Share and discuss this blog

Monday, November 30, 2015

key word analysis is not AI; one more time to get the point across to people building "conscious" computers

I am getting tired of talking about this, but there was yet another piece of stupidity published the other day.

As advancements in technology continue at an ever-increasing pace, will there ever come a day when we’ll be able to use science to cheat death? Australian startup company Humai seems to think so; it claims to be working on a way to transfer a person’s consciousness into an artificial body after they’ve died.
“We want to bring you back to life after you die,” says Humai CEO Josh Bocanegra on the company’s website. “We’re using artificial intelligence and nanotechnology to store data of conversational styles, behavioral patterns, thought processes and information about how your body functions from the inside-out. This data will be coded into multiple sensor technologies, which will be built into an artificial body with the brain of a deceased human. Using cloning technology, we will restore the brain as it matures.”

Really? OK, I am not even going to comment on this nonsense. This column is about key words.  I have had enough with claims about AI based on key word analysis, so I thought I would explain it once again, in a way that anyone outside of AI could understand.

Consider this: What does the proverb : a pig with two masters will soon starve mean?” While you are pondering that, I will mention two more proverbs to think about:

A stitch in time saves nine

You can lead a horse to water but you can’t make him drink

Understanding how we understand these proverbs will make clear why key word analysis isn’t going to lead to robot consciousness or discoveries in cancer or new Bob Dylan songs any time soon.

I have learned, (because I ask people about these in job interviews actually), that many adults have no idea what these proverbs mean and can’t explain them at all. One reason is that they may never have heard them before, but that is the key word analysis answer. “I never heard it, so I can’t look it up and say what I found.”  

In actuality, anyone who thinks hard can figure out what these mean. No computer can do that. But, remember that I am an AI person. I would like computers to be able to do this too, so I have thought quite a bit about it. Let me make clear what a person has to do in order to decipher the meaning of these proverbs. As I do this, think about how hard this would be for a computer to do.

Let’s start with the pig. English language proverbs are quite often said in farming metaphors (sailing is big too.) The first question is: why would a pig with two masters soon starve? It is a good question. Suppose it were a question on Jeopardy. Watson would lose. A smart person would win. Why? Because people who think don’t match key words. (They don’t ask themselves: where can I find a text where pig and starve are on the same page or how often are these words correlated?) What they do ask themselves is how having two masters would affect the pig. They also ask themselves other things, because sentences like this occur in actual contexts usually: 

Why is this guy talking to me about pigs? We weren’t discussing pigs.

(What is the guy who said this trying to say? We were taking about my life situation and now he is talking pigs. He must be making an analogy.)

Why would the pig starve? Well, who feeds the pig normally? Aha. Either of the masters might feed the pig. Well, what if each one thought the other was doing the feeding? Now, I get it. He is not talking about pigs at all. I have two bosses. He was telling me that neither may think they need to look out for me.

This is not rocket science. It is in fact, everyday human thinking. But such thinking is way out of bounds for what AI can do today. Tomorrow maybe. But that tomorrow would require that the computer would be able to have a conversation where one person’s goals were being discussed, where another person was giving advice that the other might follow, and where that the advice was being said metaphorically using a well known proverb.   This is what thinking looks like. It is not what key word analysis looks like.

To understand the advice given one must have a goal and ask oneself questions about how what was said relates to that goal and then figure out the answer. Could computers do that? I hope some day they can. Waston? Not so much. How about the above mentioned conscious robot company? Give me a break. We barely know what consciousness is, although I am pretty sure it has something to do with the stuff I just put in italics above.

I will let my readers figure out the stitch in time proverb themselves. Also I challenge the brilliant “AI” people at IBM give Watson a shot at it. Please let me know how it did. 

Let’s move on to the drinking horse. Why can’t you make him drink? Isn’t he thirsty? But, of course this proverb isn’t about horses. It is about education typically. It means that you can teach people but they don’t necessarily learn. Let your key word analyzer figure that out. How do I know that this is what that proverb is about? Because life is full of situations in which we try to help somebody and they refuse the help that is being offered. They don't agree, or they don’t care (or they aren’t thirsty.) You need to figure this out if have never heard the proverb before. So unless our key word analyzer has a key proverb analyzer too, the key word analyzer would be baffled by this. And if we did list the underlying meaning of every proverb in the English language, the program still wouldn't understand it, because the proverb is about goals and plans and decisions we make, and about how to learn to think differently. This is exactly what we are not yet able to do in AI, much as I would like for us to be able to do that. The AI winter that started in 1984 killed all the work on that kind of AI. That is the consequence of making ridiculous claims about what AI can do.

I will end on a joke I like: You can lead a horse to water but a pencil must be led. Watson: why is that funny? Let me know when Watson or our conscious computer has figured out the answer to that.

Thursday, November 19, 2015

The fraudulent claims made by IBM about Watson and AI. They are not doing "cognitive computing" no matter how many time they say they are.

