Dr. Peter Jansen, Cognitive Sciences
Dr. Peter Jansen is an eclectic scientist with the heart of a Trekkie. His Tricorder designs, inspired by the science of Star Trek, are open source and available online for anyone to build, experiment with and enhance. Dr. Jansen is also active in the Artificial Intelligence world, where he is taking "baby steps" in creating robots that think for themselves!
"Star Trek" has immensely inspired your scientific career - what was it that appealed to you? Do any specific episodes or childhood memories come to mind?
My dad and I used to watch Star Trek together when I was little, and so I really grew up watching these amazing scientifically-inspired stories about exploring space. He's very much a science person, and is always asking questions and curious to learn about how the world works. For work, he'd design these enormous computer controlled machines for all sorts of things -- building automotive parts, automated packing, etc., and has a workshop in the basement where he'd often bring work home with him. My dad taught me how to make, create, design, build, program, and solder from a young age, and I think the combination of having him encourage me to learn more about the world, and encourage me to build things, led me to really want to become a scientist and make the technology that I was so inspired by in Star Trek a reality. I secretly want to explore space, and when I (as a kid) learned we couldn't really do that yet, and that we hadn't even been back to the Moon in my lifetime, I was really inspired to set out and make it happen. It turns out it's a little tricky to make warp drives, so I ended up having to start a little smaller!
In this way, I don't think any specific episode of Star Trek inspired me -- it was more the combination of watching it with my dad while being positively encouraged to learn more about science and making.
My dad and I used to watch Star Trek together when I was little, and so I really grew up watching these amazing scientifically-inspired stories about exploring space. He's very much a science person, and is always asking questions and curious to learn about how the world works. For work, he'd design these enormous computer controlled machines for all sorts of things -- building automotive parts, automated packing, etc., and has a workshop in the basement where he'd often bring work home with him. My dad taught me how to make, create, design, build, program, and solder from a young age, and I think the combination of having him encourage me to learn more about the world, and encourage me to build things, led me to really want to become a scientist and make the technology that I was so inspired by in Star Trek a reality. I secretly want to explore space, and when I (as a kid) learned we couldn't really do that yet, and that we hadn't even been back to the Moon in my lifetime, I was really inspired to set out and make it happen. It turns out it's a little tricky to make warp drives, so I ended up having to start a little smaller!
In this way, I don't think any specific episode of Star Trek inspired me -- it was more the combination of watching it with my dad while being positively encouraged to learn more about science and making.
It's a funny story that my parents probably laughed about 20 years ago, but my dad had taught me soldering and basic programming in the second grade, at a time when they're just starting to graduate you from writing with a large kid pencil to a normal sized one in school. So my parents had this kid who was beginning soldering and writing computer code, while he was still so young he had to use a fat pencil!
Your most recent project, designing a Tricorder like the concept found in "Star Trek", is creating a lot of excitement among science enthusiasts and Trekkies- but how would you explain the Tricorder to the "Average Joe"?
In Star Trek a Tricorder is a handheld device that looks sort of like a flip phone, that people can use to sense or detect pretty much anything you could imagine, from life signs to radiation to the chemicals that make up an object. An analogy that I like to use is that they're sort of like a "swiss army knife" of science -- they're very general tools that contain a variety of different sensors for learning as much as possible about the world, or something that you're scanning. The Tricorders that I've released are of course much simpler than the ones on the show -- they can detect basic things relating to atmospheric measurements (like temperature, pressure, or humidity), electromagnetic measurements about magnetic fields or light, and spatial measurements about distance or motion.
Gene Roddenberry, creator of "Star Trek", was a Futurist who had a vision of the Universe where everyone worked together in the name of exploration and knowledge. Do you share this vision, and does it relate to your decision to keep the Tricorder and other projects Open Source?
Very much so. I grew up in Canada with parents who were (and are) very generous even through very tough times, and very strongly believe that life is about bettering ones self and helping the people and the world around us. Accessibility and science education is a huge part of the Tricorder project, and I would feel incredibly honoured to be able to look back in 20 or 30 years and see that I had contributed to kids being able to learn more about their worlds, and make positive change.
