I want to build general-purpose robots in the future. What to study now?
March 7, 2017 5:06 AM Subscribe
Granting the premise—however (un)realistic—that general-purpose robots will begin to enter the industrial and consumer realms within 20-30 years, what career path should I go down today in order to be able to help bring them about, and to continue working on them into the future?
Let's define general-purpose robots as robots that can learn, adapt to, and perform a fairly broad variety of tasks in a fairly broad variety of contexts. Contrast that with robots available today and in the near future, which are oriented towards very specific tasks and contexts: vacuum robots, assembly-line robots, drones, self-driving cars, etc. Narrow-purpose robots like these have their functions reflected in their forms. General-purpose robots, on the other hand, would probably look more humanoid or animal-like, as a natural consequence of needing to be able to do many different things. Today's cute precursors of general-purpose robots might be things like Honda's Asimo, or the crazy kill-bots that Boston Dynamics is building, or the wacky experiments that participated in the final DARPA Robotics Challenge in 2015.
Robotics is already an inter-disciplinary field, and I imagine that's going to become even more true / pronounced as robots become more complex and capable. So, I'd probably have to specialize, but I don't know how to choose what to study. About me: I'm a former artist (graphic/web design, illustration, 3D modeling—all self-taught) who's tired of art in general and wants to do something nerdier. (Here's some random 3D models I've done, for baseline nerd cred.) I left my design job a while back, and am considering going back to school. I've always been interested in technology in general, and computers and robots in particular, along with all the attendant fever-dreamy sci-fi stuff.
Some potential college majors:
• Robotics engineering
• Mechatronics engineering
• Electronic engineering (growth is stagnant, might be hard to find work)
• Computer hardware engineering (ditto)
• Mechanical engineering
• Biomedical engineering
• Programming / AI
• Neuroscience
• Industrial design
As I type this out, it's clear to me that this post is actually multiple questions in one:
1. How can I decide what direction(s) of study I'd enjoy and be good at?
2. Based on the set of answers from #1, what in particular should I study to be able to work on general-purpose bots later on in life?
3. What kind of work should I do after college (most likely initially unrelated to general-purpose bots) to align myself along a trajectory to work on said badass androids later in life?
To experiment and get a feel for what I like, I plan on plowing through a bunch of math, science, and programming courses on Khan Academy, and on getting a Mindstorms / Vex kit, plus maybe an Arduino. What else would you guys recommend I play around with, to get a feel for what aspect of robotics I might be drawn to?
There's answers to similar questions on here that I'm skimming through, but I figured this was different enough to warrant its own topic. Thanks in advance!
(Bonus question: I'm teaching myself Japanese, completely apart from the whole robotics thing. But if I do decide to go into robotics, should I stick with learning Japanese, or switch to Korean, or German, or something else? The subtext: do you see the USA being a leader in advanced robotics in the year 2040+, or should I plan to get out of Dodge?)
Let's define general-purpose robots as robots that can learn, adapt to, and perform a fairly broad variety of tasks in a fairly broad variety of contexts. Contrast that with robots available today and in the near future, which are oriented towards very specific tasks and contexts: vacuum robots, assembly-line robots, drones, self-driving cars, etc. Narrow-purpose robots like these have their functions reflected in their forms. General-purpose robots, on the other hand, would probably look more humanoid or animal-like, as a natural consequence of needing to be able to do many different things. Today's cute precursors of general-purpose robots might be things like Honda's Asimo, or the crazy kill-bots that Boston Dynamics is building, or the wacky experiments that participated in the final DARPA Robotics Challenge in 2015.
Robotics is already an inter-disciplinary field, and I imagine that's going to become even more true / pronounced as robots become more complex and capable. So, I'd probably have to specialize, but I don't know how to choose what to study. About me: I'm a former artist (graphic/web design, illustration, 3D modeling—all self-taught) who's tired of art in general and wants to do something nerdier. (Here's some random 3D models I've done, for baseline nerd cred.) I left my design job a while back, and am considering going back to school. I've always been interested in technology in general, and computers and robots in particular, along with all the attendant fever-dreamy sci-fi stuff.
