Brain Efficiency

I've been thinking a lot recently about how efficient the brain is. I like to spend time thinking about how neural interfaces will change the nature of humanity. Presumably, at some point, it will be possible to create computers that have intelligence that is on par with that of humans. Does that mean the end of humanity? Maybe, but maybe not.

Computers were designed to crunch numbers, and they are ruthlessly efficient at it. Unfortunately for them, most of the tasks we associate with "intelligence" are not associated with number crunching operations. Human intelligence is essentially a feat in pattern recognition - when we recognize patterns, we learn to predict the future based on previous experience. We can teach computers to perform pattern recognition tasks, but first we have to convert those tasks into number crunching operations. This is a pretty inefficient way of solving those problems, but we make up for that inefficiency by using super fast computers. Think of it as trying to drive a square peg into a round hole: its a bad idea from the start, but you might be able to make some progress if you just agree to use a humongous hammer.

So, number crunching machines are inherently inefficient at recognizing patterns. Is there another type of computing system that would be more efficient? Yes! Millions of years of evolution have placed a very efficient pattern recognition system right between your ears: your brain. Brains are insanely efficient at pattern recognition tasks. Lets see how efficient:

• The average adult consumes about 2,000 calories per day
• Of those, about 1,300 are the "resting metabolic rate" which is basically how much energy you'd burn if you just lay in bed all day and didn't move - its what you burn to keep your organs running to stay alive
• Of those, about 20%, or 260 calories, are consumed by your brain
• 260 calories in 24 hours converts to about 1.1 million joules per 86400 seconds, which reduces to 12.7 joules per second which is basically 13 watts.

That's right. 13 watts to keep the universe's most sophisticated intelligence machine operational. Astounding. By comparison, the fancypants laptop I'm using to type this blog post with consumes about 45W. The Watson computer that succeeded in playing Jeopardy reportedly uses something like 200,000W, a factor of over 15,000x more. Perhaps a more impressive feat than Watson beating Ken Jennings would have been Watson beating 15,000 Ken Jennings! And lets remember, Watson didn't 'have fun' playing Jeopardy, or parlay its experience into planning for its future: Ken did. Even super computers like Watson, with all their power, are inferior to the wonder of the human brain.

So, will a computer ever become as smart as a person? While it's hard to say, I believe that it will be damn near impossible for a computer to become as smart as a person using only 13 watts of power. I suspect that the only material that can be made to operate as efficiently as a human brain is ... a human brain. You'll never get down to 13 watts with transistors, memristors, or whatever the next great innovation is. Nothing beats neurons with respect to efficiency.

A separate question worth asking is whether a computer that can think as fast as a person (regardless of the wattage) can compete with humanity in terms of collective intelligence. I'll save that question for another day.

DARPA NESD Program

I spent the past two days at the Proposer's Day meeting for the DARPA Neural Engineering System Design (NESD) program. It was ... interesting. The program manager wants teams to create technology that can record from 1 million neurons, stimulate 100,000 neurons, and do full duplex (read and write simultaneously) from 1,000 neurons. And he wants it done in four years. And he wants this done in the context of addressing a real neuroprosthetics application such as prosthetic vision or audition. And he wants it done wirelessly. And don't forget to do your FDA IDE application, or to come up with a non-nonsensical financial model for bringing this to market. Oh, it can't be larger than 1cm^3, either. Never mind that the science of cortical stimulation for prosthetic sensory input is basically in its infancy. Or that no one can seem to work out to to keep neural electrodes viable in the brain for more than a couple of years reliably.

Phew.

On the plus side, DARPA is willing to throw up to $60M on the problem. So there's that.

My sense was that very few of the people in the room actually thought it was technically viable to do all these things in the allotted time (even though it'd still be a major accomplishment if only a subset of the desired outcomes are achieved). This sets up an interesting Catch 22: in order to be a successful proposer, you have to propose a project which you claim will meet the program's goals, even if you don't actually believe that your own goals are realistic. That only seems like a logical conundrum until you remind yourself that $60M is an insane amount of money.

