Off topic: Computer can now debate विषय पोस्ट करनेवाला व्यक्ति: Phil Hand
|
|---|
Phil Hand 
चीन
Local time: 16:27
चीनी से अंग्रेजी
This is pretty impressive. Start watching at 45.25.
https://www.youtube.com/watch?v=6fJOtAzICzw
Watson, the computer which won on an American quiz show, can now be asked a natural language question, go and mine Wikipedia for cogent responses to the question, and give the responses in ordered, fairly correct language. That's a massive advance.
| | | |
I definitely need one to write my essays at university! Thanks for sharing!
Completely off-topic: Looking back at my first experiments with artificial intelligence programming languages, I daresay that artificial intelligence has never really happened. Computers are not really capable of "thinking". Instead, they take huge amounts of textual data and make sense of them, which is of course all good and interesting, but not really artificial intelligence, or is it?
[Edited at 2... See more I definitely need one to write my essays at university! Thanks for sharing!
Completely off-topic: Looking back at my first experiments with artificial intelligence programming languages, I daresay that artificial intelligence has never really happened. Computers are not really capable of "thinking". Instead, they take huge amounts of textual data and make sense of them, which is of course all good and interesting, but not really artificial intelligence, or is it?
[Edited at 2014-05-15 05:50 GMT] ▲ Collapse
| | | | | Don't we do the same? | May 15, 2014 |
Doesn't the human brain do the same thing? We take the sum of our experiences and our knowledge and make sense of that.
Although I do believe that computers still have to develop more "creative" ways to make sense of the data. I guess our pattern recognition skills might span a larger number of very different fields and we often combine data from very different areas to make sense of informations.
| | | | Phil Hand
चीन
Local time: 16:27
चीनी से अंग्रेजी
विषय आरंभकर्ता | What understanding is | May 15, 2014 |
Tomás Cano Binder, CT wrote:
I daresay that artificial intelligence has never really happened. Computers are not really capable of "thinking". Instead, they take huge amounts of textual data and make sense of them, which is of course all good and interesting, but not really artificial intelligence, or is it?
I was looking at this and wondering whether it counts as intelligence. I think the key thing is whether it can be applied recursively. In the little demonstration in the video, the computer looks at existing evidence (Wikipedia pages) and processes them. You're right that this probably doesn't count as AI. What would count as AI is if it can take its own conclusions and do something useful with them, analyze its own output and draw conclusions. If you have a computer doing that, analyzing its own output, then I think that would count as "understanding" and "thinking." And it seems like that could be quite close now.
Coincidentally, I was reading up on Google's driverless car. It's moved off relatively empty highways and into small towns. It can now navigate through a town with reasonable success. That's pretty intelligent.
I'm quite excited by all this! I feel like there could be quite a big breakthrough very soon.
| | |
|
|
|
| Intelligence is hard to define | May 15, 2014 |
One thing is quite sure in my opinion: a person does not need to process all the information again in order to identify and make sense of new information or a new situation, the same way a driver does not need to process all signs of the road to be a good driver.
Experiences and individual pieces of knowledge we acquire in life get condensed along time in some kind of crystalline or sticky conceit that makes up our mindset. Furthermore, we can instantly decide what is generally inte... See more One thing is quite sure in my opinion: a person does not need to process all the information again in order to identify and make sense of new information or a new situation, the same way a driver does not need to process all signs of the road to be a good driver.
Experiences and individual pieces of knowledge we acquire in life get condensed along time in some kind of crystalline or sticky conceit that makes up our mindset. Furthermore, we can instantly decide what is generally interesting or relevant and what is not and are quite able to identify the subtlest deviation from a pattern or a norm. All this helps us quickly filter out useless information without a need to consider its proportional weight in a corpus of information.
To me, intelligence is this whole process of developing competences which, from a certain point in time, become fully automatic and do not require any extra processing. In fact, we forget how we acquired each individual competence: we can do complex things like reading, speaking, writing, making calculations or rebuilding a complex machine with blinded eyes if we have done it a sufficient number of times.
I keep thinking that computer science has failed to create intelligence. Instead, it creates the kind of obnoxious know-it-all nerdy types you would hardly invite to stay at your home for more than three hours. It would be hard to live with someone who cannot switch him/her/itself off and relax. ▲ Collapse
| | | | Neil Coffey
यूनाइटेड किंगडम
Local time: 09:27
फ्रांसीसी से अंग्रेजी
+ ...
| More of a sophisticated information retrieval process | May 15, 2014 |
Phil Hand wrote:
Watson, the computer which won on an American quiz show, can now be asked a natural language question, go and mine Wikipedia for cogent responses to the question, and give the responses in ordered, fairly correct language. That's a massive advance.
Mmmm isn't what it's doing quite a bit more constrained than that, though? Isn't it doing something more like this:
- recognise the concepts/semantic relationships between concepts in the input sentence
- from a database of text, recognise text that also includes those concepts/relationships between them and pull out target sentences/statements
- of those target sentences/statements, detect whether they appear to represent a "PRO" or "CON" wrt to the original set of concepts/relationships
None of these steps are trivial, but they do build on a body of incremental research that's been going on for the past few decades. The difference is that Watson is a machine with the capacity to apply these techniques to the entire Wikipedia database in a few seconds, but I don't think it represents a giant leap in NLP techniques as such.
I think I would still classify this as more of a "sophisticated database retrieval" operation than an act of "reasoning" or "debate" on the part of the computer.
[Edited at 2014-05-15 19:16 GMT]
| | | | Phil Hand
चीन
Local time: 16:27
चीनी से अंग्रेजी
विषय आरंभकर्ता | Caught out by headlinese | May 16, 2014 |
Neil Coffey wrote:
...build on a body of incremental research that's been going on for the past few decades. The difference is that Watson is a machine with the capacity to apply these techniques to the entire Wikipedia database in a few seconds, but I don't think it represents a giant leap in NLP techniques as such.
I think I would still classify this as more of a "sophisticated database retrieval" operation than an act of "reasoning" or "debate" on the part of the computer.
Yes, the word debate is clearly hyperbole - I transferred the headline from wherever I saw the link first. Like I suggested above, I think that when a computer is able to apply its analysis iteratively to its own output then we will have to concede that a form of "thinking" is going on.
However, I do think that the ability to process massive amounts of data does indeed represent a giant leap forward. In fact, it's *the* giant leap forward. Computers are already reasonably smart. The reason they can't do language is because they don't have the linguistic input. They've never been able to read (process) lots of text before. Everything had to be parsed by hand and presented to them. It is precisely the ability to interact with, parse and analyse text independently which will allow them to assimilate massive databases, and finally learn language. (By comparison, almost all of a child's linguistic input and output is paired with other cognitive data, allowing her to parse and process it.)
| | | | To report site rules violations or get help, contact a site moderator: You can also contact site staff by submitting a support request » Computer can now debate | Trados Business Manager Lite | Create customer quotes and invoices from within Trados Studio
Trados Business Manager Lite helps to simplify and speed up some of the daily tasks, such as invoicing and reporting, associated with running your freelance translation business.
More info » |
| | TM-Town | Manage your TMs and Terms ... and boost your translation business
Are you ready for something fresh in the industry? TM-Town is a unique new site for you -- the freelance translator -- to store, manage and share translation memories (TMs) and glossaries...and potentially meet new clients on the basis of your prior work.
More info » |
|
| | | | X Sign in to your ProZ.com account... | | | | | |
Login to reply/comment 