Some potential college majors:
• Robotics engineering
• Mechatronics engineering
• Electronic engineering (growth is stagnant, might be hard to find work)
• Computer hardware engineering (ditto)
• Mechanical engineering
• Biomedical engineering
• Programming / AI
• Neuroscience
• Industrial design
As I type this out, it's clear to me that this post is actually multiple questions in one:
1. How can I decide what direction(s) of study I'd enjoy and be good at?
2. Based on the set of answers from #1, what in particular should I study to be able to work on general-purpose bots later on in life?
3. What kind of work should I do after college (most likely initially unrelated to general-purpose bots) to align myself along a trajectory to work on said badass androids later in life?
To experiment and get a feel for what I like, I plan on plowing through a bunch of math, science, and programming courses on Khan Academy, and on getting a Mindstorms / Vex kit, plus maybe an Arduino. What else would you guys recommend I play around with, to get a feel for what aspect of robotics I might be drawn to?
There's answers to similar questions on here that I'm skimming through, but I figured this was different enough to warrant its own topic. Thanks in advance!
(Bonus question: I'm teaching myself Japanese, completely apart from the whole robotics thing. But if I do decide to go into robotics, should I stick with learning Japanese, or switch to Korean, or German, or something else? The subtext: do you see the USA being a leader in advanced robotics in the year 2040+, or should I plan to get out of Dodge?)
Robotics is an obvious one. You could major in comp sci and go to a top 10 robotics program for graduate school. The CS degree would give you a solid foundation in general computing and software. I met a guy recently who had gone into robotics and switched to data science for some reason. So you'd have options.
posted by deathpanels at 5:43 AM on March 7, 2017 [1 favorite]
posted by deathpanels at 5:43 AM on March 7, 2017 [1 favorite]
I teach in an information science program. We have an undergraduate major in Information Technology and Informatics. Not exactly what you're looking for, but I do want to put in a big push for you to take some coursework in science, technology, and society; sociology; human-computer interaction; and informatics. We really need the people building these systems to understand the social aspect of technology. A year-old syllabus for my class in social informatics is online as a PDF of you want to take a look at some of the 101 social side of tech stuff that we require our majors to know before they graduate. Look at the departments of communications, sociology, and information science (if the schools you're looking at have that) to see if there is coursework in this area and plan to take some of these classes. Feel free to contact me if you want to discuss further. Good luck!
posted by k8lin at 5:44 AM on March 7, 2017 [4 favorites]
posted by k8lin at 5:44 AM on March 7, 2017 [4 favorites]
If you want to build the robots, you'll need CS, EE, mechanical engineering, and computer vision, all of which are fields that already exist and should be available at any large university. There will have to be specialists in all of these areas to build general-purpose robots, but I don't think any one person is going to know all of them.
But in every tech revolution, only a tiny number of people actually build the technology, and these are often not even great jobs while the revolution is happening (working for a bunch of tiny startups is not actually fun). A much larger number of people are around to market, promote, and exploit the technology to do different things, and these people work way less hard and make way more money. I think then the best thing to do would be to go into marketing, put together the absolute best blogs and books and youtube channels about robotics, and get ready to document the revolution. Then, when a general purpose robotics company comes around, go into sales for them. You'll make a lot more money with a lot less brain power.
posted by miyabo at 6:26 AM on March 7, 2017 [1 favorite]
But in every tech revolution, only a tiny number of people actually build the technology, and these are often not even great jobs while the revolution is happening (working for a bunch of tiny startups is not actually fun). A much larger number of people are around to market, promote, and exploit the technology to do different things, and these people work way less hard and make way more money. I think then the best thing to do would be to go into marketing, put together the absolute best blogs and books and youtube channels about robotics, and get ready to document the revolution. Then, when a general purpose robotics company comes around, go into sales for them. You'll make a lot more money with a lot less brain power.
posted by miyabo at 6:26 AM on March 7, 2017 [1 favorite]
Following on from miyabo:
I doubt the robots will be able to fix themselves for a very long time, so robot repair and maintenance should be a job for a pretty long time. It currently is a job, in fact: industrial maintenance and repair. All those manufacturing and packaging lines are robots.
