Dr. Daniel H. Wilson debunks fears of robot uprisings while detailing AI’s fragmented progress—military robots like DARPA’s Big Dog and Samsung’s sentry gun rely on human oversight, despite efficiency gains, and current systems (e.g., facial recognition) exploit narrow patterns like eye-to-mouth ratios, easily fooled by simple alterations. Emergent behaviors in robot swarms mimic natural flocking but lack human-like adaptability, as AI thrives on deterministic rewards rather than complex emotional logic. Wilson’s upcoming book shifts focus from sci-fi threats to practical robotic solutions, like smart homes and Alzheimer’s care, where privacy concerns clash with predictive tech. Despite flaws—voice recognition failing due to accents or stress, chatbots collapsing under open-ended queries—AI’s future hinges on cultural adoption, with Japan and Korea leading in humanoid servants while the U.S. prioritizes niche applications like self-parking cars. The conversation underscores AI’s utility as a tool, not a replacement for human judgment, even as military and commercial uses push boundaries. [Automatically generated summary]
From the high desert and the great American Southwest, I bid you all good evening, good morning, good afternoon, whatever the case may be in your time zone, each and every one of them around the world, covered by this program, the very largest program of its kind in the world, Coast to Coast AM.
It is my honor and privilege to be now escorting you through the second half of the weekend, right on into the week, really.
And I guess it's been another sad day in the world.
We'll cover that in a moment.
It's never good news.
One of these days, before I ever do finally retire, I would love to come on and read nothing but good news.
It will not be this night.
Bombs kill eight American soldiers in Iraq.
Roadside bombs killed eight American soldiers in separate attacks Sunday.
It's really getting bad there.
A car bomb claimed more than 30 lives in a wholesale food market in a part of the Iraqi capital where sectarian tensions are nothing but on the rise.
In all, 95 Iraqis were either killed or found dead nationwide on Sunday alone.
They included 12 police, among them the city's police chief in Samarra, who died when Sunni insurgents launched a suicide car bombing and other attacks on police headquarters.
So no good news from Iraq.
The wreckage of a Kenya Airways jetliner told you about it yesterday, a crash found late Sunday in a dense mangrove forest outside Cameroon's commercial capital.
Dozens of rescue workers and journalists walking through the swamp at night did reach the edge of the crash site, but didn't find any survivors, nor did they find the main part of the plane.
They'll go back and look for that tomorrow.
It kind of closed down at dark.
Rescue workers on Sunday searched for anybody still buried in the heaps of splintered wreckage left in Greensburg, Kansas after a massive tornado obliterated, literally, most of this south central Kansas town.
Now, there are making the rounds on the internet pictures of Greensburg, Kansas, or what's left of it, not much.
Waves of thunderstorms ripped across the plain states on Sunday, drenching rubble that the Friday tornadoes scattered across Greensburg, and in addition, threatening more tornadoes tonight elsewhere.
The pictures on the internet are just awful.
I don't know what scale that tornado was, but it just literally obliterated that town.
I don't think it was an F5.
So you can imagine what an F5 would do.
In France, they've got a new president, and things are going to change.
Nicholas Sarkozy, a blunt, uncompromising pro-American conservative, was elected president of France Sunday with a mandate to chart a new course for an economically sluggish nation, struggling to incorporate immigrants and their children.
Sarkozy defeated Socialist Royale by, let's see, 53.06% to 46.94.
They had an 84% turnout in France, according to final results released early Monday.
Boy, that's big.
It was a decisive victory for Sarkozy's vision of freer markets, toughness on crime and immigration over Royal's rather gentler plan for preserving cherished welfare protections.
I almost said warfare, including a 35-hour work week that Sikozi is calling absurd.
35 hours.
Police Chief William Bratton in Los Angeles said Sunday that up to 60 members of an elite squad that swarmed into a park and fired rubber bullets during a May Day immigration rally are now no longer on the street.
Bratton said he spent the weekend viewing video of the MacArthur Park incident, and he said LAPD failures were widespread, with officers from the top on down culpable.
In a few moments, we're going to go to unscreened open line phone calls.
Anything you want to talk about us here again?
Oh, by the way, next hour, I'm really, really looking forward to this.
Dr. Daniel H. Wilson will be here talking about artificial intelligence and robotics and his belief, apparently, that robots will rise up and kill us all.
That's right.
Terminator type robotics.
Anyway, if you'd like to call in and make a comment, here are the numbers quickly.
West of the Rockies, Tolfree 800-618-8255.
East of the Rockies, 800-825-5033.
First time callers, area code 818-501-4721.
Wildcard line, folks.
Area code 818-501-4109.
We've got a lot of those.
So a good chance to get in.
Area code 818-501-4109.
And finally, if you're outside the country altogether, we've got a toll-free line for you.
Get hold of your international operator and tell her you wish to call 800-893-0903 in a moment some of the rest of the news.
By the way, last night's program with the brother, I thought was extremely educational.
And I think I said it toward the end of the program.
Many people will take it in many different ways, and it could indeed be taken many different ways.
If, I suppose, is a lesson in fundamentalist Catholicism.
I guess you could take it as an insult to the current state of the Catholic Church.
You could take it as underlining many things I've said about fundamentalism, no matter where you find it.
You could take that show and its contents many, many, many different ways, but it was extremely educational.