To be fair, its _up to_ $60M, and that's divided out among all winning teams. And each winning team will likely have a large number of teammates in order to have a prayer of addressing all the program's requirements. So the money will have to divide down a lot. But, hey, you can divide $60M a lot of times and still have real money left.

DARPA is an interesting part of the funding ecosystem. Its pretty great that someone is willing to throw big money at over-the-horizon technology. Not all technology develop should necessarily be practical if we (the US? the world?) are to make real progress. And that's actually what bugged me most about this program. The emphasis on 'addressing a real problem', jumping through the various FDA hoops, and/or trying to figure out how any of this could be turned into an end product pretty much misses the point. This research is worth doing just because its worth doing. If there was a business case to be made for any of this stuff, some company would already be on it.

Final thought: there was a lecture on ethics this morning. The speaker brought up some interesting points: most notably about the need to deal head-on with the tin-foil-hat crowd. But the bigger point seemed lost: the time to have an ethical debate is before you start a sustained, decades-long, multi-agency research portfolio on brain interfaces. The best we can do now is to make sure we design systems that are therapeutic, safe, and secure. Discussing the bigger questions of "should we engage in this research" is largely moot at this point.

Anyways, the full DARAPA call for proposals (or Broad Agency Announcement - BAA in the DARPA parlance) can be found here.

Neural Engineering Data Consortium

I am proud to announce a new effort we are undertaking at Temple called the Neural Engineering Data Consortium. The NEDC is being founded to help bring the power of big data and competitive common evaluations to neural engineering. Efforts such as this have been very effective in organizing other data intensive research areas such as human language processing.

We have been planning this effort for a while, and we're now pleased to announce that we've raised an initial round of capital to support planning and development efforts. Specifically, we seek two main goals over the next year

1. Complete a planning exercise that will identify how the NEDC should be structured, administrated, resourced, and interfaced to the neural engineering community. A big part of this effort will be getting buy-in from different elements of the neural engineering community.
2. Develop a proof-of-concept big-data corpus comprising 20,000 clinical EEGs derived from archives at Temple University Hospital. This corpus will be released freely to the neural engineering community.

We have begun the process of explaining the NEDC to the community and soliciting their feedback and participation. Last week, I presented the NEDC to participants at the 5th International Brain Computer Interface Meeting in Monterey, California. We are also planning a presentation at the 6th International IEEE Neural Engineering Conference this fall in San Diego. Furthermore, we are hosting a one day symposium at the 2013 IEEE GlobalSIP Conference. That symposium will focus on big data and will have a mix of invited speakers and contributed papers.

Our paper from the BCI Meeting can be found here.

My slides from that same meeting can be found here.

For more information, feel free to email me directly: iobeid@temple.edu.

Conferences

One critical aspect of being active in the Brain Computer Interface community is to attend the various relevant conferences in order to present your work and stay on top of new trends and developments. These are also a great opportunity to network and to create collaborations and share ideas.

I just came back from a great conference in Monterey, California. It was the 5th International Brain Computer Interface Meeting which was held at the Asilomar Conference Ground right next to the ocean. Its a bit of an isolated meeting spot and it was a haul to get there, but there's definitely something about the quiet and seclusion that facilitates a good meeting. There are only so many conferences you can attend at a Raddison in Cincinnati before part of your soul dies.

Anyways, I thought it might be useful to present a list of all the conferences one has to attend to stay relevant in the BCI community. Here's what I came up with. Meetings highlighted in green are ones I've attended, and ones in orange are ones I've sent students to.

NIH Neural Interfaces Conference
Held on even years - this meeting used to be an annual workshop on site at NIH in Bethesda, MD until 2006. Having outgrown its original venue, the meeting now rotates locations. It ends to be single-track (all attendees in the same room at the same time) and focuses on various aspects of getting reliable functional connections to neural tissue. The most recent meetings have been:

• 2012 Salt Lake City, UT
• 2010 Long Beach, CA
• 2008 Cleveland, OH

IEEE EMBS Neural Engineering Conference

This meeting is held on odd years and covers much of the same ground as the Neural Interfaces conference. It is also typically single-track.

