By Tim Ventura | 8 February 2020
Learn how artificial intelligence, humanoid robots, machine creativity and the singularity will conquer space & maybe even save our own planet. We’re joined by Dennis Bushnell, Chief Scientist at the NASA Langley Research Center to discuss robotics, AI, and sustainable, commercially viable solutions for the Earth, Mars & beyond…
Dennis, we’re going to be talking about some pretty speculative ideas, so I want to start with something proven: the Mars Rovers. NASA’s had spectacular success with rovers working in conjunction with the Mars satellite infrastructure that’s in place, and even though the focus is on human exploration right now, I’m wondering if you’ve got future plans for more robotic exploration?
Mars plans for NASA are set up partially with the National Academy of Sciences, which tells us what we should do in terms of science planning. So far, Mars exploration has been more space science than manned spaceflight, but we’re trying to use what we’ve learned from the scientific missions to inform human space flight.
The challenge with human Mars missions is making them both safe and affordable — because it’s a problem to do both of those at the same time. One solution is to send robots & AI five years or so before humans arrive, and let them prepared in advance — generating water, oxygen, habitats, you name it.
Keep in mind that one of the main costs of getting to Mars is putting the 900 metric tons of equipage required in low Earth orbit — and if we can build robotic systems to make some of that on Mars itself using ISRU, it simplifies things and keeps the cost down. It’s a solution that offers real benefits, but the current plan is to take a more conventional Apollo-style approach.
Now, do you think AI is advanced enough to take on a larger autonomous role in these future missions?
The answer is yes. In terms of AI, there are three ways to approach using it. The first is with big-data & neural nets, which is what’s been in the media since around ’12 — but this isn’t the only way to do it.
Right now, there are several billion-dollar level funded human brain projects around the world that are nano-sectioning the neocortex on replicating it on silicon. This is biomimetic AI, which is an approach that may take it to the human level and maybe beyond.
The third possibility is to generate AI the way human acquired intelligence, which is by evolving to solve problems — a process of survival of the fittest, as our brains evolved the capability to solve new types of problems. At first we evolved capabilities on a problem by problem basis, but at some point we “woke up”, which is called emergence.
We know the principle of emergent intelligence is possible because we ourselves are proof that it happened at least once. So there’s that. Maybe we can make it happen in machines — there are some people who think it already is and the internet is starting to “wake up”.
In any case, the more data that you give the current approach to AI, the better it is. Even today’s machines can do brute-force ideation by the system level evaluation of quasi-random combinatorials. It’s not AI, it’s brute force evaluations — but it’s still really good. If you combine that with AI, it gives you a serious capability to handle the unknown unknowns you’ll encounter while exploring another planet.
NASA’s R5 Valkyrie could be used for autonomous Mars construction.
— Church and State (@ChurchAndStateN) July 28, 2020
Now in terms of the actual machines that you’d use with this AI, it looks like NASA already has several different experiments in the works. We’ve got rovers which have worked really well, and the humanoid Robonaut 2, & Valkyrie — and I’ve read about several designs for Mars aircraft as well. Do you have any idea of what NASA might be leaning towards in terms of actual exploration?
NASA starts with “what do we need to know,” both for science and for human exploration, and then we figure out the best way to accomplish our goal. What I’m talking about is really a design & planning process, and it’s ongoing — a work in progress, if you will. In terms of actual plans for our upcoming missions, the best place to find those is on the web.
The reason I asked was because despite the variety of proposed machines, the rovers have been truly remarkable. As you know, Opportunity exceeded its expected mission lifespan 55 times over, driving a total distance of nearly 50 miles on the surface. That kind of success seems hard to beat.
Well, look, space has almost exclusively been the domain of machines so far. I mean, there’s a few humans up there, but not often, and not many. Space has been automatic, and it’s becoming autonomous.
You know, on the earth or even the Moon, you can tele-operate. On Mars, however, you’ve got a light-speed & bandwidth delays of anywhere from 12 to 30 minutes, so remote control isn’t an option. The further we go out, the more autonomous our machines will have to become.
Yeah. Well, and so that takes us back to the AI part and now in a paper called Frontier Technologies and the Human Future: Sustainability Solutions, you wrote about machines becoming our “mind children”. Can you describe for me how this may shape future space exploration?
