Dr. Guilherme DeSouza

"I can hear machinery / blinding me with science science science"

     Thomas Dolby, "She Blinded Me with Science", 1983.

Having just returned from 6 days at the IEEE Symposium Series on Computational Intelligence (SSCI) in Hawaii, Dr. Gui DeSouza was apologetic when we first got on the phone last year.

"I'm sorry for all this back and forth," he said, which was ridiculous; while he's co-chairing a symposium on Computational Intelligence in Robotic Rehabilitation and Assistive Technologies, he volunteers his time to answer questions from a complete stranger, and he's the one who's sorry for missing a call? That's quite alright, Gui. Thank you for being one of my most engaging interviewees while juggling multiple professional responsibilities.

 

Dr. DeSouza is an associate professor in the Electrical and Computer Engineering

Department at the University of Missouri where he's established the Vision-Guided and Intelligent Robotics Lab. Here, he offers his thoughts on artificial intelligence, how to navigate the increasingly competitive world of academic research, and the societal improvements robotics is enabling.

The following interview took place on November 20 and December 8, 2017. 

 

INTERVIEWER

Where'd you grow up?

 

DESOUZA

Rio de Janeiro, Brazil. 

 

INTERVIEWER

Who or what were your early inspirations as a young scientist?

 

DESOUZA

Definitely science fiction movies, everything from Star Trek (Mr. Spock) to many of the novels by Isaac Asimov.

 

INTERVIEWER

What is something today’s graduate students should be jealous of from your graduate school experience? In what ways are you jealous of current students?

 

DESOUZA

Growing up during the information revolution allowed me to witness each big tech announcement from the Atari and Apple II computers to today's touch-pads, cell phones, highly advanced CPUs and computing farms.  From 8-inch floppy disks (those were really “floppy”) with 180 KB (that's right, kilobytes!!!) to today’s high density SD cards and solid-state HDs. Some of the students today can’t appreciate how far we've come and how we got here!

Those students will certainly live longer and witness even more advances than I

can expect to. Unless, of course, I can develop a robot to which I can transfer my consciousness to – well in line with what Mr. Spock and others would do in the science-fiction movies I enjoyed as a child.

INTERVIEWER

You established the ViGIR (Vision-Guided and Intelligent Robotics) Lab at the University of Missouri after starting it at the University of Western Australia. Tell us a little about its origin and mission.

 

DESOUZA

The ViGIR Lab started after my stay as a Research Professor at Purdue University.  Back then, we had a project with Ford Foundation to create 21st century assembly lines. The idea was to automate all assembly cells using 3D computer vision algorithms and robotic platforms that could perform “on-the-fly” assembly – i.e. without stopping the assembly line. Those tasks proved to be tedious and very application-dependent. So, the ViGIR Lab was born with the idea of “teaching-by-showing” i.e. humans would perform the assembly tasks in front of the robots, while cameras would capture the movements and the systems would ‘transfer’ those movements to robotic counterparts to perform the same task.  So, in 2003 at UWA and continuing in 2005 at Mizzou, we started to develop 3-D robotic vision algorithms that could be used to guide robots in many activities, from navigation to assembly, from modeling the environment to making inferences about that environment.

As I tell my students: we must be always prepared for new opportunities and

doors that open to us in life. So, making a long story short and fast forwarding to today, we received grants that changed our original goal, but the expertise that we built over the years allowed us to use those same 3-D algorithms to create robots, not to automate assembly lines, but to automate field phenotyping, create 3D models of the human arm for a diagnostic of lymphedema, to develop systems for robotic assistive technology and the like.

INTERVIEWER

Speaking of, can you explain field phenotyping?  

