With the year ending, we would like to thank all our followers for reading Neurogadget throughout the year. As a New Year’s Eve special post, please welcome an exclusive interview with Steve Castellotti, one of the most successful entrepreneurs of the BCI market in 2012. Regular readers may remember a previous article about a Kickstarter campaign for the Puzzlebox Orbit, an open-source brain-controlled helicopter. A few weeks ago we met in San Francisco with Steve Castellotti, the founder of Puzzlebox, to discuss the Orbit helicopter, BCI’s, neuroscience in education, and the power of open-source technology.
Neurogadget: What was the first thing that got you interested in brain-computer interfaces?
Steve Castellotti: back in 2004 the BBC did an interview on BCI technology that the Wadsworth institute was doing. It was posted to Slashdot, and when I saw that I was amazed that they were doing mouse cursor control without invasive surgery. I ended up mocking up an interface that was based on a video demo they posted (which used targets that represented a group of letters) and you could make a text-to-speech interface based on sensory-motor control. I was in auckland NZ at the time, and the problem was there wasn’t any access to EEG tech that cost less than five figures.
NG: No commercially available BCI’s?
SC: Yeah, there was nothing available. I kept on top of the news of what was available. In 2009 I saw first the Emotiv, then NeuroSky, commercial BCI’s which cost more like two to three figures. As soon as they hit the market I jumped in, and started developing whatever I could. My first target was with Lego Mindstorms, the programmable robotics kits. I had previously used them in college, and I had worked at an elementary school that had several kits, the programmable robot seemed like an easy target. I started developing the Puzzlebox Brainstorm software (for Lego Mindstorms) but pretty quickly started getting contacts from NeuroSky, people who wanted to do special one-off projects that were not their core business model.
They were trying to foster a third party developer ecosystem, and they would occasionally toss one of these projects my way. On several instances people were looking for a television pitch, a game show or reality show, something with brain control. The brain-controlled crane project we did, in the UK was one of those. That was through the Gadget Show, who had arranged all the different players (for the project) including me via NeuroSky. It was pretty amazing, I had about two phone calls with them beforehand, and an afternoon on location to set everything up, and then they shot the show.
NG: So I take it you don’t really have a background in neuroscience?
SC: Nope. Although to be honest you don’t really need it. The most advanced thing you need to worry about is signal processing, and even that I haven’t had to do much of personally. The NeuroSky hardware is already giving you metrics, and even understanding the principles behind their metrics is pretty straightforward.
NG: How did Puzzlebox form? Specifically how did you and Hao Zhang get together?
SC: I originally incorporated in 2004. I was focused on rapid prototyping, and it was largely an LLC (limited liability company) wraparound of whatever project I was working on at the time. I would get contacted by designers who had an idea for some crazy project who needed some kind of expertise, and if I didn’t have it I would find people (who did).
As far as Hao and the Puzzlebox Orbit goes, at one point NeuroSky threw a project my way which was a television pitch (in which) the director wanted a brain-controlled helicopter. So of course I said ‘yes, I can built that for you’ and then of course had to figure out how. I ended up asking around, friends and people I knew, if anyone was an electrical engineer and got pointed in the end to Noisebridge. [Editors note: Noisebridge is a San Francisco based collaborative space for programming, hardware, crafts, science, food, robotics, art, and technology.] So I walked up on a monday night with the control (software) for the helicopter, but no helicopter, and started telling people what I was trying to do. People responded well, and eventually I got directed to a guy named Milo. Working with him, and third guy named Tony, in the course of five hours we’d pulled the whole thing apart and knew what was needed to do the entire hack, but it took another two weeks to actually carry it out.
On one particular night as I was hacking away and people were chatting about what was going on somebody pointed out a guy who was hacking on a wheelchair on the other side of the place. His name was Jake, and he’d managed to get someone to donate a wheelchair to Noisebridge. I stopped the helicopter thing for a couple of hours to port the software, which was now pretty advanced, to control the wheelchair. So then we had a brain controlled wheelchair, brain controlled helicopter, and wrote up a series of step by step guides for instructables to show how it works.