By all means, do some CS stuff. Focus on ladder logic. Get good with a wrench. If you're a clumsy nerd as I once was, getting an Airframe & Powerplant mechanic's license from the FAA (assuming you're American) will make you a jack of all trades in two years.
posted by booooooze at 6:47 AM on March 7, 2017 [1 favorite]
I doubt the robots will be able to fix themselves for a very long time, so robot repair and maintenance should be a job for a pretty long time. It currently is a job, in fact: industrial maintenance and repair. All those manufacturing and packaging lines are robots.
By all means, do some CS stuff. Focus on ladder logic. Get good with a wrench. If you're a clumsy nerd as I once was, getting an Airframe & Powerplant mechanic's license from the FAA (assuming you're American) will make you a jack of all trades in two years.
posted by booooooze at 6:47 AM on March 7, 2017 [1 favorite]
Best answer: Hi, I'm a robotics engineer. Have been building robots for the last 16 years. I have opinions.
First off - decide whether you want to contribute through academia or industry. You're more likely, in academia, to have your hands on the real technology of the future before you are in industry. But you're far more likely to be in a position to make it real and bring it to market in industry. Which of those appeals to you more?
One thing I'm observing right now is that you have a lot of people developing what they think is THE general-purpose robot AI software and they put almost no effort into making the hardware general-purpose, or they focus entirely on the hardware (arms! hands! etc) and there's no decent software to back it up and make it as useful as the embodiment suggests it could be. Finding ways to really mesh those two together will be important. And I bring this up to emphasize that not all the cool robot problems are in software, which a lot of people seem to feel - creating incredibly useful, reliable, robust, elegant, and still cost-effective hardware is actually *really fucking hard* and a company that can manage to do that in addition to making kickass software will do very well for themselves.
As for what to major in - job 1 is to figure out what interests you. Robotics is highly multidisciplinary and I'd say a team making the robots you're interested in would involve software engineers, mechanical engineers, electrical engineers, human factors/user experience engineers, artists, and industrial designers. If any of those really floats your boat, that's great. You can start to learn what really does interest you by finding people who do these jobs and learning what their days are like.
You can also play with hobby kits and learn things there, but... mindstorms and Vex are all great, but they're really very intro-level. Robotics engineers and researchers design and build their own mechanical and electrical components, they use C++ and ROS, and they are probably really familiar with Linux and its associated computer vision libraries. As you start to identify what disciplines interest you, build your vocabulary around CAD, circuit and PCB design, and robot software engineering to see if it's still stuff that's interesting.
Robots Engineering degrees I find to be questionably useful in industry. They make a lot of sense if your path is academic and you'll be doing a lot of graduate research where you have to know a bit of everything. In industry, there are no "robotics engineering" departments in robotics companies - there are mechanical engineers, electrical engineers, and software engineers, all of whom have a strong interest/speciality in robotics, so hiring managers will struggle to find where you fit in. If you can find one of the traditional disciplines and focus in it, while still learning bits and pieces of the other two, I think that's a good thing. When you do controls, for example, there's a lot of overlap between the software and electrical sides of it, but I have two kinds of engineers working on controls problems, not one, because the project benefits from those two deep banks of knowledge.
After college, find the companies whose products inspire you and work for them. You'll learn all the things you think are good and bad about those products, and the reasons you're unhappy with them will drive you through your career path to find the companies that increasingly align with what you really want to be doing. Sometimes no one will make you happy and you will be depressed with the industry. It happens. I've learned to power through.
