Category: Uncategorized

AI Neuroscience and AI Psychology: A New Era in Understanding Artificial Intelligence

by admin, posted in Uncategorized

As humans, when we encounter concepts or phenomena that we don’t fully understand, we create new disciplines to study and comprehend them. Emergent behaviors, in particular, often baffle us – these are instances where the whole is drastically different from its individual building blocks. We might comprehend the blocks fairly well, but the complex constructs they form can be mystifying. Economics, as an example, is a field established to decipher the collective outcomes of numerous interacting individuals.

Today, we are witnessing similar emergent behaviors in the realm of artificial intelligence. Large language models like ChatGPT are demonstrating capabilities that we can’t entirely explain, other than acknowledging that they’ve emerged from tools we thought we knew well. As AI systems become more complex and demonstrate capabilities resembling cognition, we may need to extend our methods of understanding and investigating them. In this context, it could be beneficial to consider the creation of new fields: “AI Neuroscience” and “AI Psychology.”

AI Neuroscience

AI Neuroscience would be a discipline focused on understanding the ‘mechanics’ of AI – the intricate layers of artificial neural networks, how they interact, and how they produce the outputs that they do. This field would delve into the structure and interconnections of AI models, much like how neuroscience studies the brain’s physical structure and the neural networks within it.

AI neuroscientists would investigate the detailed workings of AI models, looking into how information flows through the network, how weights and biases change during learning, and how different architectures impact the model’s behavior. They would strive to map the AI’s “connectome” and understand how various components contribute to the overall functionality.

AI Psychology

On the other hand, AI Psychology would be more concerned with the ‘behavior’ of AI – its outputs, interactions, and ‘decisions.’ Rather than focusing on the AI’s internal structure, AI psychology would look at how AI perceives its inputs, responds to different scenarios, and changes its behavior over time.

AI psychologists would develop tests and experiments to probe AI behavior, much like how psychologists use various tests to study human cognition, personality, and behavior. They would analyze how AI systems learn over time, how they generalize from past experiences, and how they respond to novel situations.

Why do we need them?

Splitting AI research into these two fields could provide a more nuanced understanding of AI systems. AI Neuroscience would help us understand what’s happening ‘under the hood’ of AI systems, while AI Psychology would give us insights into their behavior and interactions with the world.

This division mirrors the dichotomy in human cognition research, where neuroscientists study the physical brain, and psychologists study behavior and mental processes. Both perspectives are crucial for a full understanding of cognition, whether in humans or AI.

As AI systems continue to grow in complexity and importance in our lives, developing a more sophisticated understanding of how they work and how they behave becomes increasingly important. The creation of AI Neuroscience and AI Psychology could be a significant step in that direction, fostering a more nuanced understanding of AI and enabling us to use, regulate, and improve these systems more effectively.

——————-

PS from Gu: this was written by GPT 4, using my previous blog post as a prompt. I largely agreed with the points here and admired GPS4’s ability to organize language.

A Story of Unknown Unknowns

by Yu Gu, posted in Ideas, Robotics, Uncategorized

“…there are known unknowns; that is to say we know there are some things we do not know. But there are also unknown unknowns—the ones we don’t know we don’t know.” – Donald Rumsfeld.

If you work with robots, you know that there are lots of unknowns. Most roboticists, including me, make a living on solving known unknown problems. However, the robots would still get stuck from time to time. It actually doesn’t take much to surprise a robot. The trouble is, it’s hard to say what are the unknown unknowns, and it’s even hard to give examples. By definition, if we can describe it, it’s not an unknown unknown… Helping robots to deal with surprises is what I like/want/hope to do (can’t depend on it for funding though).  I need to find examples to tell people what I am working on…

Here is a personal travel story from over a year ago. My wife and I were taking the kids to China to see grandparents. There were many unexpected things happened along the way. I will let you be the judge on what were the known unknowns and what were the unknown unknowns.