• 2013 San Diego, CA
• 2011 Cancun, Mexico
• 2009 Antalya, Turkey
• 2007 Kohala Coast, HI
• 2005 Washington, DC
• 2003 Capri, Italy

International Brain Computer Interface Meeting

Perhaps the most informal of all meetings, this one really seems to have a sense of getting people together to just talk and exchange ideas and vision for the field. There is no official society or backing for this group - just a bunch of interested scientists and engineers hanging out to talk shop. Its also nice because it attracts a different group of scientists than the previous two meetings, with more focus on clinical and patient applications. The meeting is so informal that they don't even have a regular schedule - last week we actually voted on whether to wait two or three years until the next meeting!

• 2013 Asilomar, CA
• 2010 Asilomar, CA
• 2005 Rensselaerville, NY
• 2002 Rensselaerville, NY
• 1999 Rensselaerville, NY

IEEE EMBS Annual Meeting

This is a massive annual biomedical engineering conference. There are multiple tracks and it covers practically every topic under the sun. Submissions are typically four page papers.

• 2013 Osaka, Japan
• 2012 San Diego, CA
• 2011 Boston, MA
• 2010 Buenos Aires, Argentina
• 2009 Minneapolis, MN
• 2008 Vancouver, Canada
• 2007 Lyon, France
• 2006 New York, NY
• 2005 Shanghai, China
• 2004 San Francisco, CA
• 2003 Cancun, Mexico
• 2002 Houston, TX
• 2001 Istanbul, Turkey
• 2000 Chicago, IL
• 1999 Atlanta, GA

Biomedical Engineering Society Annual Meeting

This is another massive meeting that attracts people from across the spectrum. It tends to be a little more engineering oriented than EMBS, which tends to trend more towards physiology. Submissions are typically 1 page papers.

• 2013 Seattle, WA
• 2012 Atlanta, GA
• 2011 Hartford, CT
• 2010 Austin, TX
• 2009 Pittsburgh, PA
• 2008 St. Louis, IL
• 2007 Los Angeles, CA
• 2006 Chicago, IL
• 2005 Baltimore, MD
• 2004 Philadelphia, PA
• 2003 Nashville, TN
• 2002 Houston, TX
• 2001 Durham, NC
• 2000 Seattle, WA
• 1999 Atlanta, GA

Society for Neuroscience

This is the big daddy conference - 40,000 attendees annually. It is a great conference and a total zoo. I've only attended two or three times I believe. Its way more about biology and physiology than is suitable for a mere engineer such as myself :)

• 2013 San Diego, CA
• 2012 New Orleans, LA
• 2011 Washington, DC
• 2010 San Diego, CA
• 2009 Chicago, IL
• 2008 Washington, DC
• 2007 San Diego, CA
• 2006 Atlanta, GA
• 2005 Washington, DC
• 2004 San Diego, CA
• 2003 New Orleans, LA
• 2002 Orlando, Florida
• 2001 San Diego, CA
• 2000 New Orleans, LA
• 1999 Miami Beach, FL

There are also a handful of ad hoc conferences that come up from time to time. Many are invitation only. Figuring out how and when and where to spend one's travel dollars is serious work!

Bionic Eyes

Looks like I might have to be eating my words on bionic vision. I've been pretty skeptical on the clinical promise of prosthetic vision systems owing to my time spent in that field during my time as a postdoc. I've always thought the concept was really cool but the technical limitations just too insurmountable to provide meaningful visual sensations. My feeling was that there will be a biological (stem cell-based) solution for blindness that will render the electrical stimulus-based methods obsolete.

That may well still happen but today I'm reading that Second Sight's bionic eye "Argus-II" has been FDA approved, which is amazing. And the patients seem to like it and value whatever level of prosthetic vision it provides, even if its minimal. Good for them! I hope this makes a big splash and a lot of money.