I also just want to call into question your interest in general-purpose robots, and ask, if that is *not* the future of the industry, is this still something you want to pursue? The reason I ask is that having worked in the industry for a long time, hearing someone say "I want to work on general purpose robots in 20-30 years" is approximately like hearing someone in 2000 say "I want to work on general purpose computers in 20-30 years". There really are no general purpose computers anymore - computers are simply in everything, and you don't develop "computer software", you develop mobile software, embedded software, robotics software, enterprise software, web applications, etc etc etc. Same goes for the hardware side. I genuinely feel that this is where robotics is going, too. (See me talk about it here, even!) We've been sold the idea of your all-purpose humanoid robots by sci-fi for decades, but that's not how the industry is trending, and not how similar technologies have trended over the last years. So maybe you are right and I am wrong, and that's fine, but if that is not the case - is this still a path you're willing to follow? If not, better to figure that out now than to be deeply disappointed in 10 years.
posted by olinerd at 9:05 AM on March 7, 2017 [7 favorites]
First off - decide whether you want to contribute through academia or industry. You're more likely, in academia, to have your hands on the real technology of the future before you are in industry. But you're far more likely to be in a position to make it real and bring it to market in industry. Which of those appeals to you more?
One thing I'm observing right now is that you have a lot of people developing what they think is THE general-purpose robot AI software and they put almost no effort into making the hardware general-purpose, or they focus entirely on the hardware (arms! hands! etc) and there's no decent software to back it up and make it as useful as the embodiment suggests it could be. Finding ways to really mesh those two together will be important. And I bring this up to emphasize that not all the cool robot problems are in software, which a lot of people seem to feel - creating incredibly useful, reliable, robust, elegant, and still cost-effective hardware is actually *really fucking hard* and a company that can manage to do that in addition to making kickass software will do very well for themselves.
As for what to major in - job 1 is to figure out what interests you. Robotics is highly multidisciplinary and I'd say a team making the robots you're interested in would involve software engineers, mechanical engineers, electrical engineers, human factors/user experience engineers, artists, and industrial designers. If any of those really floats your boat, that's great. You can start to learn what really does interest you by finding people who do these jobs and learning what their days are like.
You can also play with hobby kits and learn things there, but... mindstorms and Vex are all great, but they're really very intro-level. Robotics engineers and researchers design and build their own mechanical and electrical components, they use C++ and ROS, and they are probably really familiar with Linux and its associated computer vision libraries. As you start to identify what disciplines interest you, build your vocabulary around CAD, circuit and PCB design, and robot software engineering to see if it's still stuff that's interesting.
Robots Engineering degrees I find to be questionably useful in industry. They make a lot of sense if your path is academic and you'll be doing a lot of graduate research where you have to know a bit of everything. In industry, there are no "robotics engineering" departments in robotics companies - there are mechanical engineers, electrical engineers, and software engineers, all of whom have a strong interest/speciality in robotics, so hiring managers will struggle to find where you fit in. If you can find one of the traditional disciplines and focus in it, while still learning bits and pieces of the other two, I think that's a good thing. When you do controls, for example, there's a lot of overlap between the software and electrical sides of it, but I have two kinds of engineers working on controls problems, not one, because the project benefits from those two deep banks of knowledge.
After college, find the companies whose products inspire you and work for them. You'll learn all the things you think are good and bad about those products, and the reasons you're unhappy with them will drive you through your career path to find the companies that increasingly align with what you really want to be doing. Sometimes no one will make you happy and you will be depressed with the industry. It happens. I've learned to power through.
