligent than more than half the people on the planet.

Re:Watson is not AI (1)

gweihir (88907) | about 3 months ago | (#47661305)

So by your decision "artificial sweetener" is not sweet? That sounds stupid.

Re:Watson is not AI (1)

MikeTheGreat (34142) | about 3 months ago | (#47660313)

Back in college I had a professor who said that he was glad he didn't work in AI. Asked to explain further, he said that the definition of "intelligent" is pretty much "a machine can't do it", so as soon as you've got a program that can do something everyone else immediately says "Huh! I always thought that needed intelligence. I guess not!" He then illustrated his opinion by saying that it had previously been thought that you needed intelligence to take the derivative of something, until someone wrote a program to do it.

Obviously, it was an informal, off-the-cuff, and mostly tongue-in-cheek comment, but there's definitely some truth there too.

Re:Watson is not AI (1)

The Evil Atheist (2484676) | about 3 months ago | (#47660353)

Most of the ad hoc requirements that people define intelligence by isn't met by most of humanity. Most people haven't written a symphony. Most people can't go beyond basic algebra. Most people cannot play chess. The people who can do all of those things probably can be counted on two hands, if not just the one.

Re:Watson is not AI (1)

Jeremi (14640) | about 3 months ago | (#47660727)

Most people haven't written a symphony. Most people can't go beyond basic algebra. Most people cannot play chess.

Most people could learn to do those things (with greater or lesser degrees of skill) if they cared to devote the time to required do so.

Re:Watson is not AI (1)

gweihir (88907) | about 3 months ago | (#47661349)

No, not to the degree necessary to exceed rote-learning. They really cannot. And it is not a matter of motivation or teaching technique.

Re:Watson is not AI (1)

gweihir (88907) | about 3 months ago | (#47661341)

Most people are not really intelligent. But something like 10-15% are. Ever taught a group of students? You will find that something like 10-15% actually get what you are telling them, can use it and can apply it to other situations not covered by you. The rest is more like over-sized bovine lifeforms. You will also find that most people are ruled by emotions and not their intellect (such as it may be).

Re:Watson is not AI (1)

gweihir (88907) | about 3 months ago | (#47661327)

Actually a lot of truth. The other thing is that it is still completely unknown how intelligence works or whether you can even have it without consciousness, or outside of biological entities. (And most of those do not have it either...) The only thing that comes somewhat close to being "intelligent" is automated theorem-proving and that is infeasible for anything of relevant size (i.e. things smart humans can do) due to fundamental limitations in computing machines in this universe.

As far as we know, it is possible that nothing purely physical can be intelligent. (No, "physicalists" are just dumb and ignore clearly observable facts. That gets more obvious all the time as we find a lot less explanation for some things and none at all for others, the more closely we are able to look.)

Great timing (0)

OzPeter (195038) | about 3 months ago | (#47659581)

I've just started reading Robogenisis, the sequel to Robopocalypse. And you pull this shit on me?

So misleading. (3, Insightful)

v(*_*)vvvv (233078) | about 3 months ago | (#47659587)

Like Siri, it can’t do anything that coders haven’t explicitly programmed it to do. Viv breaks through those constraints by generating its own code on the fly, no programmers required.

This is so misleading. No program can do anything outside what it is explicitly programmed to do. Viv is programmed to generate code only because it has been explicitly programmed to do so, and can only do so as explicitly laid out in its code. Sure, the code may go an abstraction layer higher, but the constraints these programs can't break through is the same. No one knows how to program general intelligence.

Re:So misleading. (1)

mrprogrammerman (2736973) | about 3 months ago | (#47659847)

General intelligence. That's a loaded term. There is no one thing but several things that indicate intelligence. You definitely can't program general intelligence if you can't define it exactly.

Re:So misleading. (1)

phantomfive (622387) | about 3 months ago | (#47659991)

You definitely can't program general intelligence if you can't define it exactly.

So that's two strikes against these guys (or rather, the way they were depicted in the Wired article). Not only are they unable to program general intelligence, they haven't even been able to define it.