I was chatting with an old friend yesterday and he reminded me of a conversation we had nearly 50 years ago. I tried to explain to him what I did for living and he was trying to understand why getting computers to understand was more complicated than key word analysis. I explained about concepts underlying sentences and explained that sentences used words but that people really didn’t use words in their minds except to get to the underlying ideas and that computers were having a hard time with that.

Fifty years later, key words are still dominating the thoughts of people who try to get computers to deal with language. But, this time, the key word people have deceived the general public by making claims that this  is thinking, that AI is here, and that, by the way we should be very afraid, or very excited, I forget which.

We were making some good progress on getting computers to understand language but, in 1984, AI winter started. AI winter was a result of too many promises about things AI could do that it really could not do. (This was about promoting expert systems. Where are they now?). Funding dried up and real work on natural language processing died too.

But still people promote key word because Google and others use it to do “search”. Search is all well and good when we are counting words, which is what data analytics and machine learning are really all about. Of course, once you count words you can do all kinds of correlations and users can learn about what words often connect to each other and make use of that information. But, users have learned to accommodate to Google not the other way around. We know what kinds of things we can type into Google and what we can’t and we keep our searches to things that Google is likely to help with. We know we are looking for texts and not answers to start a conversation with an entity that knows what we really need to talk about. People learn from conversation and Google can’t have one. It can pretend to have one using Siri but really those conversations tend to get tiresome when you are past asking about where to eat.

But, I am not worried about Google. It works well enough for our needs.

What I am concerned about are the exaggerated claims being made by IBM about their Watson program. Recently they ran an ad featuring Bob Dylan which made laugh, or would have, if had made not me so angry. I will say it clearly: Watson is a fraud. I am not saying that it can’t crunch words, and there may well be value in that to some people. But the ads are fraudulent.

Here is something from Ad Week:

The computer brags it can read 800 million pages per second, identifying key themes in Dylan's work, like "time passes" and "love fades.”

Ann Rubin, IBM's vp of branded content and global creative, told Adweek that the commercials were needed to help people understand the new world of cognitive computing.

"We're focusing on the advertising here, but this is really more than an advertising campaign," Rubin said. "It's a point of view that IBM has, and it's going across all of our marketing, our internal communications, how we engage sellers and our employees. It's really across everything that we do.”

IBM says the latest series is meant to help a broader audience—companies, decision makers and software developers—better understand how Watson works. Unlike traditionally programmed computers, cognitive systems such as Watson understand, reason and learn. The company says industries such as banking, insurance, healthcare and retail can all benefit.

Rubin said Watson's abilities "outthink" human brains in areas where finding insights and connections can be difficult due to the abundance of data.

"You can outthink cancer, outthink risk, outthink doubt, outthink competitors if you embrace this idea of cognitive computing," she said.

Really? I am a child of the 60s’ and I remember Dylan’s songs well enough. Ask anyone from that era about who Bob Dylan was and no one will tell you his main them was love fades. He was a protest singer, and a singer about the hard knocks of life. He was part of the anti-war movement. Love fades? That would be a dumb computer counting words. How would Watson see that many of Dylan’s songs were part of the anti-war movement? Does he say anti-war a lot? Probably he never said it in a song.

This is from this site: 

In our No. 1 Bob Dylan protest song, 'The Times They Are a-Changin,' Dylan went all out and combined the folk protest movement of the 1960's with the civil rights movement. The shorter verses piled upon one another in a powerful way, and lyrics like, "There’s a battle outside and it is ragin’ / It’ll soon shake your windows and rattle your walls / For the times they are a-changin'," are iconic Dylan statements that manage to transcend the times.
But he doesn’t mention Viet Nam or Civil Rights. So Watson wouldn't know that he had anything to do with those issues. It is possible to talk about something and have the words themselves not be very telling. Background knowledge matters a lot. I asked a 20 something about Bob Dylan a few days ago and he had never heard of him. He didn’t know much about the 60’s. Neither does Watson. You can’t understand words if you don’t know their context.

Suppose I told you that I heard a friend was buying a lot of sleeping pills and I was worried. Would Watson say I hear you are thinking about suicide? Would Watson suggest we hurry over and talk to our friend about their problems? Of course not. People understand in context because they know about the world and real issues in people’s lives. They don’t count words. 

Here is more from that site:

Saying that Bob Dylan is the father of folk music is probably overstepping a bit. However, saying that the vocalist is one of the most prominent writers of anti-war and protest songs in the 20th century is spot on, thus making him worthy of a Top 10 Bob Dylan Protest Songs list. The singer did change his range from anti-establishment to country to pop and back to folk again, and he remains a seminal force for those who rage against “The Man.”   

That was written by a human. How do I know? Because Watson can’t draw real conclusions by counting words in 800 million pages of text.