My hope -- and I think, the real beauty of having the project be Open Source -- is that the Tricorders become a sort of "living project" with a community of folks working to make the Tricorders capable of sensing more, make them less expensive (and more accessible), and to find better ways of visualizing complex phenomena. If you have someone who knows about sensing radiation, or about chemical analysis, it's possible for them to use their expertise to add capabilities to the project -- but the Tricorders are simple enough that adding lights, off-the-shelf sensors, or swapping screens should be well within the reach of hobbiests.
A further (and more pragmatic) reason for having the project open source is that I think it would be difficult to raise the initial resources one would require to have the first batches of Tricorders mass produced. The Tricorders are really something different -- very general scientific tools -- and that's something I don't think we've ever seen or had access to before. While they're certainly useful for science education, I think it's going to take a few months (to a few years) of people actively walking around with them in their pockets before we find specific applications where they're not just great for science education, but they're also wonderfully economic tools that people would want to use everyday in their work. Using them to find heat leaks and other issues in home inspection is one such example where, in an applied sense, they have the potential to save folks a bunch of money. When that happens, I'm sure there will be lots of folks interested in inexpensively manufacturing Tricorders, and you won't have to be handy with a soldering iron to have one -- which is of course the goal, to get them in the hands of as many kids as you can, as inexpensively as you can, where they can do a lot of good.
The Tricorder designs you have developed are meant to be easy enough for young adults to build and use. You want to establish an intuitive relationship with natural processes we can not see. Have you gotten any feedback, videos, etc of kids building and using these Tricorders?
This is a good question. The Tricorder project website with the schematics, circuit board layouts, diagrams, parts lists, etc -- the source -- has only been online for a few weeks, the response has just been so incredible and folks have been so captivated by the idea of having their own Tricorders, that the project gained a huge amount of interest within only hours of going live. It generally takes a few weeks to have circuit boards manufactured, and so I wouldn't expect to see any Tricorders out in the wild for, at the earliest, another couple of weeks. To help this process along, I ran a contest on the Tricorder project for a bunch of parts to construct your own Tricorder that just ended, and I can't wait to see those (and other) parts put to good use. I'm really excited to see not just my original designs out there, but also to see how people modify them, and what sorts of sensors and extra features they build in.
That being said, when I have the Tricorders with me I absolutely love to give them to folks to play with. I recently gave an interview on PBS, and afterwards the stage crew came up and asked if they could use them, and I loved seeing them happily scanning each other and whatever they could find in the studio. It's exactly what I made them for --people using them to see what they can't normally see, and learn about the science of the everyday world that's all around us.
A further (and more pragmatic) reason for having the project open source is that I think it would be difficult to raise the initial resources one would require to have the first batches of Tricorders mass produced. The Tricorders are really something different -- very general scientific tools -- and that's something I don't think we've ever seen or had access to before. While they're certainly useful for science education, I think it's going to take a few months (to a few years) of people actively walking around with them in their pockets before we find specific applications where they're not just great for science education, but they're also wonderfully economic tools that people would want to use everyday in their work. Using them to find heat leaks and other issues in home inspection is one such example where, in an applied sense, they have the potential to save folks a bunch of money. When that happens, I'm sure there will be lots of folks interested in inexpensively manufacturing Tricorders, and you won't have to be handy with a soldering iron to have one -- which is of course the goal, to get them in the hands of as many kids as you can, as inexpensively as you can, where they can do a lot of good.
The Tricorder designs you have developed are meant to be easy enough for young adults to build and use. You want to establish an intuitive relationship with natural processes we can not see. Have you gotten any feedback, videos, etc of kids building and using these Tricorders?