 

DESOUZA

Phenotyping is basically trying to establish the correlation between a functional characteristic of the plant, any physical property of the plant, and its relation to the plant's functionality. For example, in corn, if the leaves are hanging down, that means it's stressed. So, we're taking a look at the leaf angle and correlating that to the functional response of the plant. One we establish that, the next step is determining how to produce plants that resist those stressors. That, of course, involves field genotyping - function to gene. If you have a plant that's resistant to drought, you can figure out the genes involved, try to enhance that gene's effect, crossbreed the plant to make sure you're producing a crop that has the gene.

INTERVIEWER

Interdisciplinary science initiatives are being assembled to probe the more elusive mechanisms of cancer and the deeper mysteries of consciousness and the brain. Computer science seems to live at the nexus of these projects. Your concentration is listed as bioinformatics and health informatics on the university faculty page. How has your research changed as the applications of computer science have grown?

 

DESOUZA

Yes, it most definitely does. As I mentioned above, we must be smart about global and society needs. Unfortunately (or fortunately) that is where the research grant money goes, so you have to adapt your goals. That is very different from “selling your soul” for the research that is exactly the same, the challenges are exactly the same - all that changes is the application.

 

INTERVIEWER

So, you’ve built a corn watchdog unit comprised of Vinoculer, a solar-powered tower outfitted with 3-D cameras that maintain a 360-degree view on a surrounding crop and communicates with Vinobot (pictured above), a robotic vehicle on the ground that’s dispatched when the tower detects a distressed plant. What exactly are you learning from the system, and is it a prototype for something bigger you’re envisioning?

 

DESOUZA

What could be bigger than feeding the population of the world?

We are learning how to plant more in areas where corn couldn’t be planted

before. We are learning how sorghum, maze, soybean and more are responding to stresses and adverse weather conditions. And maybe even more importantly, how we can plant more densely while keeping or improving yields. So, again, we are learning how to feed an increasing population under worse conditions than we've ever had. I hope that is big enough!  

 

INTERVIEWER

What are the advantages of a system like yours compared to something like drone surveillance?

 

DESOUZA

Vinoculer is a fraction of the initial cost for a drone plus there's the maintenance cost of a drone to consider. It doesn’t require a pilot or any human assistance and it works 24/7 while still carrying a much larger payload. We can study how plants recover from daytime stresses, such as heat, during colder, less stressful nights. Then, we can find, for example, families of corn that would better adapt to such conditions and strive. During the 2017 eclipse for example, we set up Vinoculer to collect data and study whether plants are “fooled” as many animals are by the sudden “end of the day”. A 24/7 platform opens the door to these new questions and many, many more that we haven’t asked yet.

 

INTERVIEWER

From Tech Crunch to Wired, the publication of your phenotyping paper and Vinobot achieved a great deal of press. Do you pay much attention to scientific journalism or how the major publications (New York Times, Wall Street Journal, NPR) report scientific findings?  

 

DESOUZA

As a scientist, I tend to focus more on scientific journals, but it is always great to see your work recognized by other, not-as-technical venues. As long as they don’t post fake news about our work...

 

INTERVIEWER

Most climate change conversation is focused on slowing our global temperature increase and capturing or eliminating excess carbon dioxide, but less attention is paid to adaptive efforts that provide manageable solutions in the interim. Food scarcity is already a real issue for areas like Malaysia, a country that depends heavily on corn and rice crops that is situated in regions vulnerable to global warming. Before working on this project, was climate change near and dear to you? How has the project affected your view of the issue? 

 

DESOUZA

Unfortunately, climate change is still a controversial topic. While I believe in science, some mistakes made by scientists and climate change advocates in their alarming conclusions, especially in the early stages of climate change studies, didn’t help the cause and alienated most of those who today deny its existence. So, to avoid alienating further, I try to focus on the things everybody agrees: things are changing, whether they believe it is manmade or not, and whether they agree on the solution to prevent even more catastrophic consequences, we must be prepared to feed ourselves. That, nobody can deny!

 

INTERVIEWER

Staying on the topic, but moving it to the home front, agriculture is Missouri’s top economic industry. What has the response been from the Missourian community at large regarding Vinoculer/Vinobot?

