To make things interesting, you need a sizeable community with stated goals and ambitions. You cannot just take any niche topic and call it a failure. A failure compared to what?
I predicted the demise of the Semantic Web. It is still going strong under different names. It is all going to eventually collapse the same way expert systems failed.
But failures in research are also about not investigating some problems, or falsely considering them as being solved. Database design comes to mind: it is considered a textbook topic… but it is far from being solved! In fact, the information provided by textbooks is not even valid! The truth is that textbooks don’t tell you how to design databases in the real world.
That is, I believe, a great sin. We often insist that some problems are solved when what we advocate fails. Again, we refuse to see failures.
What happened when expert systems failed? Did all these researchers renounced their way? Some have, but a lot continued to do the same research under different guise.
Tim Berners-Lee teamed up with James Hendler to propose the Semantic Web. Hendler is a long-time classical AI guy who wrote a textbook on expert systems. You could reasonably view the Semantic Web as a distributed expert system where the expertise is encoded as RDF.
They all share a common vision of what intelligence is… this vision goes back to classical AI. To these people, intelligence is a collection of facts together with a “reasoning engine”.
This vision has failed, repeatedly, and it keeps on failing… but its proponents have no trouble finding new disciples with each generation because their vision is compelling. It is still wrong though. It does not lead to intelligence.
The VLDB statistics are even more striking when you note that the number of VLDB papers per year has roughly doubled in that time frame (from 100 to 200). So as a percentage of accepted VLDB papers, the numbers look more like 2003: 25%, 2008: 10%, 2012: 1.5%.
I have a huge grudge against XML. People are so obsessed by it, so they don’t want to tolerate a couple of binary formats here and there. At the same time, from the performance perspective, XML sucks big time.
Dror Hararisays:
Some would argue that we sort of having that mega expert system that transformed the world. Some call it “Dr. Google” and ask it all sort of questions in their native language. While the answer they get are most indirect, answer they are still.
You can find more ‘expert systems’ like that (e.g Wolfram Alpha) and while purist would object saying those are not expert system, I certainly think they are and am sure that if had we described what they do to someone back then in the early 80, they would have also counted them as expert system.
/d
Bobsays:
It was George Santayana who said it, not Edmund Burke.
I actually would have liked XML databases to succeed. It would have given database researchers decades of research problems.
I don’t understand why the research aspect of it collapsed so fast. It is somewhat puzzling. Researchers are typically more stubborn.
alansays:
okay so xml is not great as a database per se compared to mysql etc but many bodies and industry are using xml to store and carry data with standards such as HL7 and the finance standards for credit card data etc
http://www.writersofthefuture.com/time-capsule-predictions
What do you see as tomorrow’s failure from today? I mean it is of course hard to predict but what’s your hunch?
Machine translation of Russian to English.
“the spirit is willing but the flesh is weak.”
==
“the vodka is good but the meat is rotten.”
………….
Long before my time, but this article reminds of the “AI Winter”. http://en.wikipedia.org/wiki/AI_winter
Nice article, Daniel.
Would you mind commenting on the connection you refer to between expert systems and the semantic web?
Thanks!
@s
To make things interesting, you need a sizeable community with stated goals and ambitions. You cannot just take any niche topic and call it a failure. A failure compared to what?
I predicted the demise of the Semantic Web. It is still going strong under different names. It is all going to eventually collapse the same way expert systems failed.
But failures in research are also about not investigating some problems, or falsely considering them as being solved. Database design comes to mind: it is considered a textbook topic… but it is far from being solved! In fact, the information provided by textbooks is not even valid! The truth is that textbooks don’t tell you how to design databases in the real world.
That is, I believe, a great sin. We often insist that some problems are solved when what we advocate fails. Again, we refuse to see failures.
@Robert Primmer
What happened when expert systems failed? Did all these researchers renounced their way? Some have, but a lot continued to do the same research under different guise.
Tim Berners-Lee teamed up with James Hendler to propose the Semantic Web. Hendler is a long-time classical AI guy who wrote a textbook on expert systems. You could reasonably view the Semantic Web as a distributed expert system where the expertise is encoded as RDF.
They all share a common vision of what intelligence is… this vision goes back to classical AI. To these people, intelligence is a collection of facts together with a “reasoning engine”.
This vision has failed, repeatedly, and it keeps on failing… but its proponents have no trouble finding new disciples with each generation because their vision is compelling. It is still wrong though. It does not lead to intelligence.
It all makes me a bit sad.
The VLDB statistics are even more striking when you note that the number of VLDB papers per year has roughly doubled in that time frame (from 100 to 200). So as a percentage of accepted VLDB papers, the numbers look more like 2003: 25%, 2008: 10%, 2012: 1.5%.
I have a huge grudge against XML. People are so obsessed by it, so they don’t want to tolerate a couple of binary formats here and there. At the same time, from the performance perspective, XML sucks big time.
Some would argue that we sort of having that mega expert system that transformed the world. Some call it “Dr. Google” and ask it all sort of questions in their native language. While the answer they get are most indirect, answer they are still.
You can find more ‘expert systems’ like that (e.g Wolfram Alpha) and while purist would object saying those are not expert system, I certainly think they are and am sure that if had we described what they do to someone back then in the early 80, they would have also counted them as expert system.
/d
It was George Santayana who said it, not Edmund Burke.
@Harari
I am sure that some would call Google an expert system. The only problem is that it works nothing like an expert system.
In fact, if you went back in 1980 and told people who Google works, they would not believe you that such a thing is possible.
Heck. They would not believe that Wikipedia is possible.
@Itman
I actually would have liked XML databases to succeed. It would have given database researchers decades of research problems.
I don’t understand why the research aspect of it collapsed so fast. It is somewhat puzzling. Researchers are typically more stubborn.
okay so xml is not great as a database per se compared to mysql etc but many bodies and industry are using xml to store and carry data with standards such as HL7 and the finance standards for credit card data etc
@alan
Yes, that is why I wrote:
Storing data in XML for long-term interoperability is an acceptable use of XML.