Day One, Dec 24, 2019, Christmas Eve, Morgantown, Pittsburgh, Houston

This was the day to fly to China. The flight was scheduled to leave at 4:24pm. We left home at 12:59 and the weather was perfect; sunny, 50 plus degrees, with very little wind. As we drove past Washington PA, the air became foggy. It was almost like a dust storm coming at us; but it was fog. An alert appeared on the phone, said our flight to Houston (connecting to Beijing) was canceled. We called United Airlines (famous for dragging an old man off plane) while still driving towards the airport. The first lady (not the First Lady) from United was not very helpful, and it was a little difficult to understand her sometimes. She said the reason for cancellation was due to “severe weather conditions”, which sounded bogus to us at the time. She was not able to find any alternative solution (e.g., all other flights were full) and offered to refund the tickets. I told her that was not an option we would consider and asked to speak to her manager. The manager lady had many more options. She offered to check other airlines for solutions and suggested we could go through other cities such as Singapore or Hong Kong. That sounded a bit more exciting. We pulled over at the airport entrance waiting on her to find something for us. Unfortunately, the phone cut off after about 35 minutes into the call and we didn’t have a way to reach the manager lady again…

We went ahead and parked at the short-term lot and went to the United counter. The screen was now showing several canceled flights, including two flights to Houston.  While waited in the long line I dialed the United number again, in hope to connect back to the manager lady. This time it was a guy, who was quite helpful. I also gave him my number just in case. He helped us to find a flight to Houston at 9pm, which would give us just about 1-hour layover time in Houston. Sounded like a feasible option, and the best chance we had; I took the suggestion. The kids and I went to move the car from short-term to extended parking. I found out for the first time that you can drive to extended from short-term without having to pay for the latter… Looks like a loophole in the system. The fog was getting denser by then and the visibility was ~ 100m. It became clear that the weather poses a risk to flights. The road ahead of us would be full of unknowns.

Passing through the airport security was uneventful. We had dinner and started to wait at the gate. Many more flights were canceled. The fog outside looked very dense. Our airplane was supposed to come from San Francisco, but it had to land in Chicago first to wait for an opportunity to come over to Pittsburgh. The plane kept getting delayed (e.g., the phone showed that it taxied for 30-40 minutes in Chicago) and the passengers waited at the gate became more and more worried. There were about 15 Chinese, hoping to get on the same Houston flight to Beijing. As the window of opportunity getting narrower, people started to discuss the distance between C11 and D12 gates in Houston, how fast can they run, who can run the fastest, would the airplane wait for so many passengers, etc. I went to talk to the lady at the counter, she searched around and offered me a Plan B: if we missed the flight in Houston, we would then be automatically put on the next flight to San Francisco, and connect there to Beijing. I never knew they could do that (i.e., a prebooked contingency plan) and happily accepted the offer. Of course, there was another flight to San Francisco from Pittsburgh at 9pm (after many hours of delay as well) that we could get on, but I wanted to try our luck in Houston first.

The plane for the San Francisco flight came from Houston (sounds confusing, right?) and it was supposed to land at 8:22, and it did. People cheered. The United staff were also happy, known that a plane could safely land in that kind of condition. I got a brief moment to chat with the Pilots for the San Francisco flight. Apparently, to fly in that kind of weather required airplanes with special equipment and training of the pilots. They had to use autopilot during landing as there was almost no visibility for the pilot to do anything meaningful. Taking off was not so much an issue.

Our plane (from Chicago) was not so lucky. It was first predicted to land at 8:23pm but kept getting delayed. Eventually it landed at 8:56 and the scheduled take off time got delayed from 9:00 to 9:46. This would leave about 19 minutes in Houston to catch the Beijing flight. The Chinese passengers got into more vivid discussions of the possible options, but there was still a slim slice of hope. A plan was formulated: everyone needed to call Air China to hold their plane on the ground a little longer, everyone needed to tell the flight attendant to let people under time pressure to exit the plane first, the first (fastest) person who gets to the gate could tell the airline to keep the door open a little longer, among others…

Boarding was relatively fast. People patiently waited for the door to close. After the lights flashed a few times, the pilot announced that there was a maintenance related issue.  A smoke alarm needed to be reset but can not be done through software. Someone had to physically get down under the plane to check it out. Some passengers started to get impatient. The procedure took 20-30 minutes, wear out the remaining hope of catching the flight in Houston. Finally, the airplane started moving, but instead of headed for the runway, it was asked to go through a de-icing procedure. That helped to seal the deal. By this time, nobody would have believed to be able to get to Houston on time. It actually felt more relaxing this way. We wouldn’t need to run, and we had a backup plan in the pocket…

It was about 12:30am when we landed in Houston. The flight to China had long gone (12:04 am). We started to line up at the counter. The lady at United was already prepared (!). She had our boarding pass to San Francisco waited for us. For the boarding pass from San Francisco to Beijing, it was a different story. She could only print one (my son, Anderson’s) and we would need to get the remaining tickets at the gate, in San Francisco.

