[WriterOfMinds]

IBM's Watson was much lauded when it performed at Jeopardy, but where is it today? https://www.nytimes.com/2021/07/16/technology/what-ever-happened-to-ibms-watson.html

TL;DR: They were hoping to use it for ambitious things like fighting cancer. Instead, it's become yet another customer service chatbot. But it does about as well at NLP as the more recent big names like Google.

[infurl]

It hasn't led to the revolution that IBM was originally predicting, but after a lot of expensive trial and error, Watson does seem to be finding success making incremental improvements to business practices. As they say towards the end of the article, Watson technology helps with the little things so humans can focus on the big things.

[HS]

I thought this section of the ''Comparison with human players'' on Watson's Wikipedia article seemed to present a clue.

Watson's basic working principle is to parse keywords in a clue while searching for related terms as responses. This gives Watson some advantages and disadvantages compared with human Jeopardy! players. Watson has deficiencies in understanding the contexts of the clues. As a result, human players usually generate responses faster than Watson, especially to short clues.

It may be difficult for a natural language processor to work with big concepts because the meanings of words relative to each other exist in a limited way. I think numbers have been so successful in modeling natural processes because that's a system with obvious internal relativity. The state of ''8'' relative to ''3'' is self evident. This can be mapped to any natural phenomenon exhibiting a similar difference relative to itself.

We may lack adequate contextual understanding with words alone because they don't intrinsically show how they relate to other words. A word is not designed to be a useful representation of reality, a symbol like a word is supposed to represent a useful representation of reality.

Maybe an NLP could be mapped to numerical representations of the natural phenomena which the language represents. Then words could be compared to each other based on the underlying math.

[frankinstien]

I remember those matches on Jeopardy and realized back then that it was using a simple yet powerful technique. As I watched the game on TV I got this epiphany: "Could Wikipedia be used to play the game by a machine?". So I googled using the clue words in the game and Wikipedia was the first link for each of those clues! Not only that but the first one or two sentences in Wikipedia provide a summary of the subject matter. So with a few tricks to use those sentences, Watson could easily win the game! I later found out that, indeed, Watson does use Wikipedia and its stored on its disk storage! So Waston didn't cheat by going online but used its own memory from its assimilation of knowledge where some of it is from Wikipedia. 

[MagnusWootton]

I wonder if there is a way to get it to work, (like be a working intelligent chat bot) but its just swept under the rug with all the rest of the possible Ai.!!

[DaltonG]

IBMs a strange company. What happened to Watson was the same thing that happened to OS/2 and Lotus Smart Suite. I remember a quote that IBM had spent a billion dollars developing OS/2 then left it by the side of the road because they didn't know how to market software in the PC market. I suspect they are still alive because of government contracts for their research and development divisions. Only someone like the NSA could afford to buy something like Watson and cover the cost of running it.

[MagnusWootton]

May be corporate lies.

[Dat D]

IBMs a strange company. What happened to Watson was the same thing that happened to OS/2 and Lotus Smart Suite...

Seems like IBM is doing more researches than business. IBM should do a name change to IRM, International Business Machine to International Research Machine