Of course, what upsets me most is not Watson but what IBM actually says. From the quote above:

Unlike traditionally programmed computers, cognitive systems such as Watson understand, reason and learn. 

Ann Rubin, IBM's vp of branded content and global creative, told Adweek that the commercials were needed to help people understand the new world of cognitive computing.

I wrote a book called The Cognitive Computer in 1984:

I started a company called Cognitive Systems in 1981. The things I was talking about then clearly have not been read by IBM (although they seem to like the words I used.) Watson is not reasoning. You can only reason if you have goals, plans, and ways of attaining them, and a comprehension of the beliefs that others may have and a knowledge of past experiences to reason from. A point f view helps too. What is Watson’s view on ISIS for example? 

Dumb question? Actual thinking entities have a point of view about ISIS. Dogs don’t but Watson isn't as smart as a dog either. (The dog knows how to get my attention for example.)

I invented a field called Case Based Reasoning in the 80’s which was meant to enable computers to compare new situations to old ones and then modify what the computer knew as a result. We were able to build some useful systems. And we learned a lot about human learning. Did I think we had created computers that were now going to outthink people or soon become conscious? Of course not. I thought we had begun to create computers that would be more useful to people. 

It would be nice if IBM would tone down the hype and let people know what Watson can actually do can and stop making up nonsense about love fading and out thinking cancer. IBM is simply lying now and they need to stop.

AI winter is coming soon.

Friday, November 6, 2015

Learning and Technology buzz words examined in order to enable massive peer to peer learning

I have just returned from another learning and technology conference where the people there used so many buzz words concerning the latest learning solutions that I am beginning to think madmen have taken over the field.

Here are some of my favorites with some comments:

Massification: this means that we are now dissatisfied with the idea that we can only stuff 1000 people into a classroom to hear a boring lecture, so now we want millions in a virtual classroom. Wow! A big improvement! Think of the money we can save/make.

Flipped Classrooms: since classrooms are such a good idea, we should make them in a new way that allows people to be bored by the lecture at some other time and then have a discussion about a lecture they didn’t care about (and probably didn’t watch) in the first place.

Social Construction of Knowledge: apparently we can’t know anything without discussing it on social media. Then, we know what everyone thinks about it. This is perhaps why people post pictures of what they had for lunch so they can make they sure their knowledge of what they had for lunch was correct and the lunch actually happened.

E-assessment: this is very important because if we didn’t assess everybody then how would we know that they know what we told them to know. (We don’t really care if they can do anything with all this knowledge.) We must turn all materials into quizzes as efficiently as we can so that people can pass tests and then immediately forget what was on that test knowing they will never need that stuff again. Do it online, and it is way better. I am not sure why. But I know we need to assess everything all the time and do it fast.

Learning Incentives. I hear that if you put really boring material in an animation then it becomes less boring. I hear that if you pay people to get good grades then they will learn more. I hear that if you offer people a good meal they will learn stuff whenever they are hungry.

Learning Analytics. This means that we can know how you learned, what you learned, and when you learned it. I am still trying to figure out when I learned that learning has become a moronic field. I have analyzed it but I don’t have enough data.

Blended Learning. This means we will do the crap we always did, but some of it will be online.

Nano-learning. This means that no one wants to take a course any more, which is a good thing because courses are usually quite tedious, especially when they are full of “content.” So, now people want to learn in small chunks. The next time I want someone to be my doctor I will inform him that I need to make sure that he learned to be a doctor through nano learning. We will see how well that works out.

Content. This is stuff we need to put in online courses. It is the same stuff we had in the courses that weren't online before. It is usually massive amounts of text. Why reading text on a computer is better than reading it in book is not something I can explain, but I am sure that content is king, and there must be a lot of it. We can ignore it as we always did. The idea that the computer allows you to do things rather than read things has apparently not been considered.

Learning Styles: everyone learn differently, or so I hear. So, that means there are some people who don’t learn by trying something out, figuring out what they did well and what they did wrong, and then possibly getting help from others, and then trying again. I wonder who those people are.

Collaborative Learning. This is when people learn together rather than learn in a world where only they exist. In the world that I know, where other people do exist, all real learning is collaborative, so this shouldn’t be worth mentioning. Even if you figure something out all by yourself, you will be telling someone, they will be reacting, and you will change your point of view slightly. When I said that all learning is a conversation, apparently what I meant to say is that it is collaborative. That is a very nice word.

Learning hasn’t changed. We all (including other mammals) learn in the same way, by trying things out, and hopefully getting one-on-one teaching from a parent or mentor when we are in trouble and need help. Learning has always been like that, and always will be like that. It is school that needs to change, not learning. Learning needs technology to the extent that it can transform school-based ideas into the way we always learned before there was technology and school. But, silly me, I thought that all these people who work on learning would know better than to copy school and then think they can improve learning by adding technology and cute buzz words. I was wrong.

Oh, I left out badges. We don’t need no stinkin’ badges.