Day Two, Dec 25, Christmas Day, Houston, San Francisco

We had about 4 hours to spend in the San Francisco airport, but the first thing we wanted to do was to get our boarding passes. First, we went to the gate where another flight to China was boarding. However, the people there told us that they worked for United and only people from Air China could help us with the tickets. With nobody from Air China we can find in the airport, I called their number. The guy on the phone was 1. not very patient, 2. claimed this must be addressed by United since they reserved the tickets; and 3. offered to refund the tickets… He also told me that our names were in the system but there was no guarantee that we would be allowed to board the plane. We had to walk to another terminal to find the United office. The lady there suggested us to exit the airport security to talk directly to Air China at ticketing to get the boarding pass (the reason been that we were required to show them the passports, and only the Air China people could issue the tickets). So, we did that. The people at the Air China were neither patient nor helpful. They told us that 1. the plane was full and there was no room for more people (in the meantime, the lady on the phone in the background just offered tickets to four “important visitors” for the same flight…); 2. the United people did not do their job right and give us the seats; and 3. it can only be addressed by United. While listening, I was planning vacation plans in San Francisco in case we could not make to China. We were suggested a “black uncle” at the United counter as someone maybe willing to help. So we were at the United line and talking to the “black uncle” a few minutes later. Without much trouble, the nice guy replaced our tickets and told us that the Air China people would be waiting for us. We had to wait in the long Air China line, again, but finally we got our boarding passes. BTW, through this process, I learned that airline tickets and boarding passes are two different things… I also asked the lady at the counter to double check our checked baggage and she confirmed. Soon, we were on a Boeing 747 to Beijing.

Day Three, Dec 26, Beijing

This was a short day with only about 15-20 hours, and most of them were in the air. We were chasing the sun the whole time. At first the sun was faster, and finally set behind the horizon. As we got closer to the North pole, the airplane was able to gain some ground.

The kids were surprised that we had to go through the Custom (they were a little tired). The line was not too long, and the process was smooth. Finding our luggage was not so easy though. There were just a few bags left and none of them was ours. Once again, we were at the customer service and they couldn’t locate the bags in the computer system. A big problem was that all our heavy winter Jackets were in the checked bags and it’s freezing outside. Another problem was that we would be heading to Baotou (a very cold city in northern China) in a day so it was not clear where should the bags be delivered. The nice lady there gave each of us a red blanket, so we wouldn’t be too cold waiting for the Taxi outside. That made the four of us looked like Tibetan monks in the Beijing Streets.

Day Four-Seventeen, Dec 27-Jan 09, 2020, Beijing, Baotou

Many interesting things happened but I am going to skip this part.

About the bags, they were able to find them in Houston and delivered them to Baotou directly. Before that, we borrowed clothes from relatives. We then had to drag four large bags back from Baotou to Beijing on the train.

On Jan 7, there were a few seconds on the TV news about a new virus found in Wuhan that made some people sick.

Day Eighteen, Jan 10, Beijing

It was the time to think about going home. I checked the tickets online and only Anderson’s was shown. I was not surprised. I picked up the phone to call Air China and was told to “talk to United”. Talked to the United and was assured “everything is solved, just not showing up on the website.”

Day Nineteen, Jan 11, Beijing, Washington DC, Pittsburgh, Morgantown

It was going to be a long day with over 30 hours. We arrived at the Beijing airport in the morning. Still no tickets can be found in the system… Expected to pick a fight, I walked to the Air China custom service. A nice lady there was happy to help. She even walked out her station to help us to get the bags checked…

We got to DC no trouble, then we jumped on a wrong airport shuttle bus. It was until everyone else got their bags for us to find out that our bags were at a different part of the terminal… That wasted a few hours but we did barely made to the flight to Pittsburgh that night.

After Came Back Home

A day after we came back, my wife had to fly to Dallas to help her sister with the birth of twin babies (a known unknown). No one was thinking much about the coronavirus in the US for the first week. People heard about it in the news, but that seemed to be a distant thing in China. At the end of the second week, people (mostly Asians) started to be cautious. Things went progressively (exponentially) bad after that…

Some Random Thoughts

Unlike robots, we rarely get totally stuck in normal lives, probably because we have accumulated a lot of different experiences growing up. But that doesn’t mean we can’t get stuck. Sometimes it just takes a couple steps out of our normal routines to find such examples.