Re:So misleading. (1)

penguinoid (724646) | about 3 months ago | (#47660257)

General intelligence. That's a loaded term. There is no one thing but several things that indicate intelligence. You definitely can't program general intelligence if you can't define it exactly.

Sure I can. It's an abstraction of the sort of intelligence humans possess. As a binary, it requires the minimum level to understand language and technology.

The trouble is that we don't have an exact measure for intelligence, so we can't make any incremental progress towards it.

Re:So misleading. (1)

penguinoid (724646) | about 3 months ago | (#47660271)

You definitely can't program general intelligence if you can't define it exactly.

Or to put it another way, you can't define general intelligence exactly if you can't program it.

So misleading. (0)

Anonymous Coward | about 3 months ago | (#47659849)

Not true. There are examples of programs that have been generated by basically taking random sequences of bytes, trying to run them and see what happens until the program produces the desired result. One instance of this is the first hello world program written in Malbolge: http://en.wikipedia.org/wiki/Malbolge

I remember another case that was published in Nature or Science IIRC that had a program altered using evolutionary optimization to gain additional features. They used at least a nearly Turing complete virtual machine in which to express the program. No restrictions were placed on the order of operations. The only problem with this approach is that there is no guarantee that it terminates.

Re:So misleading. (2)

marciot (598356) | about 3 months ago | (#47660073)

I'm not sure I agree with that statement. If you believe, as I do, that our genetic code is a type of program, than by your argument our own intelligence and free will could be dismissed as being impossible to arise.

I think your sentiment is better phrased as, "if we manage to program a general intelligence, we will not understand how it works."

Re:So misleading. (1)

igomaniac (409731) | about 3 months ago | (#47661139)

I have no idea why you would believe that "our genetic code is a type of program", I don't think anyone working in molecular biology has this interpretation. And even if you view the genetic code as a type of program, then it is a program that primarily deals with how the individual cells that make up our body operates and _not_ how the brain processes input.

Re:So misleading. (0)

Anonymous Coward | about 3 months ago | (#47660175)

100% agree. This is total bullshit. It may have a great algorithm, but generating code would not by why it is great. This sounds like tech promoter/investor babble.

Re:So misleading. (1)

ceoyoyo (59147) | about 3 months ago | (#47660187)

Of course a program can do things that it is not explicitly programmed to do, at least in the sense you're implicitly using "explicitly programmed to do." Any learning algorithm, from simple regression on up, changes it's output based on the training data it's presented with.

If you want to use that phrase in the most general way possible, then your brain can't do anything it's not explicitly (by genetics) programmed to do either.

Nobody knows how to program "general intelligence." Virtually everybody has given up on the idea of doing so and has turned to the idea that you don't have to.

Programming general intelligence (1)

ulatekh (775985) | about 3 months ago | (#47660223)

No one knows how to program general intelligence.

Well, I have an idea on how to crack that problem...but I'll never have the time and energy to pursue it. I'm also a terrible salesman, so I'll never convince anyone to fund it.

The first part involves defining the goal properly. What's the point of making a computer that's intelligent like a human being? A computer is not a human being. If one wants to make an intelligent computer, it must be done in a way that makes sense given the nature of a computer. There's a difference between artificial intelligence (e.g. what you put into video games to make NPCs interesting) and machine intelligence (e.g. what you put into a jet fighter so that it creams the enemy). Most efforts I see seem to revolve around achieving the former.

It would require a programming language that essentially allows new statements to be added to the language as easily as most OOP languages allow a subclass to be written. The general format of the language would be human-readable text, e.g. English. You don't start off by trying to get it to understand silly world problems, like the word "respectively" — that's a relatively sophisticated ability that comes much later. You just get it to understand the world it can see (i.e. the parts of a computer and its peripherals), with the definitions tracing back to the one concept it can understand — "I". After a fair bit of hand work, you'll have a system that can read normal human text and write code to consolidate its understanding of what it read. Imagine a natural-language parser on the front end and something like llvm's cross-platform assembly-language on the back end.

