Watson

[AndrewB]

[Aliotroph?]

Impressive. Interesting how it had extremely low confidence about the correct answer for the art period question, as if it picked the wrong subject for the question.

[Technician]

Just try and get an accurate answer out of it in a few decades.

[Aliotroph?]

The vibe I got from Watson was a lot like this:

[lupinx-Kassman]

Cool. Probably the only time I would want to watch jeopardy is for something like this.

[Scuba Steve]

[Aliotroph?]

It doesn't seem like Watson's answers come any slower when it has to guess, but most of the questions I saw the other guys get were when Watson had low confidence. Maybe you really need tenth-second reaction times.

[E_O]

I, for one, welcome our new computer overlords.

[Technician]

I can't wait until they give him a CGI face.

[gggmork]

I saw the nova program on this earlier. It doesn't use the internet but it has stuff like a static version of wikipedia and imdb etc in its memory. It seems to brute force through a massive memory with the help of machine learning. I wonder if a next step could be to make it have goals and ask questions itself (kinda ignoring that jeopardy makes the 'answer' a 'question') as well as answer them. Human goals like "mate" are pretty much programmed in. I think for any sort of technological singularity to occur there probably needs to be physical embodiment. A human can physically alter/interact with the world with their hands etc, and we can observationally test facts like "does water fall out of a cup" or something. A bodyless computer couldn't perform such experiments so would seemingly have to accept facts from others on faith. I'm not even sure if 5+7=12 is an fact/axiom 'known' by a computer. Its sort of just the behavior of digits that carry over rather than the computer knowing it as a fact, and we told them to carry over that way. I could be wrong, but humans don't have 100% known facts either.
If mere bits somehow became extremely intelligent by learning natural language and reading the non-bs portions of the internet, possibly extrapolating its own unique educated guesses and ideas/ questions/answers beyond that, it might know a method to embody itself. Maybe nothing will be much of a threat until it starts replicating like real life. Or I guess it could manipulate the bodies of humans somehow temporarily at least.

[Csonicgo]

You can tell Ken and Brad had a blast with WATSON.

[Justince]

r_init: getpostnumberforname: post not found!

Edited July 28, 2017 by Justince

[bytor]

Fun to watch but I found it creepy to see Jeopardy contestants obsoleted by a computer. Seen too many 'evil computer takes over' movies, I s'pose. Watson looked neutral, unobtrusive, almost cute. Imagine if they stood something there that looked like a Terminator.

"Logan 5, did you find sanctuary?"

[Quasar]

Some of the criticisms of Watson are very interesting because they show how much some people really don't "get" it.

For example, my favorite was this one:
"It only knows things it was told to know by humans."
- As do most humans. You spend the first 18 years of your life getting programmed in an extremely inefficient manner with all manners of facts. You don't start from scratch with naught but the scientific method and rediscover everything on your own.

Watson doesn't "think," and all the engineers on it will readily tell you that. But what it is doing is a LOT closer to thought than what any other machine has accomplished before. It is using parallelism (much like the human brain) to run out multiple algorithms at once, following dynamically created and weighted connections between facts (somewhat similar to neural pathways and how they encode information).

Language is undeniably core to human intelligence, and Watson's approach to understanding it, while far from perfect, is an evolutionary leap forward. When you watched the show, tell me you couldn't help but think it was "smart." It's engrained into our own perception that things which can understand language are intelligent.

The appearance of intelligence in machines will increase. At first it is only appearance. But then the actual intelligence of machines will appear as well, and by that time there will be no obvious way to tell any more. The mystery of intelligence and what "causes" it will remain eternal, more than likely, because we won't ever know exactly what "one development" in a complex system finally leads to it. It may prove to be something over which we truly will have *no* control.

[Maes]

Heh this creepily reminds me of Asimov's first story in the "I, Robot" novel. Of course, keeping with those 50s retro-futurism's flawed predictions, of which not even Asimov was entirely immune to (quite the opposite), household and general purpose humanoid robots were already available, but they lacked the power of speech.

In that first story, a little girl who lost her robotic companion saw one of the first "talking robots" in a tech exhibition, who was little more than a large cabinet on wheels, and could only answer very simple questions to entertain guests (and a human operator evaluated each question and judjed whether it was suitable for the robot to digest). When left alone with the little girl unsupervised, the questions she asked sent it in a state of tilt, especially when she told it that she was looking for "My friend Robbie, a robot just like you, Mr. Robot". The concept of "just like itself" was overwhelming for the primitive electromechanical-pozitronic "brain" ;-)

[Csonicgo]

Quasar said:

It may prove to be something over which we truly will have *no* control.

