by Richard Van
Intro: You’re listening to inQUERY, the podcast show run by the professional writing students here at York University. We're exploring the technology of today and creating the new ideas of tomorrow.
Richard Van: Hello, this is our podcast show inQUERY. I’m Richard Van. Today, I’d like to talk about machines and intelligence.
So I think everyone listening to a podcast like this right now is listening because they want to know something. Sure, every time I go to university for a lecture, I’m going because I want to learn something. Why do people want to learn? For most people, they want to learn something because they want to gain skills for, let’s say a job, or maybe because they need it to make something new and exciting like Twitter.
(Music: “Circle Round” by Spinning Clocks. Licensed under Creative Commons Attribution 4.0 International License.)
And why does this relate to intelligence? Basically, intelligence is the ability to acquire knowledge and skills, and apply it. Ever since humans as a species were created, we have been storing and applying knowledge. It’s only natural, right? We go out, gather some data and learn how to do something. And then we create some amazing things like spaceships
But what if we could create something that would generate knowledge by itself? How would we create something that could understand and add to human intelligence?
I’d like to bring your attention to devices we use every day. Devices like laptops and smartphones. These are applications of human knowledge that we use often, to stay in touch with friends and family, and to enhance our productivity. These days, we have so much to deal with at any given time that we often need some help staying on top of things. It shouldn’t be surprising that people have created programs to use on phones and laptops for the purpose of managing their needs and schedules.
(Music: “Is There Life In Here?” by Jim Rooster. Licensed under Creative Commons Attribution 3.0 United States License.)
If you regularly use a phone, you’ve probably heard of something called Siri, or Google Now. If you use Windows 10, you might have heard of Cortana. These are called intelligent personal assistants, and they’re meant to help people with planning and executing tasks. But in a different sense, they’re highly specialized chatterbots. Chat bots.
Chatterbots are programs which essentially chat with people like us. They’re conversational agents which take in our input and return a response. They’ve mostly been in the realm of tech demos and hobbyist projects, but more recently chat bots have been giving tech support pointers. They respond to us every day when you call in to help phone lines. And of course, they’ve also taken the role of intelligent personal assistants, and programs like Siri are indeed chat bots.
You can find out about more chat bots in the past in my blog post “Traditional Chat Bots: Chat bots in the last 50 years”. (This can be found here. https://stephanie-bell-m08b.squarespace.com/blog-season1/d3d1aef4-217e-4794-974e-2f9218e74c8f)
(Music: “No-End Ave” by Jahzzar. Licensed under Creative Commons Attribution-ShareAlike 4.0 International License.)
After surveying 18 people at York University, 100% (18/18) of survey respondents told that they have used an instant messaging program before. 61% (11/18) of respondents said that they had an intelligent personal agent in their back pocket, and it was a tool they used regularly as well. So chat bots are used quite often by respondents, and many respondents are open to the idea of talking with chat bots.
Now chat bots have traditionally been weak at holding actual conversations (Chakrabarti). Sure, they could tell you what the weather is like today, or which song is on the top of the Hot 100 music charts, but they usually wouldn’t help if they needed context from something you previously told them. Chat bots typically fail to respond appropriately in that situation. But that’s where intelligence comes into play. The solution to this problem lies in creating chat bots with artificial intelligence. Chat bots that can learn.
So how can chat bots learn, and how will smarter chat bots be useful? The key to learning lies in a field in computer science called machine learning. It concentrates on creating algorithms which can interpret sets of data and learn from them (Kohavi, Provost). Basically, these learning chat bots should be able to take into account anything you tell it previously before forming a response. This means that they will be able to answer any question you may have given proper context. With all of the data that companies track and store on the Internet, it means that eventually these smarter bots will give you a unique and (technically) correct answer. And the best part is that these smart bots aren’t far off.
I've linked Jeremy Howard’s TED Talk, "The wonderful and terrifying implications of computers that can learn" in the transcript: http://www.ted.com/talks/jeremy_howard_the_wonderful_and_terrifying_implications_of_computers_that_can_learn/). In just the first minute, Howard talks about a man named Arthur Samuel, and talks about in 1956, how Samuel was able to train a computer to play checkers so well that it beat the Connecticut state champion in checkers by 1962 (Howard). Fast forward to 2010, IBM’s Watson machine competed in Jeopardy! against 2 of the show’s champions and convincingly won against them (Ferrucci et al.). We can definitely create specialized machines to gather, learn and understand human knowledge.
Now those two examples weren’t particularly chat bots, but very recently Google unveiled a “neural conversational model” for natural language understanding (Vinyals, Le)… but we can just call its implementation a chat bot. Here’s an excerpt of the bot talking with a human about philosophical issues, voiced by me:
Human: what is the purpose of life ? Machine: to serve the greater good . Human: what is the purpose of living ? Machine: to live forever . Human: what is the purpose of existence ? Machine: to find out what happens when we get to the planet earth . Human: where are you now ? Machine: i ’m in the middle of nowhere .
Richard Van: This bot’s responses were all generated through training the chat bot on English words. And the paper published by the Google scientists that created this machine shows more conversations where the bot forms its responses through what it has learned.
(Music: “Last Dance” by Jahzzar. Licensed under Creative Commons Attribution-ShareAlike 3.0 International License.)
So in the end, machines do have the ability to learn and have intelligence similar to what us humans have. This was displayed through applying machine learning to chat bots. Machines can draw upon all human knowledge, as shown by machines such as Watson. They can also create knowledge, as shown by Google’s machine creating sentences by itself. But what is to come in the future? Will these machines be able to communicate like humans? That’s for us to decide.
Chakrabarti, Chayan. "Artificial Conversations for Chatter Bots Using Knowledge Representation, Learning, and Pragmatics." (2014). https://repository.unm.edu/bitstream/handle/1928/24299/ChakrabartiDissertation.pdf
Ferrucci, David, Eric Brown, Jennifer Chu-Carroll, James Fan, David Gondek, Aditya A. Kalyanpur, Adam Lally, J. William Murdock, Eric Nyberg, John Prager, Nico Schlaefer, and Chris Welty. "Building Watson: An Overview of the DeepQA Project." The AI Behind Watson — The Technical Article. AAAI, Fall 2010. http://www.aaai.org/Magazine/Watson/watson.php.
Howard, Jeremy. “The wonderful and terrifying implications of computers that can learn.” TED. December 2014. Lecture. http://www.ted.com/talks/jeremy_howard_the_wonderful_and_terrifying_implications_of_computers_that_can_learn
Jahzzar. No-End Ave. Free Music Archive, 2014. MP3.
Jahzzar. Last Dance. Free Music Archive, 2012. MP3.
Jim Rooster. Is There Life In Here?. Free Music Archive, 2015. MP3.
Kohavi, Ron, and Foster Provost. "Glossary of terms." Machine Learning 30.2-3 (1998): 271-274.
Metz, Cade. "Google Made a Chatbot That Debates the Meaning of Life." Wired.com. Conde Nast Digital, 26 June 2015. Web. 25 Nov. 2015. http://www.wired.com/2015/06/google-made-chatbot-debates-meaning-life/.
Spinning Clocks. Circle Round. Free Music Archive, 2015. MP3.
Vinyals, Oriol, and Quoc Le. "A neural conversational model." arXiv preprint arXiv:1506.05869 (2015). http://arxiv.org/pdf/1506.05869.pdf