This is a good question. The Tricorder project website with the schematics, circuit board layouts, diagrams, parts lists, etc -- the source -- has only been online for a few weeks, the response has just been so incredible and folks have been so captivated by the idea of having their own Tricorders, that the project gained a huge amount of interest within only hours of going live. It generally takes a few weeks to have circuit boards manufactured, and so I wouldn't expect to see any Tricorders out in the wild for, at the earliest, another couple of weeks. To help this process along, I ran a contest on the Tricorder project for a bunch of parts to construct your own Tricorder that just ended, and I can't wait to see those (and other) parts put to good use. I'm really excited to see not just my original designs out there, but also to see how people modify them, and what sorts of sensors and extra features they build in.
That being said, when I have the Tricorders with me I absolutely love to give them to folks to play with. I recently gave an interview on PBS, and afterwards the stage crew came up and asked if they could use them, and I loved seeing them happily scanning each other and whatever they could find in the studio. It's exactly what I made them for --people using them to see what they can't normally see, and learn about the science of the everyday world that's all around us.
Everyone these days is attached to their cell phones. Is there a difference between the functions of your Tricorder designs and what's available as an app? Have you or other people experimented with building this technology directly into a cell phone?
There's a great deal that's different between a Tricorder and a smartphone of today, although someday I imagine the two might become very similar devices. Smartphones and tablets often contain a few sensors -- most often light sensors for automatically adjusting the brightness of the screen to increase battery life, or accelerometers/gyros/compases to automatically rotate the screen from portrait to landscape, or for motion-based video games. Modern smart phones are extremely capable devices with a lot of graphical capabilities, and I've been experimenting with pairing Tricorders and smartphones together as one possible route to go, though I imagine for kids you'll definitely want an inexpensive standalone device that doesn't require an expensive phone.
Let's talk about some of your other work. You say you are working on Artificial Intelligence, using your Cognitive Science background to teach Robots how to interact and learn about their world the same way babies do. Can you elaborate on that?
You bet. I just finished my PhD, where my research focuses on trying to figure out how people represent things like concepts and language in their brains, and then tries to get computers to do something similar, and learn concepts and language like babies do. There's a lot of
research out there that suggests that infants know a lot about their worlds and have a brain full of concepts (like about what a dog or house or car is) well before they learn their first words, and that it's this rich pre-linguistic conceptual system that really helps bootstrap their earliest acquisition of language. A lot of early artificial intelligence research (and still some today) focused on trying to make a thinking machine by having a large group of researchers design a complicated knowledge system (sort of like a fancy database), and then try and figure out ways to get a computer to make inferences using that knowledge. Sometimes it would work, but usually only for very limited tasks, or for very specific and limited fields of knowledge.
The problem is, it's really hard to build, say, a tree. We're finding out that a better question to ask is, how could I build a /seed/, that if I design it simply and elegantly enough, will grow into a tree by virtue of nothing but it's exposure to the world and some simple rules that govern how the seed works.
My research focuses on trying to design something like a very early version of that seed. These take the form of cognitively-inspired self-organizing neural network models that are functionally similar to the neural networks in brains, and that (in my research) acquire concepts and language something like how we think babies do, through the sensory and motor experiences babies get by existing in and exploring their physical environment. Most of my work uses simulated worlds or data sets, but in the last few years folks have developed very sophisticated and inexpensive humanoid robots, and so I hope that one day soon I can get my hands on one and have it roaming about learning aspects of language in a way that resembles something like a very young infant.
It's important to note that this cognitively-inspired approach to artificial intelligence and language learning is just one theoretical camp in language learning. More traditional artificial intelligence techniques usually rely on massive statistical processing, which is particularly good for certain kinds of tasks -- Google and searching being just one example.
What do you imagine our relationship with artificially intelligent technology looks like in the future?