Okay. First let me define this term: it came out of Hans Moravec’s book on robots. The idea is that humans are becoming cyborgs. We have cochlear implants for hearing, artificial retinas for seeing, artificial hearts for living, and now we’re even printing artificial organs. DARPA is working on brain implants for soldiers, and the idea of uploading human minds into machines has gone from being science fiction to something plausible. So maybe the future of space exploration will be done by human-contaminated machines.
Human contaminated machines? Are you talking about putting your goals into the machine as a set of instructions or rules for exploration?
Well, it’s more than the goals. In theory you could put in your entire personality into a machine — and you are the machine! After all, your brain makes you who you are. The rest of the wet electrochemistry you’re composed of is simply there to keep the brain working.
So this is big picture then — you’re talking about changing the substrate that a human lives on basically.
Yeah. That’s one way of putting it.
Since we’re talking about big picture ideas, let me ask about the singularity. Do you see the singularity occurring, and if so, when would you guesstimate that we would anticipate to see it.
The singularity is occurring now, and it continues to happen over time as artificial intelligence and information technology develop.
Think about it for a moment: in today’s world, we do everything remotely: we telecommute, tele-shop, tele-educate, tele-travel, tele-commerce, tele-socialize, use tele-medicine & tele-politics. It’s tele-everything — and it’s becoming more autonomous with more AI built into it over time.
The singularity is where this convergence of digital trends becomes so pervasive that it’s faster than humans can keep up — but in reality, it’s happening slowly enough, that God, that people are accommodating to it, and largely, they’re okay. So I don’t view the singularity as a single event, like seeing a big storm cloud on the horizon and then being inundated with a downpour.
— Church and State (@ChurchAndStateN) July 28, 2020
I see what you’re saying. We’re adapting to it, normalizing to it, the machine is waking up, but it’s happening slowly enough we don’t realize it. For us, this is the new normal.
In your papers you’ve mentioned Marshall Brain, who said that we’re creating an intelligent species to compete with ourselves, and also Dr Steven Thaler for his work in machine ideation and creativity. Can you elaborate for me about creative machines that can outperform us in traditionally human areas of expertise?
Thaler was messing around with neural nets decades ago — before anybody thought they were worth very much. Anyhow, he trained a neural net, then deprived it of all rational input, and it basically began to dream, just like people do — producing new quasi-random combinatorials, new ideas.
Thaler took what’s called a critic neural net that collected all of these streams and evaluated them at the system level for various problems & metrics as solution spaces. This is an area that he’s refined extensively and literally wallpapered the space with patents over the years — but he’s producing real results.
Thaler’s process has been used to produce better toothpaste for Palmolive, better warheads for the Air Force, and ideate improvements in a lot of other areas. His work is an example of machines getting bigger & better, to the point where the machine is actually able to outperform a human being at creative tasks. The machine knows more, and it solves problems a lot faster.
So over here at NASA, we turned the original AI that IBM developed loose on cancer, and within a few years it was doing cancer diagnosis & treatment four times better than a human — and it was doing it four times cheaper because the machine can’t cash paychecks.
So AI works for cancer — but what about space exploration? Well, by using Thaler’s approach of brute force generation & evaluation of ideas, at machine speeds and at machine total knowledge, the machines can review millions of ideas in milliseconds, and come up with far better solutions than humans.
Keep in mind that Thaler’s approach is not AI, it’s just quasi-random combinatorial production with superbly fast & knowledgeable engineering system evaluation.
The singularity is happening, kinda sorta. https://t.co/W2NqMKqskV
— Futurism (@futurism) October 4, 2018
I’d like to switch gears a bit & touch on the environment. In your paper, “Where is it all going? Prospects for the Human Future”, you discussed the challenges with the Earth’s overburdened ecosystem. Can you tell me about some of the issues & possible solutions you’ve come up with from an environmental perspective?
For the Earth’s ecosystem? Well, we’ve been pursuing an unlimited economic growth mantra in a finite ecosystem. What that means is that we have way too many people asking way too much of the planet. People are just about agreed that currently we’re short about 50% of a planet now, and we are driving the ecosystem and a deficit each year more. As the rest of the planet tries to come up to Western living standards, we’re going to be short three or four planets. At some point we’re going to have to abandon today’s growth mantra and towards what you could call “green growth” — something I’ve written about extensively.