Some examples of us getting stuck are probably taking a math exam or trying to solve problems (e.g., when doing research). I also feel that we sometime get stuck when facing a system built on procedures (you cannot move to step B if step A was not completed).  Is this a problem of our decision-making process or how the system was designed?

Some profound events that would have a huge impact on our life are happening somewhere (far away, close to us, or in plain sight) when we are busy worrying about other things…

 

Last Quarter Moon

by Yu Gu, posted in Astronomy, Uncategorized

This is an 9-panel mosaic of the last quarter moon. The seeing was excellent at the beginning and it deteriorates a bit towards the end. My C11 also has some field curvature and collimation imperfection. But I think the end result looks quite good. I also added a Mars image from the same night (~8hours earlier) for comparison of their visual sizes.

See a much larger version of this photo on my Astrobin page.

Or click here for a full resolution (about 300 meters per pixel) version (make sure to zoom in!).

Starting a Deep Sky Gallery

by Yu Gu, posted in Astronomy, Uncategorized

After NEOWISE, I  am getting more interested in learning deep sky astrophotography.

With a recently acquired Astrophysics 92mm F6.65 Stowaway and a 0.8x field reducer/corrector, I really don’t have excuses.

Here are the first two images taken from my light-polluted backyard that I am quite happy about: M27, Dumbbell planetary nebula, and M13, the Hercules globular cluster.

The camera used was an unmodified Sony A7RII, no filter, 100 x 30s exposures at ISO 800 (to avoid the star-eating problem…).

Comet NEOWISE

by Yu Gu, posted in Astronomy, Uncategorized

In almost 45 orbits around the sun, I have only gotten to see a handful of comets.

Halley’s Comet’s return in 1986 was the single event that ignited my interest in astronomy. I was 11 back then. All of a sudden, everyone in school was talking about Halley. All the kids wanted to see it, and I wanted to see it too! The “instrument” I had at the time was a pair of toy 2X binoculars. I begged my parents to buy me a real telescope, but they could not afford one. One day, my dad borrowed a telescope just to check it out, but it was too expensive. It was about 50 Yuan (or 10 USD), roughly my parents’ salaries for a month.  I used my 2X binoculars to scan the sky every evening and I did not find anything. Nobody I knew seemed to have seen Halley, except for some rumors floating around. But we were just kids. We knew that if we take care of ourselves, we may get to see it in year 2061!

The interest in astronomy and telescopes stuck with me ever since. Years later, I have gotten to know several people sharing the same story: Halley helped them to discover the passion that they didn’t know existed, despite that they never got to see it.

Fast forward four years (1990), I was in high school. My dad’s best friend gave him a pair of 10×50 binoculars, after he visited America. I started to scan the night sky from my balcony anytime I can. One night, I came across a fuzzy cotton ball near zenith, which seemed strange. The next night, it was still there, but moved a little. Maybe I found a comet? I wrote a letter to Beijing planetarium about this finding. To my surprise, I received a reply a few weeks later. What I saw was a comet discovered by Levy (a famous comet hunter) a couple months earlier. That was enough excitement to keep me up at night! I became more diligent at scanning for comets; but with no success.

Five years later (1995), I was a college student in Shanghai. School load and light pollution had made astronomy a distant memory. That was until Hale–Bopp showed up, one of the great comets of the 20th century. It was supposed to be not only visible, but also spectacular to the naked eyes! But not for people in Shanghai. If you have sharp eyesight and patience, you maybe able to find 5-10 stars in the Shanghai night sky in those days. I had a homemade 3cm scope with me and I found Hale–Bopp from the top of a tall school building: a faint glow in the orange sky!

Another four years had passed before I arrived in the US (1999). One day, about one year later, I realized that I could actually afford a decent telescope, with my humble graduate research assistantship. I went all crazy on telescopes and planetary imaging. I have seen a couple comets over the years but nothing too exciting. Oh, I also saw the 2009 Wesley Jupiter impact event. Not sure if it was caused by a comet, but close enough to the famous comet Shoemaker–Levy 9 impacts that I missed.

Now, year 2020, comet NEOWISE is here to visit the sun. My son is now at the age when I was helplessly hunting for Halley in the sky and my daughter is only 2-years behind.  The whole family got up early one day to see NEOWISE rising before the sun. Now the comet is in the night sky, we can observe it from the comfort of our front porch. I have never taken a photo of a comet before NEOWISE so it’s an excellent opportunity to practice!