"Jan, I want to play WoW."
"MARK CAN I TELL YOU MY DREAM FIRST??!"
"Jan, run the program, please... stop crying or I pull the plug"

[gggmork]

" When you watched the show, tell me you couldn't help but think it was "smart." "

I'm pretty sure all jeopardy replies are a single noun (fused into the form of a question). It might be interesting to take all the nouns out of each jeopardy "answer" (sigh, I mean the stuff that alex trebec says which is basically the "question"), put them in google and see if the correct noun "question/answer" is easy to find in the first result.
It might be a whole new can of worms to think up complete sentence answers like "why/ how does __"... "because... uh...". Then again, if the goal is to surpass human intelligence, there's not much of a hurdle. Just make its "brain" 99% occupied with sex and religion. "how are rainbows formed". "xenu made them, will you fuck me alex?"
Ha ha, I love sean connery in (saturday night live?)'s parody. I really like that shirt alex. No that's a really great shirt. I just have one question. Do they make it for men? hua hua hua hua

[magicsofa]

gggmork said:

I really like that shirt alex. No that's a really great shirt. I just have one question. Do they make it for men? hua hua hua hua

Yer sittin on a gold mine Trebeck!

[Aliotroph?]

Maes said:

Heh this creepily reminds me of Asimov's first story in the "I, Robot" novel. Of course, keeping with those 50s retro-futurism's flawed predictions, of which not even Asimov was entirely immune to (quite the opposite), household and general purpose humanoid robots were already available, but they lacked the power of speech

That always made me wonder if Robbie would have died from the same idea. He did go to length to imply producing speech, rather than just understanding it, took brain power of such magnitude they could barely do it, and yet nobody cared what was asked of the household robots.

[Maes]

Well, the lack of speech sounded more like an artificious impediment introduced by Asimov as a plot element, or simply a lack of foresight by Asimov. He considered a humanoid, fully-autonomous, self-contained, self-powered, dexterous and intelligent robot (enough to to interact with human beings and be trusted to be a companion to a little girl) to be feasible, yet he described speech production as cumbersome and awkward.

In reality technology went the other way around: computerized speech synthesis was possible not much after the novel was written, while we're still stumbling with creating a mass-producible robot that could do a fraction of what Robbie -an 'early' model by Asimov's standads- was capable of. Think flying cars and household nuclear reactors, for instance. In any case, Robbie was described as being fully able to read and write, so the idea could have been just as easily expressed to him.

OTOH Asimov had written another independent robot novel, Cal, where a famous writer had got a household robot servant which was able to speak but -initially- was seriously lacking in his ability to express written thoughs -only after a hardware upgrade he was able to write fluently. As much as I like Asimov as a sci-fi writer, he can sometimes be very inconsistent in his technology's timeline.

In the same I, Robot novel, at Chapter Three (the one with the infamous QT robot) there is indeed a robot who becomes not only self-aware, but distrustful of humans (at first), and then condescending in their regards due to his perceived superiority (the passage where he compares himself to humans and proved his superiority rationally at every point he raises, is distubingly convincing). Even though the robot,in its logical zeal, had made one logical fallacy -that no being could create another being superior to itself, something easily disproven. Still, it's Asimov.

[Jimmy]

I'm just popping into this thread to say that Watson is not a sign of the coming robot apocalypse, nor will it be the instigator of a violent and bloody war between man and machine. It's a brilliant piece of computer engineering, and is going to pave the way for all manner of subsequent technological advancements which will ultimately benefit humankind in a multitude of ways, and this is an opinion I hold very strongly.

Also, I'm 100% in favor of assigning the the head of Norman Lovett to be the avatar of the next interactive computer system to be developed.

That is all. :)

[Chopkinsca]

Jimmy91 said:

I'm just popping into this thread to say that Watson is not a sign of the coming robot apocalypse, nor will it be the instigator of a violent and bloody war between man and machine.

Aww, way to get my hopes up.

[Quast]

Quasar said:

Some of the criticisms of Watson are very interesting because they show how much some people really don't "get" it.

For example, my favorite was this one:
"It only knows things it was told to know by humans."
- As do most humans. You spend the first 18 years of your life getting programmed in an extremely inefficient manner with all manners of facts. You don't start from scratch with naught but the scientific method and rediscover everything on your own.

Watson doesn't "think," and all the engineers on it will readily tell you that.

Computers have no ego and super ego. When an ai does something remarkable it was not specifically programmed to do (aside from crashing) then wake me up. Computers cannot just decide to do shit on their volition. Human beings on the other hand, can express their own will without being taught 'facts'.

[AndrewB]

Quast said:

Computers cannot just decide to do shit on their volition. Human beings on the other hand, can express their own will without being taught 'facts'.

Bullcrap. Computers can be programmed to be essentially the same as humans in every significant way. That includes the ability to do silly and non-preprogrammed stuff on a whim just to reaffirm one's own sense of free will. With the success of Watson, they've succeeded in emulating one component of the human brain for one specific task. All they have to do is emulate about 20 or 30 more components, program them toward specific goals (our version of food and sex) and you've pretty much got a sentient/self-aware being.

[DuckReconMajor]

Quast said:

Human beings on the other hand, can express their own will without being taught 'facts'.

What can you do that you weren't taught? Oh, sorry a computer can't scratch its dick yet.