Now, in terms of the global climate change, after many years of struggle we’re actually creating solutions now for the climate because technology is finally able to reduce the cost of the renewable energy below fossil carbon cost. Not much happened with renewables until we got the cost down below fossil carbon — and now it’s going way below.
Once the price was right, greed came into play renewables are becoming popular. Currently 33% of all the electricity on the planet is from renewables, the cost of storage & generation has dropped some 80 to 90% in the last few years, and the costs are still dropping.
So greed works. What I’ve written about are various ways to avoid the huge losses from damaging the ecosystem along approaches like halophytes — salt plants grown on deserts using seawater. These are useful because 44% of the Earth’s land is wastelands or deserts and 97% of the water is seawater. This is the last natural resource that’s of any size the humans have not exploited — and it would allow us to land, water, food, energy, and climate issues and make build massive new industries in the process.
— Aeon+Psyche (@aeonmag) July 2, 2017
What about electric vehicles? That’s something we waited decades for, and how they’re here. Does that factor into your ideas for renewables?
Oh yeah. As renewables become a larger part of the electrical grid, using the electricity they produce for transportation will also grow. This is exciting because several types of transportation are now poised to transition to electric — and it has a lot of cost & environmental benefits.
The efficiency of electric propulsion is a factor of 2. Now in contrast, we’ve striven mightily to get only a 5% improvement in gas turbine engines, and electric is a factor of 2. Okay. There’s a huge list of benefits to electric vehicles in terms of efficiency, weight reduction, reduced maintence & lower pollution — and the bottom line is about a factor of four cheaper energy costs to go anywhere.
There’s another idea that I’m excited about from the Rocky Mountain Institute. They’re the best energy conservation group that I know of, and they have designs for buildings that produce energy instead of use energy.
I’ve seen estimates that between 30% to 50% of the energy use can be reduced as we go electric. So that could mean we’ll only need about half of the total number of exajoules that we use now because of increased efficiency.
I’ve also seen talk about renewables combined with conservation making energy literally too cheap to meter — and energy like that will make seawater desalinization possible for a lot more applications. It will also allow us extract the 44 minerals that are in the seawater out instead of land mining, which is one of the worst things we’re currently doing to the ecosystem.
In a way, this brings us back to talking about commercial opportunities, so I’d like to ask you about space commercialization — not just in orbit, but also what you’ve written about colonization & commercialization of the Moon, Mars, and beyond. I’m wondering when we’ll start to see real ideation for these projects as commercial ventures?
The current planning is what you see. It’s humans back to the Moon after all these years, and humans on Mars along with Musk’s efforts on that. If he’s as derring-do and successful as he has been in many of his other endeavors then it will serve to expedite this process.
In terms of Mars, you can make just about anything that humans will need on Mars itself, given the energy and In-Situ Resource Utilization. Mars is interesting because it has lots of resources, and as we develop faster transportation, with nuclear rockets and VAZIMR, etc, we can turn Mars in into the Walmart for the inner solar system, and maybe beyond.
Ultimately the question for commercialization is making money — and even now, commercial space is very successful. We’re putting up 20,000 to 40,000 LEO satellites to give everybody high speed internet & communications, and there’s also a large market for positional Earth utilities. The question here is how to commercialize space beyond geosynchronous orbit — and there are lots of ideas and even more unknowns for that.
For a while it was thought that mining water on the moon was going to be commercially viable — but we did a study of SLS LEO access numbers, and it turned out to be cheaper just to bring it up from Earth. There’s also talk about space mining of asteroids containing huge amounts of platinum, but if that was done seriously, it would probably crash the market here on Earth and the price would go down.
That being said, there are a number of ways to make money in space, and it’s something I’ve written about as well. One near-term idea is to use an energized tether to pick up space debris, put into a space junk yard, and then remanufacture & repurpose it. There’s an awful lot of high quality aluminum and a whole bunch of other materials up there.
About Our Guest
Dennis Bushnell is the Chief Scientist at NASA Langley Research Center, and is responsible for technical oversight & advanced program formulation. Bushnell obtained his M.E. from the University of Connecticut in 1963 and his M.S. from the University of Virginia in 1967, both in Mechanical Engineering.
Reprinted with permission from the author.
The ‘Space Architects’ of Mars | The Age of A.I.
The New Mars Rover Is The Most Advanced Yet, Here’s Why
8 NASA Robots That Will Study The Mysteries Of Space
Be sure to ‘like’ us on Facebook