The really interesting question is how willing people will be to undergo the surgery in exchange for a fairly primitive visual system. My experience with blind patients during my postdoc is that they live very full complete lives and, while I'm sure they'd love to have their sense of vision back, they didn't think of themselves as missing out on part of life. Its a big grey area, but one that I'm looking forwards to seeing explored.

Double Arm Transplant - The Ultimate Prosthesis

Army Sgt. Brendan Marrocco has become only the seventh person in the world to receive a double arm transplant after losing all four of his limbs in combat. This is amazing and it really makes me appreciate that no matter how fancy the best electronics and robotics become, they are no substitute for Mother Nature. I'll be interested to see how his arm control progresses. Time to add @BMarr86 to my Twitter feed! Check out those scars! The full story is here.

Human Brain Machine Interfaces

Great article from today's New York times describing the Donoghue group doing cortical BMI in humans to control a prosthetic arm. This is the long awaited "next step" from the lab that did human cursor control a few years ago. The results are impressive and an important step forwards.

The article can be read here.

Common Evaluations

We've been having an interesting discussion at work about the best way to improve the overall rate of progress in decoding brain signals. As a community, we've been making progress, but improvements have been painfully slow. Typically, one group will announce a big breakthrough and publish it in a splashy journal, but the reality is always that "the devil is in the details" - these experiments become awfully hard to replicate since experimental details are rarely divulged. For example, if you are conducting an EEG-based BMI experiment, you might have kept the lights dim or the room temperature cool or some other detail. You might have had to reject certain trials for some reason or another - these details are crucially important to making the experiment work. And therefore, having one lab build on the success of another is damn near impossible. As a corollary, it is nearly impossible to say which lab is making the most progress or which decoding technique is the best, because there is no such thing as an apples to apples comparison in this field as of yet.

I'd like to change that.

I learned from my department chair that human language technology research had a similar problem about 30 years ago - various labs were making outrageous claims about transcribing text to speech, but since everyone was using their own proprietary data set, the claims were very hard to sort out. The solution came from the National Institute of Standards and Technology (NIST). NIST decided to institute an annual challenge to the community: transcribe these phone conversations, detect speaker language, etc. NIST provided the data sets, so that everyone was working off the same data, and finally it became possible to objectively compare the performances of various labs and algorithms. As the project grew year after year, it became necessary to set up the Linguistics Data Consortium (LDC) to design and create ever more sophisticated data sets. Part of the challenge is to properly design a data set such that all the control cases are properly addressed and that the desired algorithms can be properly tested. Once the data set is designed, LDC collects and disseminates the data. LDC will also archive and disseminate data from independent laboratories. The LDC is hosted by the University of Pennsylvania and currently has about 50 employees and has amassed about 500 data libraries in over 60 languages.

The advantages are manifold. First, by having a community-wide competition, attention and energy is focussed on the most important problems. Program managers from federal funding agencies can be instrumental in setting these goals. Secondly, by having common data libraries, the community is able to effectively ferret out the real differences between various algorithms and techniques - this is a boon for overall progress. And finally, relative to the overall amount of money being spent by funding agencies to fuel all this research, the cost of collecting and distributing the data is relatively minor by comparison.

The following graphic shows how the Common Evaluations paradigm of NIST and LDC has propelled progress in the human language technology field. As time has progressed, the challenges have become progressively harder and yet progress is always forthcoming. As a bonus, these data libraries have also become a real boon for industry players who wish to incorporate language technology into their products: these companies now have "industry-standard" data sets to build their algorithms around. So its not just good for progress in research, but also in industry.

The challenges in making such a data consortium work are also manifold. First, it won't work without the consensus of the scientific community that this is a valuable exercise. If the main labs and key players refuse to participate, then the whole exercise becomes less useful. The main way to resolve this potential problem is to (a) directly engage the community and sell them on the importance of the concept and (b) to convince program managers at funding agencies to insist that their PIs participate in the consortium. Beyond the engagement issue, there are secondary problems such as funding, scope, organization, and so on. But none of these issues are show-stoppers. We believe there is a need for an LDC-like operation in the neural engineering world, and we are pursuing efforts to start such an endeavor, to be hosted (naturally) at Temple University.