I also just want to call into question your interest in general-purpose robots, and ask, if that is *not* the future of the industry, is this still something you want to pursue? The reason I ask is that having worked in the industry for a long time, hearing someone say "I want to work on general purpose robots in 20-30 years" is approximately like hearing someone in 2000 say "I want to work on general purpose computers in 20-30 years". There really are no general purpose computers anymore - computers are simply in everything, and you don't develop "computer software", you develop mobile software, embedded software, robotics software, enterprise software, web applications, etc etc etc. Same goes for the hardware side. I genuinely feel that this is where robotics is going, too. (See me talk about it here, even!) We've been sold the idea of your all-purpose humanoid robots by sci-fi for decades, but that's not how the industry is trending, and not how similar technologies have trended over the last years. So maybe you are right and I am wrong, and that's fine, but if that is not the case - is this still a path you're willing to follow? If not, better to figure that out now than to be deeply disappointed in 10 years.
posted by olinerd at 9:05 AM on March 7, 2017 [7 favorites]
Response by poster: Hey everybody. Thanks a lot for all the different input. It's great to hear different viewpoints, and have people point out problems and challenges, and options I hadn't even considered. Really appreciate it. I'm taking notes on what everybody's saying, and I'll be looking into it all.
A couple replies to specific people:
miyabo: ooh, computer vision sounds sweet. And both you and Kalmya pointed out that I could get a job in robotics that doesn't involve any actual engineering, which is an interesting thought. Dunno if I'd last long in sales in particular though, haha. I do feel like I at least want to explore engineering of some kind, to see if I'd be into it. I'm not afraid of hard work, as long as I'm doing something I enjoy. (Famous last words?)
booooooze: another interesting thought. I do like fixing things.
olinerd: hot damn, thank you, this is a gold-mine. I'll check out your video in a sec. Some questions and comments directed mostly at you, but maybe others have input too:
posted by pableaux at 4:39 PM on March 7, 2017
A couple replies to specific people:
miyabo: ooh, computer vision sounds sweet. And both you and Kalmya pointed out that I could get a job in robotics that doesn't involve any actual engineering, which is an interesting thought. Dunno if I'd last long in sales in particular though, haha. I do feel like I at least want to explore engineering of some kind, to see if I'd be into it. I'm not afraid of hard work, as long as I'm doing something I enjoy. (Famous last words?)
booooooze: another interesting thought. I do like fixing things.
olinerd: hot damn, thank you, this is a gold-mine. I'll check out your video in a sec. Some questions and comments directed mostly at you, but maybe others have input too:
- Re: academia vs. industry, hard to say, and I was thinking about that as I wrote up my original post. I guess I'm putting the cart before the horse in general here, since I haven't really explored robotics yet at all, and don't know what areas and what kind of work might draw me in.
- Re: artists, how much work / influence does an artist have at robotics companies today? (I'm guessing by "artist," you mean "industrial designer" in this case?) Like, is the look / styling of a given robot determined more by an engineer (who has to decide on materials and other functional stuff that ends up inadvertently contributing to the overall aesthetic of a robot), or is an artist / industrial designer involved to a large degree? Do you see roles for artists in robotics growing as time goes on? If I wanted a smooth transition to working in robotics as an artist, what kinds of things should I study?
- The question of whether I'd like working in robotics in general, even if general-purpose bots and androids never come about, is another interesting one, and something I've thought about too. Again, I guess I have to explore, see if I'm into the work in general, regardless of whether or not I'm working on crazy sci-fi stuff. But regardless, I gotta imagine even if most robots in the future are purpose-built for narrow tasks, a small subset of them will be somewhere between narrow-purpose and general-purpose. For example, elder-care robots in Japan, or disaster recovery robots (like the very primitive ones in the last DARPA challenge) will by nature need to be fairly general-purpose, won't they? And innovations in those areas might trickle down to consumer robots and toys, and economies of scale will bring down costs, etc etc. Are you saying you guys in the industry have serious doubts about any of this ever happening in a major way in 20-30 years?
posted by pableaux at 4:39 PM on March 7, 2017
Best answer: In order:
- academia vs industry: fair enough! Just something to think about.