The wheelchair was actually controlled with an Emotive’s headset. Every once in a while I’d get a call from Emotiv saying ‘hey so-and-so wants to do a piece on the wheelchair, can we get a demonstration?’. This thing had been sitting in a hackerspace for the past two years, so it was always a ‘yeah I would love to, but I’m going to need a little notice to make sure this thing is operable’. So finally they gave me notice around July of this past year, Discovery was coming out, could I set it up? I said sure, but after talking to someone at Noisebridge I was told I had to talk to this guy named Hao, who had been hacking on this thing for the last while. That was how I met Hao Zhang, who is my partner in the Orbit project.
We spent about a month making sure the wheelchair, which he was using for his Dorabot project, was going to be brain controllable. I really got to know the guy, really respected him as an engineer, and found him very dependable. When I was pulling late nights he would stick around, we just worked together really well. As soon as that project with Discovery finished, I started talking to Hao about doing a more interesting project. All along over the previous two years people were always asking us to do brain controlled helicopters, schools, all sorts of TV programs like Nova Science Now and another episode on Discovery, so on and so forth.
It seemed like an obvious choice to do a Kickstarter project using a helicopter. Our first thought was to make it 3D printable, something that people could make entirely from scratch… There are several 3D printers in Noisebridge and in fact Milo, the guy who I did the first brain-controlled helicopter with, has his own company called Type A Machines, which we used to make our Puzzlebox Pyramids (base for the Orbit).
NG: Speaking of the Puzzlebox Orbit, were you expecting to get such an overwhelming response to your campaign on Kickstarter? You got over 700% of your originally asked for funding, that’s pretty incredible.
SC: That was really unexpected. When it started off we were shooting for around twenty thousand dollars (12,370 GBP), and were going to pitch it at that but decided you know what, lets just make it lower. We really just wanted to gauge interest, see how people would respond to this project. If they responded well it was worth our time to do this, just wanted to get a feel for things. After halving what we were originally going to ask for, (once the campaign started) we blew past it before the first week was out. Usually for the first couple days Kickstarter promotes you on the site, and you sell a few, and then sales tail off to a flat level till the end when they put you back on the front page again. Every day we waited for it to taper off, and it never did. Every day we were selling a few and I think that kept us on Kickstarter’s front page. At the very beginning two thirds of our sales came through Kickstarter. They provide you with some tools that show where your backers are coming from. By the end it was less than a third from Kickstarter because of all the media coverage that it had spawned.
It actually worked out really well because in the beginning we were able to write to individual media outlets who had already covered Kickstarter projects and say ‘hey we’ve got this thing, if you’re interested in covering it’, after that we started getting contacted by people like CNN, Wired, and some of the other bigger media outlets. The time the CNN article, which was really the biggest one for us, was posted on 6:30am on Black Friday. On that day we were just about to hit thirty thousand, and by the end of that weekend I think we had passed well beyond forty thousand. That article by itself was responsible for about an eighth of what we took in, and also spiraled into a couple of television opportunities which we’ve been busy responding to over the past couple of days. It’s really gone well.
NG: Why choose Kickstarter as a funding platform?
SC: It was the name recognition, initially, although we did also look at Indiegogo due to the Muse launching there a few days before we were ready to post ours. We’d already submitted for approval (with Kickstarter), and it took a full week for them to approve our project before it showed up, which was a surprise. Because of the proportion of responses we got through Kickstarter, especially in the beginning, it ended up being worthwhile.
Another thing, almost by coincidence, there aren’t very many technical projects on Kickstarter anymore. It’s a lot of writing a book, or producing an album, a lot of that sort of thing.
NG: Seems like when it started out there were a lot of gadgets and technical things, but not so much anymore.
SC: There have been a lot of restrictions placed. There were a few instances where people posted projects and then never produced anything, basically ran with the money. They changed the rules at one point that you can’t show a mockup of a prototype, you have to show the actual prototype, restrictions like that. Because we were within their guidelines, we ended up in their tech section on the tech page on the first page almost the entire campaign. The fact that most of the other tech projects had left ended up being in our favor.