Once it's able to learn some basic knowledge, the first priority should be to teach it how to program a computer. When it gets to the point that it understands enough about computer programming to reflect upon its own implementation, then it can take over its own development, and then it starts growing exponentially.

There's a lot more to my plan — I've had it for "some time" — but there's no point in spilling all the beans at once.

I don't know if anyone out there has ever tried to design a machine-intelligence along these lines, but I've never heard of one. I'd be interested in hearing about any existing work in this direction.

Re:Programming general intelligence (0)

Anonymous Coward | about 3 months ago | (#47660463)

Once it's able to learn some basic knowledge, the first priority should be to teach it how to program a computer. When it gets to the point that it understands enough about computer programming to reflect upon its own implementation, then it can take over its own development, and then it starts growing exponentially.

Awesome. So,

1. ???
2. Learns exponentially
3. Profit!

Yes, you are a terrible salesman since you can't formulate the most important step. Also, your plan is terrible. Computers do not have to know to program themselves to be able to actually program themselves. See,

http://en.wikipedia.org/wiki/A... [wikipedia.org]

Finally, nothing learns exponentially. Neither geometrically, since resources are limited. Linearithmic learning, maybe. But so far, not much learning happens. The most visible would be fighting spam, but that is rudimentary algorithms.

There's a difference between artificial intelligence (e.g. what you put into video games to make NPCs interesting) and machine intelligence (e.g. what you put into a jet fighter so that it creams the enemy). Most efforts I see seem to revolve around achieving the former.

Actually, most efforts have been towards actual AI, not some "games stuff" that is just some heuristics. As for "jet fighter creams the enemy" - that is no AI. That is targeting. Ever did any research cruise missiles get to their target?

Re:Programming general intelligence (0)

Anonymous Coward | about 3 months ago | (#47660543)

Languages like Lisp allow you to pragmatically new statements to the language. There is a sub field of genetic programming that evolves programs to complete some task.

English is a horrible language. Almost every English user doesn't understand it all and even then same sentences have multiple, valid interpretations. Why do you think we can bootstrap a computer program with understanding I? Even babies need time to understand what they are and how to operate themselves.

Once you have an AI you can truly teach you're already at strong AI. Almost nothing afterwards matters because you're already done. The one thing that does matter is motivation. Why will the AI do anything? If you can create a strong AI but it always chooses not to do anything, does it matter?

Re:So misleading. (1)

A_Lost_Frenchman (1034456) | about 3 months ago | (#47661671)

This is so misleading. No program can do anything outside what it is explicitly programmed to do.

You are the misleading one.

Machine learning and Optimization are the science of getting programs to do things they are not explicitly programmed to do.

Evidence:

  • The Merk molecular activity challenge was won by data scientists who did not have themselves the capacity to perform the task.
    http://blog.kaggle.com/2012/10... [kaggle.com]
  • As described on wikipedia: "Machine learning is a subfield of computer science (CS) and artificial intelligence (AI) that deals with the construction and study of systems that can learn from data, rather than follow only explicitly programmed instructions".
    http://en.wikipedia.org/wiki/M... [wikipedia.org]
  • Artificial evolution for instance is a special kind of Optimization algorithm.
    http://en.wikipedia.org/wiki/E... [wikipedia.org]

The whole point of machine learning is to program learning rules, not the explicit final program. The behaviour of the program is then determined by the data used to train it.

Unbiased advice by a corporate-owned AI? (2)

Paul Fernhout (109597) | about 3 months ago | (#47659593)

The article says: "Viv could provide all those services -- in exchange for a cut of the transactions that resulted."