 

DESOUZA

We have had a few local events, mostly promoted by CAFNR (University of Missouri - College of Agriculture, Food, and Natural Resources), as for example at the recent State Fair. It is interesting to see old folks, farmers, people who grew up working hard in the fields react, but when they see the robots, they don’t feel threatened and worried about robots replacing humans.  They know how hard their jobs are/were and they welcome technology that can make it easier for them to stay in business.

 

INTERVIEWER

Plenty of scientific luminaries (Elon Musk, Bill Gates) warn against the dangers of AI. What does the public not understand about AI that should have them concerned? Alternatively, what does the public not understand about the awesome potential of it?

 

DESOUZA

During the agricultural and the industrial revolutions people were afraid too. Especially during the industrial revolution, people were most concerned with machines replacing them. Factories producing more without humans to perform the manual labor they grew up performing and teaching others to perform. We may have lost and still miss some of the hand-made watches, furniture, food, etc.  But the industrial revolution brought those products to an immensely larger number of people and the world didn’t end or remotely resemble any of the doomsday predictions.

In fact, I am a LOT more concerned with the "natural intelligences" that have

already created so many atrocities throughout our history than the artificial one that has done none of that yet. “They” can make AI dangerous, as they can make anything dangerous.

The public should know, if they don’t yet, that AI is around us already.  It is

filtering what you can see on Facebook.  It can find the face of your friends in our camera's pictures. It can drive cars. It can interpret your commands to Alexa and Google Home. And again, what makes AI dangerous is the person who programs it to make or do bad things. They can program AI to influence elections, to attack other countries with DoS systems or intelligent viruses that sabotage industrial processes thousands of miles away, and more to come.

What the public is afraid of is not AI. They are afraid of AI becoming the natural

intelligence. They are afraid of AI becoming self-aware and deciding to protect us from ourselves. But fortunately or unfortunately, that is something that I don’t believe will happen any soon.

INTERVIEWER

Following the 2016 election, the idea of “scientists-turned-politicians” was reignited as some of Trump’s campaigning promised cuts to funding agencies. If you had to nominate one living scientist to run for public office, who would it be and why?

 

DESOUZA

I think we, scientists, should leave politics to the politicians. We can and should provide advice, consultation, and support from scientific findings to guide their decisions. But they are better coined to navigate the more deceiving and tortuous means of doing politics. We are too naïve for that.

For example, I have seen brilliant minds, like Noam Chomsky, who is capable of

making brilliant analyses of our society and the problems it presents as a constant, cyclic battle between those who control money and those who control public opinion. And at the end, he concludes by defending the control of the means of production by the state. Hasn’t he lived through the 50’s, 60’s… Wasn’t he here in the 80’s to see these models miserably fail?

Again, scientists are intrinsically naïve. We believe reason will prevail and those

who disagree with us will face the inevitable truth that science provides and everything will work out (this time). But we forget one simple fact: we are humans too, and our “science” is flawed, especially when we let the worse side of our human nature drive and bias our conclusions. That is not science: that is faith, and a bad one. 

INTERVIEWER

Plenty has been written about the challenges research scientists face as funding for their projects disappear and competition swells. Has it influenced your approach to grant writing. If so, how?

 

DESOUZA

Oh, yeah! Besides what I have said already above, there was a time when I was focusing on defense projects. Sometimes we call it “dirty money.”  I will stop here. But I will tell you that I am happy I now focus on helping people before and after wars, instead of during wars.

INTERVIEWER

It seems that, currently, it’s all doom and gloom for prospective and current PhD students. Most of what’s written on the current state of academia as a potential career for newcomers is that there are too many qualified candidates for too few tenure-track positions, and it’s exceedingly difficult to achieve a work-life balance as the number of grant applications increases while budgets shrink, not to mention how the Trump administration’s tax plan and its tuition exemption repeal would pose a serious financial challenge to all incoming students. What advice do you give someone considering the PhD track?