NG: What plans do you have now for the Orbit, given the amount of extra funding that you’ve gotten? Are you thinking of taking it further, or was it intended as more of a one-off project?
SC: I don’t think it was ever intended as a one-off. We targeted the Orbit as our first consumer product, and wanted to see how far it would go based on the reaction. Because this got such good response, we got a lot of support from NeuroSky, and we’ll be presenting at their base at CES and seeking physical retailers, trying to make it a larger thing. We are pretty enthusiastic about that.
Mainly the next few months, until the Puzzlebox Pyramid comes out, we’ll just be developing the software. We’ve already published our Android app; unfortunately the iOS app still has to be approved by Apple before people can use it. We’ve contacted everyone who says they need holiday delivery on iOS device and said ‘hey by the way the app might not be approved by the time you actually give the hardware gift, if you want we’ll give you a refund’ but so far nobody has wanted to take us up on it.
The software is as basic as we can make it, because we wanted something that fulfilled everything we said it would do, but be from a programming perspective very very simple. We’re trying to encourage people to pick up the source code and be able to make modifications themselves.
Although we sold a lot of helicopters, now we want to see the users pick up the mantle and help improve the software and do new things with it.
NG: That leads to our next question: You’re releasing the hardware specifications and the software for free. Kickstarter as a platform is a very open-source mode of funding. Why is open source so important to you?
SC: The open philosophy is the point. The flying brain-controlled helicopter is interesting, but at the end of the day it’s a bit of a gimmick. However, it does have real substance behind it. The whole idea of producing a toy that you can not only take apart, but follow guides to take apart, and really understand how it works, the principles behind it, and use it as an educational toy is the entire reason for us doing this project on such a scale. For whatever reason Science in the US has taken a beating lately from political interests, and it really shouldn’t be. We don’t have to worry about that kind of stuff, and can produce something that is open and understandable, and has these interests at heart. Although we sold a lot of helicopters, now we want to see the users pick up the mantle and help improve the software and do new things with it.
NG: Are there any particular ways you are looking forward to people hacking the Orbit, or are you hoping to be surprised?
SC: I’d like to see a variety of developments. I’d love to see an artist make new graphics for the software, or an engineer port the control mechanism into a car, or another mechanism; we’re planning on providing guides on how to do that. I’d love to see some new feature that we never considered get integrated by somebody. We’re going to produce the guide material so it leads to this kind of development, leave opportunities for people to make an impact on the Orbit themselves. Especially if it’s a highschool student or a new college student, we hope that they’d be able to use it to get recognition for themselves.
Over the next few months we start with the most simple stuff, and whether we have to or users help, we’re going to get as far as we can. We want people, especially the early backers, to feel like they were part of an evolution (of the device) and part of a community, and that their desires are respected.
We were asking people who said they would be using iOS for the Orbit about the possibility that it might not be approved for the holiday, over half of them said the source code or some alternate mechanism of distribution would be sufficient. That was a huge surprise, we didn’t expect that.
Watch the Puzzlebox Orbit helicopter Kickstarter campaign video:
NG: It sounds like a lot of people are picking it up for the express purpose of hacking it as opposed to just a toy for the holidays.
SC: Yeah, exactly. We hope that the people who are technical do hack it, so that the people who are not will be able to benefit from their advances. Hopefully, people who have no idea about the whole principle behind it will start to get a firsthand exposure to what makes it work.
NG: Someone’s published a guide to hacking it to control your TV, further apps are created, that sort of thing?
SC: Exactly. In fact that’s one of the first guides we’re planning on publishing. With an infrared LED you can actually splice it into a stereo cable, plug it into a microphone cable, and use some free software like Audacity to record infrared signals. Even though it’s supposed to be an emitter, it can operate the other direction. We’re going to use Audacity as our visualization tool for waves. Whether it’s analog or digital, and explain ‘here’s a digital wave from infrared, here’s an analog wave from your brainwaves’ and if you can understand something about one, you can apply that understanding to the other.
It’s the same thing with math (behind the technology), we have a very simple explanation for what a wave is, how you measure it, cycles per second, etc. We have a way of explaining that that’s easy for a ten year old to understand. They don’t know the math component, but they can understand the concept behind it. You can chart it, and see it the result, and comprehend it instead of using rote memorization to explanation it.