This is a bit coloured from my own research. I imagine that the accessibility of inexpensive humanoid robots that are progressively more able to have rich sensory experiences similar to infants is for the first time going to enable collecting rich datasets and experiments in conceptual and linguistic knowledge representation that simply haven't been possible until now, and are really essential to meaningful conceptual and linguistic processing going on. By "meaningful", I mean "semantic", in that computers will go from the extremely simple linguistic processing we have today largely centered around search and co-occurrence, and really begin to understand the /meaning/ behind the concepts and words that they're seeing and using. They'll know not just a word -- say, "cat" -- but also know from experience what a cat (and many other things) look like, how they behaves, how they might interact if you put them in a room together. I think transitioning from the "symbolic" computation that we've had for the last 70 years into "semantic" computation will really enable thinking machines, and I can't even imagine the discoveries and chances to our lives that that will create, from medicine to science and mathematics.
Are there any other projects in development that you can share with us?
One of my mentors in graduate school often had a bunch of independent projects going on at once, and I very much find myself the same. Something that I've become very interested in over the last 5 years is 3D printing, or more formally "rapid prototyping", or in Star Trek language, something like a "replicator". My dad and I built our first 3D printer a few years ago now, and I remember the absolute amazement we both had as we watched something that we'd built together slowly /itself/ build an object out of plastic. It was a very simple object -- just a box -- but we were both absolutely captivated, and sat in awe for 45 minutes as it slowly built a taller and taller box. That got us hooked. Today we've built a couple inexpensive 3D printers that are more sophisticated than that first one, and some have even printed off parts and upgrades for themselves from the thriving community of open source 3D printers and objects that have arisen in the past couple of years.
The inexpensive open source 3D printers of today are very cool, but still only capable of building objects out of a single material, usually plastic. I've begun sketching and collecting parts for a machine that can act as both a 3D printer, circuit board mill, pick and place machine, etc., all in a single machine, that I'm hoping will demonstrate autonomously constructing a very simple device all on it's own that contains not just plastic, but also electronics. I'm thinking of starting with making something simple, like a watch, and have the lofty dream of the machine constructing a little toy car or robot that rolls itself off the printer. I don't think that there are any technological reasons why we couldn't construct such a machine -- a factory on your desktop -- I think we just really have to want to do it, and put the time in. The RepRap project has the goal of constructing a 3D printer that can itself print out all the parts for another 3D printer, thus making a truly self-replicating machine. I think we're still a ways off from that, but the project has spawned a huge amount of interesting work, and creative non-traditional approaches to design. Like many things, it's the journey that we're especially excited for, and what we'll learn and develop along the way -- if we ever get too close to our goal, we usually take what we've learned and move our goal ahead to find even more interesting and exciting problems to solve. That's (at least partially) what Science is about!
There's a great deal that's different between a Tricorder and a smartphone of today, although someday I imagine the two might become very similar devices. Smartphones and tablets often contain a few sensors -- most often light sensors for automatically adjusting the brightness of the screen to increase battery life, or accelerometers/gyros/compases to automatically rotate the screen from portrait to landscape, or for motion-based video games. Modern smart phones are extremely capable devices with a lot of graphical capabilities, and I've been experimenting with pairing Tricorders and smartphones together as one possible route to go, though I imagine for kids you'll definitely want an inexpensive standalone device that doesn't require an expensive phone.
Let's talk about some of your other work. You say you are working on Artificial Intelligence, using your Cognitive Science background to teach Robots how to interact and learn about their world the same way babies do. Can you elaborate on that?
You bet. I just finished my PhD, where my research focuses on trying to figure out how people represent things like concepts and language in their brains, and then tries to get computers to do something similar, and learn concepts and language like babies do. There's a lot of
research out there that suggests that infants know a lot about their worlds and have a brain full of concepts (like about what a dog or house or car is) well before they learn their first words, and that it's this rich pre-linguistic conceptual system that really helps bootstrap their earliest acquisition of language. A lot of early artificial intelligence research (and still some today) focused on trying to make a thinking machine by having a large group of researchers design a complicated knowledge system (sort of like a fancy database), and then try and figure out ways to get a computer to make inferences using that knowledge. Sometimes it would work, but usually only for very limited tasks, or for very specific and limited fields of knowledge.
