Roger Schank - about IBM Watson

Original author: Roger Schank
  • Transfer
American theorist of artificial intelligence, one of the founders of cognitive psychology and entrepreneur Roger Schank published in his blog a personal opinion about the Watson system. Here is the translation of this entry.

image

IBM doesn’t do cognitive computing, no matter how many times they repeat these words.

Yesterday I corresponded with an old friend, and he reminded me of our conversation, which took place almost 50 years ago. I tried to explain to him what I do for a living, and he tried to understand why making computers to understand something was more difficult than analyzing keywords. I explained the concepts of sentences, and the fact that sentences consist of words, but people actually do not use words when they think - except when they need to get to the underlying ideas, and this is where computers have difficulty.

After 50 years, keywords are still in the first place among people trying to get computers to work with the language. However, this time the key word with which people deceive the public is AI. People declare that the AI ​​exists, the AI ​​thinks, and that we should all be afraid of him, or rejoice in his arrival - I forgot what it was.

At some point, we have achieved some success in the understanding of natural language with computers, but in 1984 the “winter of AI” began. It started because of too many promises about what the AI ​​can do and what it really could not do. (Then everyone advertised expert systems - and where are they now?) Funding stopped, and with it the real work on the processing of natural language stopped.

But people are still running around with keywords, since Google and the like use them for "search." With the search, everything is fine, as long as we count the words - these are exactly the kind of tasks that are involved in data analysis and machine learning. Of course, by counting words, correlations can be made, and users can find out which words are often associated with which ones, and use this information. But users had to adapt to Google, and not vice versa. We know what can be entered in the search bar, and what is not, and we perform such a search, with which Google will most likely be able to help us. We know that we are looking for texts, not answers, and we are not going to start a conversation with some entity that knows what we want to talk about. People learn through dialogue, but dialogue with Google will not work. I can pretend to be talking to Siri,

But I'm not worried about Google. It satisfies our needs quite well. I’m worried about the exaggerated claims that IBM makes about its Watson program. Recently, they had an ad with Bob Dylan, which would be very funny if she hadn't pissed me off so much.



Frankly, Watson is a hoax. I do not claim that the program can not process words, and it can even be useful to someone. But its advertising is deceiving you.

Here is an excerpt from Ad Week:
The computer boasts that it can read 800 million pages per second, identifying key topics in Dylan’s works, such as “time is running out” and “love is dying out.”

Ann Rubin, IBM Branding Vice President, told Adweek that advertising was needed to help people understand the new world of cognitive computing.

“We focus on advertising, but it’s much more than just an advertising campaign,” says Rubin. - This is IBM's point of view, and it is permeated with all our marketing, our internal corporate communication, the way we work with suppliers and employees. It permeates everything we do. ”

IBM says the new series of advertisements should convey to a wider audience - companies, decision makers, software developers - an understanding of how Watson works. Unlike computers that are programmed in the usual way, cognitive systems such as Watson understand, reason, and learn. The company says that industries like banking, insurance, healthcare, and retail can benefit from the system.

Rubin said that Watson can think better than the human brain, in areas where finding answers and connections is difficult because of the abundance of data. “You can think of what to do with cancer, risks, doubts, competitors, if you accept the idea of ​​cognitive computation,” she says.

Seriously? I'm a 60's kid, and I remember Dylan's songs pretty well. Ask any of this era about Bob Dylan, and no one will tell you that the main theme of his songs is “love passes.” He was a protest singer and sang about the vicissitudes of life. He participated in the anti-war movement. Love passes? This gave a stupid computer that can count words. How could Watson know that many of Dylan’s songs were part of the antiwar movement? Has he often repeated the word "anti-war"? Yes, this word may not have sounded at all in his songs.

From the rock site :
In Bob Dylan's No. 1 protest song, “The Times They Are a-Changin,” Dylan got all in and crossed the popular protest movement in the 1960s with a civil rights movement.
Short verses superimposed on each other and made a powerful impression, and words like:

There's a battle outside and IT is ragin '
It'll soon's-shake your windows and rattle your walls
For the Times for They are a-a changin'

There goes the battle, and she erupts
Soon it will shake your windows and stagger your walls
As times change

, Dylan’s cult claims have outlived their time.

But he does not sing about Vietnam or civil rights. Therefore, Watson would not have known that there was some connection with these concepts. It is possible to speak on some topic, using words that do not speak about it directly. Fundamental knowledge means a lot. A few years ago, I asked a 20-year-old man about Bob Dylan, and it turned out that he had never heard of him. He didn't know much about the 60s. Little does Watson know. Words cannot be understood without understanding the context.

Suppose I tell you that my friend is buying too many sleeping pills, and I'm worried about him. Will Watson say I'm hinting at suicide? Will he quickly run to a friend and talk to him about his problems? Of course not. People understand the context because they know how the world works and what problems people have. They do not count words.

And here's another from another site:
Calling Bob Dylan the father of folk music will probably be an exaggeration. But one can definitely say that this singer is one of the eminent authors of anti-war and protest songs of the twentieth century, so he is completely worthy of creating the rating of the 10 best protest songs of Bob Dylan. The singer changed his repertoire from defiling the establishment to country music, pop music and folk music, and remains productive in the field of protests against those in power.

It was written by a man. How did I find out about this? Because Watson cannot draw conclusions by counting words on 800 million pages of text.

What upsets me most is not Watson, but IBM statements. I quote:
Unlike computers that are programmed in the usual way, cognitive systems such as Watson understand, reason, and learn.

Ann Rubin, IBM Branding Vice President, told Adweek that advertising was needed to help people understand the new world of cognitive computing.

I wrote the book " Cognitive Computer " in 1984.

I opened Cognitive Systems in 1981. What I was talking about at that time in IBM was obviously not read (although they clearly liked the words I used). Watson does not reason. You can argue only if you have goals, plans, ways to achieve them, an understanding of what others can believe in, and knowledge of past experience, on the basis of which conclusions can be drawn. Also not interfere point of view. What does Watson think, say, about terrorism?

Stupid question? Really thinking beings have their own views on terrorism. The dogs do not have it, but Watson does not reach the dog (for example, the dog even knows how to get my attention).

In the 80s, I came up with such a thing as " precedent-based reasoning": it should have allowed computers to compare new situations with old ones and based on this change what the computer considered to be the result. We managed to build some useful systems. And we learned a lot about how a person learns. I thought that we created computers that would soon surpass people in thinking or become rational? No, of course. I thought that we started creating computers that could be more useful to people.

It would be nice if IBM hesitated the hype and told people what's the real way Watson, and would cease to drivel about love and leaving cancer deliberation. The IBM deceives us, and it needs to stop.

AI Winter is coming.

Also popular now: