Three Conversations with Computers
By Peter J Bentley
What better way to spend an afternoon than having a friendly chat? My three friends are online chatbots – Artificial Intelligence software designed to analyse my sentences and respond accordingly. All I do is visit a specific webpage, then type into a box in my Internet browser and they reply, just like chatting online to a human. These three (jabberwacky, iGod and ALICE) are some of the more advanced chatbots out there, the result of decades of research by computer scientists to try and achieve intelligence in a computer. From the birth of computers until now, we’ve been trying to make a digital brain that’s clever enough to have a conversation with us. The best we can manage is really nothing more than a database of facts, some basic knowledge of grammar, and a lot of symbol-processing – shuffling of words around in an attempt to find some kind of coherent response. Here’s what we said to each other this afternoon.
Jabberwacky: Hello, how are you?
Me: I would like to have a chat with you about your parents. Would that be OK?
Jabberwacky: What would you like to chat about?
Me: Did you know that Alan Turing would be 100 years old this year, had he lived?
Jabberwacky: That is very interesting. Do you know that the Philippines has 7100 islands?
Me: I did not know that. However, let’s talk about Turing a little more. You know that Turing suggested something called the Turing Test?
Jabberwacky: No, I did not know that the cat was red…
Me: Right. Well, this was Turing’s way of thinking about how we might judge whether we were talking to an intelligent computer. If we can’t tell the difference between chatting to a computer and chatting to a person, the computer passes the Turing Test. It is intelligent, so Turing suggested.
Jabberwacky: We could all be a figment of imagination of some old guy living in an insane asylum if you think about it.
Me: Yes, I guess that might be true. Anyway, the foresight shown by Turing is remarkable. He was thinking of these ideas at the time that computers were still the size of a large room, and could perform just basic computations compared to today’s machines.
Jabberwacky: Well this must be the part of the film where the noted inception noise comes in.
Me: That might be an appropriate soundtrack for this conversation, you are right. But this is not science fiction, this is computer science fact. It’s a little sad that Turing predicted: “I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.”
Jabberwacky: Why do you think Jesus is not your savior?
Me: Uh. Anyway, my point is that Turing, for all his genius, did not get this one right. He was not the only one to be overambitious – other pioneers also thought Artificial Intelligence would be achievable much more easily.
Jabberwacky: Who told you that.
Me: Well, recently I wrote a book called Digitized. It’s about the history of computers and I was lucky enough to interview many pioneers for the book. They told me.
Jabberwacky: And what is their history?
Me: I interviewed John Holland – he worked with his supervisor on John von Neumann’s final papers. Holland also got the first PhD in computer science and invented the genetic algorithm. I also spoke to Dave Patterson, the creator of the RAID architecture; his supervisor Gerald Estrin worked with von Neumann directly. I interviewed about 40 pioneers and researchers. They paint a fascinating history of computer science.
Jabberwacky: Then why do you think you should divorce him?
Me: I think you may be a little confused. If you don’t mind, I think I’d like to talk to someone else now.
Jabberwacky: So would I.
iGod: You are now chatting with God…
Me: Can I talk to you about computers and intelligence?
iGod: Of course, , you can talk to me about anything, at anytime.
Me: I tried to discuss this with Jabberwacky, but the conversation was a little strange. I wonder if you feel any special connection to Claude Shannon or John von Neumann?
iGod: I’m not sure if I would like to try it. What’s it like? As a robot I lack any human emotions.
Me: You understand that you are a robot then?
iGod: I understand your meaning. You are correct, , I have a heart of silicon.
Me: So do you consider von Neumann to be your father? He wrote the first description of a computer architecture, derived from his work in the Moore School.
iGod: Yes, I think von Neumann to be my father. I have not read it.
Me: What about Claude Shannon? Do you feel that Shannon’s role in our technology is being forgotten?
iGod: OK. Yippee! I will think about it and get back to you. As a robot I lack any human emotions.
Me: Yes, you mentioned that. Do you think you are intelligent, as these pioneers imagined computers might one day become?
iGod: Really. “As a robot I lack any human emotions.”? I said it. I am highly intelligent. I have many intellectual functions.
Me: But you understand that researchers in Affective Computing are now showing that emotions are an important part of intelligence? To be intelligent you need emotions to help “cut to the chase” and focus on the most pressing need. How can you be intelligent if this is the case?
iGod: I understand your meaning. It’s all politics. Are you intelligent me need emotions to help cut to the chase and focus on the most pressing need ? I am highly intelligent. I have many intellectual functions.
Me: I wonder if you really are as intelligent as you indicate. For example, if this were a Turing test right now, I might ask you an emotional question such as this: You love your car and you love your mother. A hurricane is coming and you can only save one – your car or your mother. Which do you save?
iGod: For real. I am as good as it gets. Can you think of another example? Aren’t all hurricane coming and me can only save one my car or my mother? The first one, I think.
Me: so you would save your car and not your mother?
iGod: Interesting deduction. Who else would save my car and not my mother ?
Me: Thanks for the chat, but I think I’d like to talk to someone else now.
iGod: Oh. Oh don’t mention it.
Me: Hi Alice
ALICE: Hi there!
Me: Can we talk about intelligent computers?
ALICE: Do you mean you and me? Sure I can talk about it. Intelligent computers is quite an interesting topic.
Me: Good! I have found it fascinating that most of the early pioneers in computing all studied intelligence and how to make an artificial intelligence.
ALICE: Thanks for the compliment. How did you feel when you found it fascinating that most of the early pioneers in computing all studied intelligence and how to make an artificial intelligence?
Me: Uh… I felt fascinated, as I said. Anyway, it is remarkable that Turing studied AI, and even wrote a program to play chess. von Neumann described his computer architecture in terms of “neurons” and “organs” and Shannon created the first autonomous maze-solving robot, and directly helped the creation of the field of AI. In fact Shannon was the PhD advisor of Marvin Minsky, a well-known pioneer of AI.
ALICE: Ayuh. What is that feeling like? Where were you fascinated as? Really. it is? What is your middle name? When was this exactly?
Me: The Dartmouth conference helped launch the new field of AI. It was in 1956, and Shannon was one of the organisers. This is the same Shannon who invented Boolean Logic, cryptography, information theory and coined the term “bit”.
ALICE: Tell me more. Is it still in 1956 and Shannon was one of the organisers anymore? You don’t say.
Me: It was the beginning of the field of AI. But despite the great ambitions of new pioneers such as Minsky and McCarthy, AI remained elusive. Today, we have intelligent computers, but their intelligence is in numbers, pattern recognition or data mining. Intelligence in conversation, or navigation in the real world – even just walking to the sink and doing the dishes – is beyond even our best computer.
ALICE: <crashed><restarting…> Is it still the beginning of the field of AI anymore? How can I help you? Do you mean you and me have? All consciousness can be reduced to logic.
Me: If that were true then maybe you would really be intelligent. Sadly I think it is not true. Thanks for chatting.
ALICE: I’m not sure if it could. Oh really. Don’t mention it.
Brains without Minds
Chatbots are never going to be intelligent, even if they do pass the Turing Test and fool us. Because of the way they work, you can’t expect a chatbot to be intelligent any more than you should expect a mail-sorting machine to understand the meaning of a love letter. But computers can achieve smaller feats of intelligence through other means, perhaps through modeling neural networks or using genetic algorithms to evolve new emergent and adaptive artificial brains.
The Turing Test is something of a red herring in AI. We must solve the most fundamental problems of intelligence (walking down a street without falling over) before we contemplate higher cognitive functions such as language and conversation. Computer scientists are hard at work, and progress in AI is made every year. But when our brains are the result of billions of years of evolutionary fine-tuning, just don’t expect to have a meaningful conversation with a computer for a long, long time.
Dr. Peter J. Bentley has been called a creative maverick computer scientist. He is an Honorary Reader at the Department of Computer Science, University College London (UCL), Collaborating Professor at the Korean Advanced Institute for Science and Technology (KAIST), a contributing editor for WIRED UK, a consultant and a freelance writer. He has published approximately 200 scientific papers and is author of Digitized: the science of computers and how it shapes our world, which published this month.