At the end of the day, if you approach (the technology) from the right perspective, it’s not that hard to understand, you can use analogies that are graspable.
NG: Do you have any other upcoming projects that you’re considering?
SC: After we feel like the initial backers are satisfied, or beyond satisfied, we’ve gotten the software and everything else beyond expectations, we’ll update the PC software to support the Orbit as well; eventually we want every BCI headset out there to be supported too. We are and will stay very closely tied to NeuroSky but we want to support everyone’s efforts.
We’d really like to see sensorimotor cortex coverage in a BCI headset [Ed note: the sensorimotor area of the brain controls movement and touch sensation]. That was the thing that really got me into BCI’s, but it still costs five figures to get that kind of equipment. I think that if there is a groundswell of interest that the algorithms are out there , Bci2000 and OpenViBE and similar software, all the pieces are there. We are going to demonstrate that there is sufficient interest to build that headset. The first question people always ask us is ‘can I fly the helicopter horizontally with my brain’ and we are happy to explain that no, this is why, you need electrode coverage in this other part of your brain to do it.
So, we are going to encourage someone else to produce that headset, and we’ll support them, and if somebody doesn’t, we’ll do it ourselves. That’s our next big project, if somebody else doesn’t come out with that consumer-grade sensorimotor coverage headset then that will be our next major undertaking.
NG: The closest thing right now is the Emotiv, with fourteen electrodes. They don’t cover any of the sensorimotor area?
SC: No, they don’t cover the key places for sensorimotor control. If you look at it, there’s a lot of coverage on the frontal cortex, and then there is a massive gap right around the sensorimotor cortex, the area where some of the most interesting things are… My understanding of their design was they started off with thirty two electrodes, average EEG coverage, and while designing it worked out what information they could pull out. Then they started trying to figure out how many sensors they could remove to decrease the cost of the headset without losing any information. I think because they came at it from an engineering perspective, they missed the neuroscience aspect to it, that if you had an electrode spread to these other areas you might have been able to do other things. That said, there are still all sorts of problems with size of head and (electrode) positioning, hairstyle, all sorts of problems you have to overcome.
I think what they’ve done is a really good job with the current gear, but it’s going to require another headset to really do the interesting things that people want to see. The way I like to describe it is the NeuroSky headset is a drum. Anyone can pick up a drum, and with a bit of practice get a rhythm going. If your rhythm is right, you’ll be able to control something. The sensorimotor area is like a guitar. Anyone can pick it up and make noise, but it requires practice, and a lot of practice, to get good enough at it to control something.
Thats the sort of expectation we want to set with that sort of sensorimotor control, even with the proper coverage and the right decoding algorithms, it’s not just going to be pick it up and go.
NG: On that note, what kind of upcoming tech are you looking forward to? Or is that covered mostly by ‘sensorimotor coverage’ on a consumer BCI?
SC: You already have companies, NeuroSky especially doing frontal cortex coverage. There are a lot of people doing the attention/relaxation type of thing. Except for Emotiv, no-one else is really focused on control, and it is the hardest thing to do. I think the only way we’re going to do it is with a groundswell of support, open communities and scientists and so on. We don’t necessarily have any expertise beyond those other groups, but if we leverage a community of experts I think that’s how that problem (of sensorimotor control) is going to get solved.
Our whole position is we can give away everything, be completely open, as long as we push the technology forward. It’s like our mantra, just push the technology forward. Whoever is doing that, and has their hand on the cart, is by definition at the edge and moving forward.
There are always ways of being profitable, making money; there are all these technology companies that have to worry about patent restriction and so forth that actually impede the innovation that they are meant to encourage. We think this is an alternative way to do things. And to be perfectly honest, if you look at a technology like mobile phones, they’re kind of limitless. Look at where they started from, and how far they’ve come, and you know they’ll continue to grow and change as time goes on.
We’re interested to see how far the control of BCI’s can go, and I don’t believe we’ve seen it yet even from research labs.
NG: Thanks very much for speaking with us.
SC: For sure, my pleasure.