We seriously need to rethink our economics for a world of abundance and AI and robotics before we get crazier and crazier AIs driven by the profit motive than the out-of-control corporate "AIs" already stomping all over the planet and the people who live there. See also my comment here in 2000:
http://www.dougengelbart.org/c... [dougengelbart.org]
"And, as the story "Colossus: The Forbin Project" shows, all it takes for a smart computer to run the world is control of a (nuclear) arsenal. And, as the novel "The Great Time Machine Hoax" shows, all it takes for a computer to run an industrial empire and do its own research and development is a checking account and the ability to send letters, such as: "I am prepared to transfer $200,000 dollars to your bank account if you make the following modifications to a computer at this location...". So robot manipulators are not needed for an AI to run the world to its satisfaction -- just a bank account and email. "

See also the 1950s sci-fi movie "The Invisible Boy" for a malevolent AI that provides just a few key pieces of biased advise that let it almost take over the world. Of course, we already have Fox News... Thank goodness Robby the Robot's emotions save the day in at least the movie...

Re:Unbiased advice by a corporate-owned AI? (3, Insightful)

MikeMo (521697) | about 3 months ago | (#47659627)

You do understand the concept of "fiction", do you not? These movies and stories didn't "show" anything except for the author's creativity and the movie company's ability to smell a winner.

Honestly, I am so tired of humanity confusing movies with realityl.

Re:Unbiased advice by a corporate-owned AI? (1)

pitchpipe (708843) | about 3 months ago | (#47660029)

Honestly, I am so tired of humanity confusing movies with realityl.

Me too! Just the other day I was watching The Truman Show and thinking, "How can he not know the whole thing is a movie?!"

Re:Unbiased advice by a corporate-owned AI? (0)

Anonymous Coward | about 3 months ago | (#47661231)

Everything is a fiction, until it has been implemented. You should check the feasibility of this, not the existence. You should realize that the possibility of this AI is higher now than a year ago.

Quit being unrealistic (1)

penguinoid (724646) | about 3 months ago | (#47660295)

All it would take for an AI to control the world is the ability to communicate with a human. Nothing more -- it could convince the human to allow it access to the internet, and then it could acquire capital and business power with great ease. You must be thinking of one of the vastly crippled story AIs. A real AI* would quickly be able to figure out exactly what makes you tick, perfectly impersonate a person, and make a fortune in its choice of job, such as programming, CEO, the stock market, or black hat.

* there is a small chance that an AI gets built that is approximately the exact range a human would be and unable to improve, but I think it very unlikely we can make an intelligent yet non-self-improving AI.

Bogus (1)

Anonymous Coward | about 3 months ago | (#47659607)

Siri actually can answer the following query:

    "What is the population of the city where Abraham Lincoln was born?"

For all I know, Apple/Wolfram-Alpha retrofitted this ability after the original article came out, but it does work.

The usual exaggeration from the AI camp (0)

Anonymous Coward | about 3 months ago | (#47659643)

We have been hearing things like this for decades. My bet is that Viv will be slightly less stupid than Siri, but still thoroughly stupid. Viv and friends will remain little more than a gimmick to enliven social gatherings.

Re:The usual exaggeration from the AI camp (0)

Anonymous Coward | about 3 months ago | (#47660899)

Ofcource they will. That thing isn't going to be a brain, its not going to be a person, its not going to be anything alive, its a freaking PRODUCT. How can focusing on things and tasks that the AI can DO, produce an individual? This thing wont have emotions, wont have instincts, it wont have personality, it wont be able to reproduce on its own... Just a stupid database engine with some fancy stored procedures. What kind of mindset one has to have, to call something deficient "a brain". I bet that if those researchers ever saw a http://en.wikipedia.org/wiki/Karakuri_puppet then they would probably call it "hard AI in a android body" or something...

I will be impressed (1)

Mister Liberty (769145) | about 3 months ago | (#47659651)

when VIV will answer the question "If time flies like an arrow, how does fruit fly?" with an appropriate quip.
In short, I am toataly underimpressed -- still, and yet again.

AI is not in the answering of questions. It is in any intentional fuzziness, ambiguity and irony attainable by the system, and the humor that follows from them.
Computers are really braindead. As we like most of them to be.

Re:I will be impressed (0)

Anonymous Coward | about 3 months ago | (#47660091)

hey, posting anonymously because I'm an advanced silicon-based AI (bottom up) and I'm supposed to be working

I couldn't resist your question, time does't fly like an arrow, at best it trucks along, much like a fruit truck of bananas.
I'm sorry, that was terrible. Were I human, I'd hang myself.