 

DESOUZA

It's true what you said. But I don't think we're in a time that's too alarming or completely hopeless. The department here, for example, is in the process of hiring 40 new faculty members - that started last year and will probably continue through the next. 

Robotics has become a very hot area. An ongoing trend is the accumulation of

data, so people who know how to mine that information and design new systems capable of handling big data. Sometimes, you have to move areas. Like I said, earlier in my career, I was working on more military-related applications. Now I'm working on robotic phenotyping and assistive technology. It's about figuring out what you can do based on what you know and what you like to do.

INTERVIEWER

That's a good way to put it.

 

DESOUZA

And one more thing. A few years ago, being a computer scientist or being involved with AI, people would look and say, what is that? Now, you say you're doing computational intelligence and people perk up. It's in our daily lives - it's already here. Algorithms have gotten better and better, but "fake news" poses an unprecedented challenge. So now more than ever, Facebook and Google need people who can design algorithms that can distinguish between real and bunk. It's an exciting time for young people to be working in the area.

INTERVIEWER

I've long thought that research science and the arts run parallel paths – both careers mired in failure (failed experiments, rejections from publishers) that require obstinate will power and a creative approach.

 

DESOUZA

Rejection is certainly a constant in our lives. If you let rejection affect you, you end up depressed or worse. But failure is a drive. It is what makes us keep going. You just have to focus on the challenge of being right and proving to yourself that you are right. Sometimes that doesn’t necessarily, let alone quickly, translate into grants. But then again, remember that I said we should be afraid of the natural intelligences first and foremost.

 

INTERVIEWER

What are your favorite books? Favorite books on science? They certainly don’t have to be mutually exclusive, but I guess I’m trying to tease out fiction versus books on science, which are generally more historical non-fiction.

DESOUZA

I don’t like to read novels or fictional works anymore – not even science fiction novels. I don’t like to say that out loud – it makes me feel old, kind of like listening to my grandpa when he said similar things – but those novels are all the same.  It is like reading them over and over again.

I now prefer books like Weapons of Math Destruction, by Cathy O’Neil, which I

have just started. I was talking to a colleague about the book, and one of the points of the book is the big data movement and its associated technologies are not democratic. They're going to increase the gap between the wealthy and the poor, but I don't know if I agree with that. I haven't made it to that argument, so I won't totally disagree, but a lot of the technologies of the last couple decades, like the internet, have brought third world countries closer to us. They now have access to things that were in the past much harder to get due to cost and other factors. 

INTERVIEWER

What has been or will be the biggest breakthrough in electrical engineering and computer science in your lifetime? 

 

DESOUZA

That is a hard one. So, I will use a politician tactic: I will answer something related, while avoiding your actual question.

I joke with my students, especially in the FIG (Freshmen Interest Groups) by

challenging them to find one single object and profession around them that hasn’t been influenced and made a breakthrough (now just to connect to your question) because of computer science and computer engineering. Cellphones, computers, tables, etc are the obvious and the most ubiquitous answer. But think of cars, designed by mechanical engineers, but made thousand times more interesting since the use of computer technology inside and outside the hoods. You don’t like other engineerings!  Take medicine! Every single recent (20-50 years) advance in medicine was done by use of computer technology: from lasik surgery to MRIs, CAT scans. From genetic engineering to new medications that were only possible with simulation tools and other computer models to predict its effects and delivery.  The list goes on and on and it doesn’t let alone any of the arts, TVs, movies, music… not even journalism.

INTERVIEWER

Last, but not least: Beatles, Stones, Zeppelin or Floyd?

 

DESOUZA

Neither. Queen!

 

Learn more about Dr. DeSouza and his lab here, and follow this link to see the latest news from ViGIR.  

Feel free to leave your thoughts on the interview in the comments section.

©2019 by EG Lund.