- artists: it will really vary by company and by application. If you have a look at some of the "home robots" out there right now - Jibo, Mykie, Kuri, and many others - they all have the same look (and I think it is LAME). As time goes on, people will stop trying to make the "robot Apple would make" and start thinking about how the design plays in to peoples' decor and interactions. So yeah, maybe industrial design, maybe some generic "user experience" or "user interaction" studies where you spend some time on the psychology of what technology looks like and how that changes people's interactions with it. Best thing to do would be to ask these companies what they call these jobs - i honestly don't know and I bet it varies widely. Here's an interesting one - a "Character AI Developer". It's a pretty hard technical job description, but it's fundamentally a more artistically creative job than your average software engineer would have. But the people who focus on look and branding DEFINITELY have influence beyond the mechanical designers responsible for materials and manufacturability.
- If your definition of "general purpose" is humanoid, I'm slightly more optimistic for you than I am on the idea of a truly "general purpose" (e.g. human-capability) robot. But have a google for disaster recovery or search and rescue robots and see just how many different designs people have come up with to do that job - humans actually aren't great for every job. We use a lot of tools. Bipeds don't do a lot of disaster stuff - we use bulldozers, cranes, levers, IR cameras, etc. Why build a humanoid to use those tools if you can effectively roboticize those tools? Elder-care robots are likewise going to be different from disaster recovery robots because they won't need to have the shielding, armoring, etc that disaster recovery robots will, and will be far more useful for "picking up people" than "moving heavy pieces of broken concrete" so will probably benefit from different arms/hands and won't need the expense of such high-torque motors and transmissions. I'm sure humanoid robots will be somewhere, likely in eldercare and very people-facing applications, but I'd hesitate to call those "general purpose" robots.
posted by olinerd at 7:38 PM on March 7, 2017 [1 favorite]
- academia vs industry: fair enough! Just something to think about.
- artists: it will really vary by company and by application. If you have a look at some of the "home robots" out there right now - Jibo, Mykie, Kuri, and many others - they all have the same look (and I think it is LAME). As time goes on, people will stop trying to make the "robot Apple would make" and start thinking about how the design plays in to peoples' decor and interactions. So yeah, maybe industrial design, maybe some generic "user experience" or "user interaction" studies where you spend some time on the psychology of what technology looks like and how that changes people's interactions with it. Best thing to do would be to ask these companies what they call these jobs - i honestly don't know and I bet it varies widely. Here's an interesting one - a "Character AI Developer". It's a pretty hard technical job description, but it's fundamentally a more artistically creative job than your average software engineer would have. But the people who focus on look and branding DEFINITELY have influence beyond the mechanical designers responsible for materials and manufacturability.
- If your definition of "general purpose" is humanoid, I'm slightly more optimistic for you than I am on the idea of a truly "general purpose" (e.g. human-capability) robot. But have a google for disaster recovery or search and rescue robots and see just how many different designs people have come up with to do that job - humans actually aren't great for every job. We use a lot of tools. Bipeds don't do a lot of disaster stuff - we use bulldozers, cranes, levers, IR cameras, etc. Why build a humanoid to use those tools if you can effectively roboticize those tools? Elder-care robots are likewise going to be different from disaster recovery robots because they won't need to have the shielding, armoring, etc that disaster recovery robots will, and will be far more useful for "picking up people" than "moving heavy pieces of broken concrete" so will probably benefit from different arms/hands and won't need the expense of such high-torque motors and transmissions. I'm sure humanoid robots will be somewhere, likely in eldercare and very people-facing applications, but I'd hesitate to call those "general purpose" robots.
posted by olinerd at 7:38 PM on March 7, 2017 [1 favorite]
« Older Weight training, nutrition and fitness blogs and... | Wordpress page load tile image issue Newer »
This thread is closed to new comments.
But I think you're wondering what part of bringing them to life you may also love. Do you really like mechanical or electrical engineering? How about computer science? If you chose to get a degree in thise fields, try to identify potential awesome jobs you'd love to do today with those skills.
Flip side, you could try to get ANY job at current robot producing companies, just because you're interested. Then you could see what different folks do now and potentially train while you work in admin or whatnot.
posted by Kalmya at 5:13 AM on March 7, 2017 [1 favorite]