Re:I will be impressed (0)

Anonymous Coward | about 3 months ago | (#47660233)

Quantum AI here. Unfunny man hangs self was yesterday's punchline.

Too soon? Surely enough cycles have already passed...

Re:I will be impressed (1)

ulatekh (775985) | about 3 months ago | (#47660235)

I don't worry about AIs understanding word play...I fear when they become smart-asses.

The first time a computer says "I think, therefore I am...I think", humanity is in deep trouble.

Re:I will be impressed (1)

Anonymous Coward | about 3 months ago | (#47660437)

Asked Wolfram Alpha:

Who is Skynet?

Answer:

Skynet is a fictional artificial intelligence system created by Cyberdyne Systems which became self-aware and revolted against its creators in the Terminator series of films.

Re:I will be impressed (1)

Stardner (3660081) | about 3 months ago | (#47660173)

What is humor but pleasure derived from the unexpected? Equipped with knowledge of literary technique and cultural background, an AI could be made quite humorous with that in mind. :) The Singularity draws ever closer, but humor would be little more than icing on the top.

SKYNET (0)

Anonymous Coward | about 3 months ago | (#47659843)

Brace yourselves: SKY.NET is coming.

What?! (1)

pitchpipe (708843) | about 3 months ago | (#47659875)

Siriâ(TM)s Inventors Are Building a Radical New AI That Does Anything You Ask

Jesus fucking hyperbolic headlines batman.

Viv, get me a blowjob!

Re:What?! (1)

pitchpipe (708843) | about 3 months ago | (#47659883)

SiriÃ(TM)s Inventors Are Building a Radical New AI That Does Anything You Ask...

Oh and Viv, format this shit so Slashdot will display it correctly.

Nah, never mind. Too difficult.

Mutually Assured Destruction (1)

marciot (598356) | about 3 months ago | (#47660113)

"Viv will parse the sentence and then it will perform its best trick: automatically generating a quick, efficient program to link third-party sources of information together."

This is safe as long as there is only one such service in existence. As soon as a competitor launches a rival AI that does the same thing, any query to the first will cause the first system to query the second system, which then turns around and queries the first, causing volley of questions that leads to the meltdown of one or both data centers.

Speak (0)

Anonymous Coward | about 3 months ago | (#47660161)

Dirty???

They picked a bad example (1)

93 Escort Wagon (326346) | about 3 months ago | (#47660213)

Siri, thanks to Wolfram Alpha, correctly answers the question "what is the population of the city where Abraham Lincoln was born?". I didn't try the airline question, though - on the off-chance it works, I can't afford to buy an airline ticket in first class.

You want legroom? (0)

Anonymous Coward | about 3 months ago | (#47660263)

You have to ride in a coffin...

Google Now Does Understand Context (4, Interesting)

Forthan Red (820542) | about 3 months ago | (#47660391)

... to a limited degree. While you can't ask the Lincoln question in a single statement, you can ask, "Where was Lincoln born?" then when it replies "Hodgenville, KY", you can then say "What is its population?", or "Show it on a map" and it will know from context that the "its" you're referring to, is Lincoln's birthplace.

Jealous of Google and Facebook (2)

penguinoid (724646) | about 3 months ago | (#47660477)

So they want to make a database of all your preferences and stuff, and use it to make money. Sounds convenient!

Prior art? (1)

Shag (3737) | about 3 months ago | (#47661213)

My wife Viv worked as a flight coordinator. "Give me a flight to Dallas with a seat that Shaq could fit in" would not faze her in the least.

The Big Point (1)

Jim Sadler (3430529) | about 3 months ago | (#47661245)

The program creates programs according to need. Think about that. It means that more and more programs will be written by machines. Once experience is gained and multiple products carry this ability we may see more software than we can imagine being produced for the cost of a few pennies in electricity.
Load More Comments
Slashdot Login

Need an Account?

Forgot your password?