An NSF REU Proposal on Human-Swarm Interaction

by Yu Gu, posted in Research, Robotics, Students, Uncategorized

I am sharing a REU (Research Experience for Undergraduates) proposal. When I first started writing this proposal, along with my colleagues Dr. Gross and Dr. Klink, we had no idea where to start. I searched the internet and found a couple sample REU proposals. Several friends also kindly shared their successful proposals. Now, it’s my turn to give back!

The project is about human-swarm interaction: how can one person effectively manage a 50-robot swarm in achieving a high-level goal? In the project, the undergraduate students need to design the robots, the test environment, a simulator, human-machine interfaces, along with swarm control algorithms. They would also perform a variety of experiments to demonstrate these capabilities.

During the first year (2019) program, eight undergraduate students from around the country worked together with several WVU students for 10 weeks. They designed and built 50 robots (!) and developed the basic swarm control software. Here is a video:

Here is a paper written by the students. One major difference between our REU Site and most other REU programs is that all students were working together on a same project. They had to work as a team (I called them a swarm…) to achieve the overall project goal.

This year (2020), however, we had to cancel the program due to COVID. We will be back in motion again next year!

The proposal received a C (Competitive) from the NSF review panel. The individual reviewer ratings were VG/G, VG, VG, VG/G (VG – very good, G – good). In general, the reviewers were excited about the project research ideas, but had concerns about our detailed program designs. Of course, we were very inexperienced in this area at the time.

Attachment: REU Proposal

The CV that Gets You the Faculty Interviews

by Yu Gu, posted in Faculty, Job, Robotics, Students, Uncategorized

If you are a Ph.D. student or a post-doc, you probably have thought about the prospect of becoming a professor at some point, even though this option may not be high on your list. If you only thought about getting a faculty position shortly before applying, you probably won’t get an interview; because it takes planning and time to build up a CV that is competitive for the faculty job market. When is the best time to start planning your academic CV? During your undergraduate years. Ok, if you missed that, and I don’t think I know anyone who didn’t, the earlier the better. I only started caring about my CV three years after my Ph.D., and I had to pay for that …

A strong CV is simply the most important part of your faculty job application package, and sometimes, it may be the only thing that a search committee member reads.

Sun Tzu once said: “If you know the enemy and know yourself, you need not fear the result of a hundred battles.”

There would probably be more than 100 battles for a faculty-job applicant. Let’s first think about your opponents: the search committee members. We are talking about a bunch of professors who may or may not be in your field. They have all failed numerous times in the past but those memories have faded. They are now trying to get through busy daily schedules and fulfill the service duty with the minimum time/energy costs. Reading tens, if not hundreds, of applications is just not that fun for them anymore. To prevent these people from making arbitrary decisions there is likely some kind of forms that the search committee members must fill. These are probably pre-determined performance metrics that evaluate your record or potential on teaching, research, and service. What does all this mean? It means your CV needs to be simple (easy to find relevant information) and well rounded. You shouldn’t be lagging your competitors in any of the major categories (e.g., publication, teaching, funding, and service). An easy way to find out the going credentials is to check out the CVs of newly hired faculty members in your field; they are not hard to find on the internet. Scoring a zero in any of these categories is not going to look good for you.

You might wonder how would you get teaching and funding experiences when you are still a graduate student? You can if you try hard enough. For example, you can offer to teach or co-teach a class; you can submit fellowship proposals; and you can attend teaching and grant writing workshops. A reasonable search committee would consider your available opportunities as a student, but you must demonstrate that you are proactive and have the potential. This would also help you later during the interview process.

Would this be enough? Not yet. Having all the boxes checked may get you a decent first impression. It may also take away some of the arguments for someone on the committee to strongly against you. What you really need now is a few strong supporters on the committee. In other words, you must impress them.

What makes your CV look impressive? Here is my rule of thumb: the most impressive things are the ones many people have tried but could not get. If you have done something unique but no one on the search committee has thought about trying, then sadly, it’s likely to go unnoticed. What are some examples of impressive things? publications in highly selective journals/conferences, best paper awards, prestigious fellowships/grants, a new theory, solving a known hard problem, …. Of course, you need to be good, work hard, and have abundance of luck, to get even one of these. Many people focus energy on achieving the seemingly least uncertain goal: increase the number of high-quality publications. However, if you have special talents, you may find a different path to success.