The problem is, it's really hard to build, say, a tree. We're finding out that a better question to ask is, how could I build a /seed/, that if I design it simply and elegantly enough, will grow into a tree by virtue of nothing but it's exposure to the world and some simple rules that govern how the seed works.
My research focuses on trying to design something like a very early version of that seed. These take the form of cognitively-inspired self-organizing neural network models that are functionally similar to the neural networks in brains, and that (in my research) acquire concepts and language something like how we think babies do, through the sensory and motor experiences babies get by existing in and exploring their physical environment. Most of my work uses simulated worlds or data sets, but in the last few years folks have developed very sophisticated and inexpensive humanoid robots, and so I hope that one day soon I can get my hands on one and have it roaming about learning aspects of language in a way that resembles something like a very young infant.
It's important to note that this cognitively-inspired approach to artificial intelligence and language learning is just one theoretical camp in language learning. More traditional artificial intelligence techniques usually rely on massive statistical processing, which is particularly good for certain kinds of tasks -- Google and searching being just one example.
What do you imagine our relationship with artificially intelligent technology looks like in the future?
This is a bit coloured from my own research. I imagine that the accessibility of inexpensive humanoid robots that are progressively more able to have rich sensory experiences similar to infants is for the first time going to enable collecting rich datasets and experiments in conceptual and linguistic knowledge representation that simply haven't been possible until now, and are really essential to meaningful conceptual and linguistic processing going on. By "meaningful", I mean "semantic", in that computers will go from the extremely simple linguistic processing we have today largely centered around search and co-occurrence, and really begin to understand the /meaning/ behind the concepts and words that they're seeing and using. They'll know not just a word -- say, "cat" -- but also know from experience what a cat (and many other things) look like, how they behaves, how they might interact if you put them in a room together. I think transitioning from the "symbolic" computation that we've had for the last 70 years into "semantic" computation will really enable thinking machines, and I can't even imagine the discoveries and chances to our lives that that will create, from medicine to science and mathematics.
Are there any other projects in development that you can share with us?
One of my mentors in graduate school often had a bunch of independent projects going on at once, and I very much find myself the same. Something that I've become very interested in over the last 5 years is 3D printing, or more formally "rapid prototyping", or in Star Trek language, something like a "replicator". My dad and I built our first 3D printer a few years ago now, and I remember the absolute amazement we both had as we watched something that we'd built together slowly /itself/ build an object out of plastic. It was a very simple object -- just a box -- but we were both absolutely captivated, and sat in awe for 45 minutes as it slowly built a taller and taller box. That got us hooked. Today we've built a couple inexpensive 3D printers that are more sophisticated than that first one, and some have even printed off parts and upgrades for themselves from the thriving community of open source 3D printers and objects that have arisen in the past couple of years.
The inexpensive open source 3D printers of today are very cool, but still only capable of building objects out of a single material, usually plastic. I've begun sketching and collecting parts for a machine that can act as both a 3D printer, circuit board mill, pick and place machine, etc., all in a single machine, that I'm hoping will demonstrate autonomously constructing a very simple device all on it's own that contains not just plastic, but also electronics. I'm thinking of starting with making something simple, like a watch, and have the lofty dream of the machine constructing a little toy car or robot that rolls itself off the printer. I don't think that there are any technological reasons why we couldn't construct such a machine -- a factory on your desktop -- I think we just really have to want to do it, and put the time in. The RepRap project has the goal of constructing a 3D printer that can itself print out all the parts for another 3D printer, thus making a truly self-replicating machine. I think we're still a ways off from that, but the project has spawned a huge amount of interesting work, and creative non-traditional approaches to design. Like many things, it's the journey that we're especially excited for, and what we'll learn and develop along the way -- if we ever get too close to our goal, we usually take what we've learned and move our goal ahead to find even more interesting and exciting problems to solve. That's (at least partially) what Science is about!
Find out more about Dr. Peter Jansen, scope out his Tricorder designs and learn more about the future of Artificial Intelligence on his website! http://www.tricorderproject.org/
No comments:
Post a Comment