I want to quickly bring up the discussion of quality vs quantity here. I have seen CVs with over 100 papers not being appreciated by the search committee. A CV with just 1-2 good papers (unless they are on Nature or Science…) may also not be convincing enough. What if this person was just lucky or was working in a very small niche area? In robotics, for a freshly graduated Ph.D., I think 2-3 first-authored top journal papers and 2-3 first-authored top conference papers, plus a similar number of non-first-authored high-quality papers would make your CV look quite impressive. It’s certainly not easily achievable within the Ph.D. period. Adding more low-quality papers to the mix would only serve to reduce the perceived quality of your CV.

OK, say you have gotten a decent impression for your application package and the committee will meet next Wednesday to decide on a list of candidates to be interviewed. By the time the committee meets, the contents in many applications may have gotten mixed up in their heads, because they all look so much alike… The only chance you have is if you got something stands out, pointed out by one of your advocates on the committee. Other members would quickly flip to that page on your CV. If they also agree and cannot remember anything bad to say about you, then congratulations, most likely you would get an interview.

So, let me summarize it: create a CV that maximize the “expected” number of supporters (e.g., having something impressive) and minimize the number of naysayers (e.g., being well-rounded).

How to get there? This is the part about “knowing yourself.” Discover your true passions and strengths and come up with a plan for yourself. Well, …, that’s easy to say than done. Without known better, one way that worked well for me was to start by making myself look bad: creating the Google Scholar profile and making it public; posting the CV online; asking myself awkward questions like what are my top 3 contributions to the field? I found things usually get better after I learned to not avoid myself.

Attachment: my CV back in 2008, over 3 years after my Ph.D., which, unsurprisingly, did not get me anywhere.

Shake Your Camera to Take Sharper Photos

by Yu Gu, posted in Future, Ideas, Uncategorized

Computational photography is changing the way how photos are taken. More and more cell phones are using computation to offset the small lenses allowed on the phones and the progress has been amazing. What I am still waiting on is a way to allow shaky cameras to take sharper photos. Arguably, a shaky camera can provide more information of a scene than a steady camera. It seems like our brain-eye (and inertial?) system can process it, which gives us a stable (and sharp!) perception of the world while moving. Since most phone users are not so good at holding their cameras steady, why not taking advantage of the shaking? Even better would be allowing a shaky telescope to provide a sharper view of the planets! Has this been done before? Can someone point me to a product or a paper using this approach?

So, What Do We Do with It?

by Yu Gu, posted in Astronomy, Photography, Research, Robotics, Uncategorized

I am more of a telescope collector than a sky watcher. I have about a dozen telescopes of different designs: achromatic, apochromatic, doublet refractors, triplets, Newtonians, Maksutov-Cassegrain, Maksutov Newtonian, H-alpha solar scope, Dobsonian mounts, German equatorial mounts, roof prism binoculars, porro prism binoculars… you name it. I know way more about telescope designs than constellations of the sky or features on Mars. Most of my telescopes spend years collecting photons in a very dark place: my closet.

I am more of a camera lover than a photophile. I have several cameras from the film era to the mirrorless age. I have a couple dozen lenses with focal lengths ranging from 14mm-500mm, not counting telescopes.

I have learned to accept this. There is nothing wrong with being obsessed with equipment, I told myself, the hobby is supposed to be fun!

I also like robots. My lab, IRL, has about two dozen robots, plus a 50-robot swarm. The UAV lab I worked in before had about a dozen UAVs. Most (but not all) of these robots and UAVs were custom developed. I, as someone who always like toys, had a hand in the design of most of these systems.

So now, what happens when we have all the hardware we ever wanted? Of course, we can only get close, but not there. There is a pretty big difference between “wants” and “needs”, and we often rationalize “wants” as “needs”. As engineers and perfectionists, seeing small issues with the current setup makes us feel itching. We are constantly dreaming up next design iterations. We are telling ourselves better robots will make our research better.

Is that true? Do we really need more/better robots to do better research? Maybe to some degree. If we don’t have the appropriate tools, we can’t do certain experiments. If we don’t have high quality equipment, some work may be very hard to do (e.g., mapping without 3D Lidar or robotic pollination without a precision manipulator). I think another important reason for having the best robots, like having the best telescopes/cameras, is that we have no one else but ourselves to blame for the underperformance…

So, let me ask again, what happens when we have all the hardware we ever wanted? What do we do with it? The answer is simple: let’s focus on research. Instead of rushing to start on the next generation design and letting the existing robots collect dusts, let’s make them do things nobody else can dream of or believe!

What Makes a Good Grand Challenge?

by Yu Gu, posted in Ideas, Robotics, Uncategorized

I am a big fan of Grand Challenges.

I was super motivated when reading about John Harrison, a self-taught engineer (carpenter) in the 18th century, who won the longitude reward through decades of perfecting clockmaking skills.  I also watched several DARPA Challenges with great interests. I have participated in, for three years, NASA’s Centennial Challenge on Sample Return Robot. Those three years left me with countless memorable moments to be enjoyed for the rest of my life.

What I like most about Grand Challenges is that they give people excitement and hope. Grand Challenges allow someone, who otherwise would not be known by people, such as John Harrison and Charles Lindbergh, to shine through their courage, dedication, and talent. They also can accelerate technology development, by bringing together a broader range of conventional and unconventional innovators and solutions.

However, many Grand Challenges failed to achieve these effects, for a variety of reasons. Here are a few factors I think maybe worth considering when designing a new Challenge.

  1. It needs to be relevant. If a Challenge addresses one of humanity’s most urgent needs, more people would likely to follow and participate.
  2. It must be a Challenge. A Grand Challenge needs to be hard. It should be a jump from any of our known abilities. It may sound impossible at first, but It’s so cool that it makes people imagine. The Challenge shall also not be too big a jump, otherwise everyone would fail (which is an acceptable but not desirable outcome).
  3. The Challenge description must be clear, rigorous, and stable. Like any games, there should be no ambiguity and room for interpretation. The actual tests must also precisely match the description. Unfortunately, quite often, the organizers did not fully think through all the issues at the beginning. They would come up with a set of rules that cause confusions (and potentially unfairness) and then they dumb down the challenge after most participants failed (this happens more often than you may want to believe!).
  4. Human factors must be kept at a minimum. One of the Grand Challenge’s greatest strengths is that it gives everyone a fair chance. You do not have to be a world renown thinker/scientist/engineer, you do not have to be rich, you do not even need to have a stable job; as long as you have a good idea, the skills, and the will (easy to say than done), you have your fair chance of winning. The success of a Grand Challenge should be defined by beating the problem, not anyone or anything else. If we allowed humans (e.g., the Challenge organizers) to pick winners based on their pre-conceived ways of solving the problem, John Harrison would had no chance against big name astronomers at the time (note: the Longitude Board, including Newton’s preference on finding an astronomy-based solution did cause hardship to Harrison for many years…). Let the results speak for themselves!
  5. Teams shall come up with their own resources, at least initially. This one may sound strange to you. Would it not be rewarding people with deeper pockets and leave the poor guys out of the fight? It might, but let’s consider the alternatives for a moment. What if the organizer picks a few promising teams, give each of them a few $M, so they don’t have to be sidetracked by fund raising and other resource constraints?  The question would then be: based on what criteria? prestige? track-record? how likely a team’s idea may work? If you read the history of Grand Challenges, you would know that none of these are reliable indicators of success. What this funding approach does is effective disincentivize the selected teams to push envelopes hard (they already have the cake; the final Challenge prize is just the icing) and block out all other competitors. In my opinion, just like any startups, each team needs to fight for its own survival the entire time. If you think you have a good idea, try to convince someone to fund you, or join another team with adequate resources. I think the phased approach being used by NASA Centennial Challenges is very good. Let teams compete for some initial phases (e.g., a simplified Challenge with a low-entry barrier) on their own dime, provide teams some funds once passed the initial phase. This record of success also helps teams to raise more funds from other sources.
  6. Give it a longer time frame. Most government funding mechanism have a short time horizon, but that is not necessarily good for getting the best outcomes. If a problem is of such importance to the society (e.g., determining longitude), why not keep the challenge alive for decades until it’s solved (luckily, it was!)? Short term focus leads to more applied solutions, discourages risky/crazy ideas, and more likely leads to the picking of lower-hanging fruits. Grand challenges for picking lower-hanging fruits? Does not sound good!
  7. Follow up after the challenge. Don’t let the whole thing ends the moment a victory is declared. Each participant probably has developed something unique/valuable; creating mechanisms (with funding) to support them working together for a little while may spark more innovations.

Of course, we all live within the real-world constraints. I will continue to be excited whenever a new Challenge is announced!