Okay, scientists in the U.S. say that initial data from a new way of scanning Mars has now shown up to half of the red planet's surface may contain ice.
The new method of scanning for water offers vastly more accurate readings than anything previous.
The data could prove vital for the Phoenix Mars mission, which launches this August and which will put a lander on the surface and it will dig for ice.
The new data shows wide variation as to how deep below the surface the ice is.
The deposits far beyond the ice that is known to exist in the planet's north pole could be so large that were they to melt, they'd literally deluge the planet in water, forming an ocean that would cover the whole planet.
In Britain, the British government is taking Big Brother pretty literally by adding lip-reading technology to some of the 4 million or so surveillance cameras now already in place in order to identify terrorists and criminals by watching what everybody says.
Electronic Design is reporting the Home Office is interested in a project being pursued by a senior lecturer in computer vision at the University of East Anglia in Norwich, England.
Computer-based lip-reading technology would help video surveillance systems spot people planning a crime or a terrorist attack by literally watching suspects' lips move for clues.
They did that several years ago, began it actually at a Super Bowl game that I went to, so now it'll move from facial identification to what's the guy saying?
Boy, I'll tell you.
Consider someone who's just died of a heart attack.
His organs are intact.
He hasn't lost blood.
All that's happened is his heart has stopped beating.
That's actually quite a bit the definition of clinical death, actually, and his brain has shut down to conserve oxygen.
But what actually has happened?
Has he really died?
As recently as 1993, when Dr. Sherwin Newland wrote the bestseller, How We Die, the conventional answer was that his cells had died.
The patient couldn't be revived because the tissues of his brain and heart had suffered irreversible damage from lack of oxygen.
This process was understood to begin just, oh, four or five minutes after the heart stops.
If the patient doesn't receive cardiopulmonary resuscitation within that time, and if his heart can't be restarted soon thereafter, he's unlikely to recover.
The dogma went unquestioned until researchers actually looked at oxygen-starved heart cells under a microscope.
What they saw amazed them.
According to Dr. Lance Becker, an authority on emergency medicine at the University of Pennsylvania, after about an hour, we couldn't see evidence the cells had yet died.
We thought we'd done something wrong.
In fact, cells cut off from their blood supply died only hours later.
But if the cells are still alive, why can't doctors revive someone who's been dead for an hour?
Because once the cells have been without oxygen for more than five minutes, they die when their oxygen supply is resumed.
It was that astounding discovery, Becker says, that led him to his post as the director of Penn's Center for Resuscitation Science, a newly created research institute operating on one of medicine's newest frontiers, treating the dead.
Biologists are still grappling with the implications of this new view of cell death, not passive extinguishment like a candle flickering out when you cover it with a glass, but rather an active biochemical event triggered by re-perfusion, the resumption of oxygen supply.
The research takes them deep into the machinery of the cell to the tiny membrane-enclosed structures known as microchondria, where a cellular fuel is oxidized to provide energy, microchondia control, the process known as apoptis,
I guess it is, the programmed death of abnormal cells that is the body's primary defense against cancer, it looks to us, said Becker, as if the cellular surveillance mechanism cannot tell the difference between a cancer cell and a cell being re-perfused with oxygen.
Something throws a switch that makes the cell die.
With this realization came another, that standard emergency room procedure has it exactly wrong, backwards.
When someone collapses on the street of cardiac arrest, if he's lucky, he'll receive immediate CPR, maintaining circulation until he can be revived in the hospital.
But the rest will have gone 10, 15 minutes or more without a heartbeat by the time they reach the emergency department.
Then what happens?
We give them oxygen, says Becker.
We jolt the heart with paddles.
We pump in epinephrine to force it to beat so it's taking up more oxygen.
Blood-starved heart muscles suddenly forced and flooded with oxygen, precisely the situation that leads to cell death.
Instead, Becker says we should aim to reduce, that's reduce oxygen uptake, slow metabolism, and adjust the blood chemistry for gradual and safe re-perfusion.
In other words, we're doing it exactly backwards.
Speaking of hearts, here's a man in Montreal with no heartbeat, yet he lives.
He is a 65-year-old Quebec man who received a new long-term mechanical heart last month, which is being described as the only living Canadian without a pulse.
Dr. Renzo Cecier implanted the Heartmate 2 mechanical heart into Gerald Langvin in a three-hour operation November 23rd.
Officials at McGill University Health Center say the device, which is about the size of a flashlight battery, might last up to 10 years.
Now, that's quite a few more years than conventionally you get what they currently have, lasts two or three years.
So he has no detectable blood pressure whatsoever and no heartbeat, and yet he lives.
Cost about 100 grand.
They paid for it.
Good news for those of us who were, and I certainly was, very dismayed at the cancellation of Dead Like Me.
I thought that was a great show.
The dark comedy about the trials and tribulations of a group of grim reapers going to come back to life, at least briefly.
Dead Like Me, the popular Showtime TV series starring Mandy Hatkin and Ellen Muff, I interviewed Ellen Muff, is returning as a made-for-DVD movie.
MGM told Moviehole.com that Stephen Herrick, the Mighty Ducks, will direct the new film.
No word yet who's going to be starring in the film, but I hope they keep the original cast.
Well, all right, there you have it.
All the rest of the news.
Let's now move to one of the wildcard lines and say, good morning.
This is conspiracy theory, I know, but I'm kind of thinking about the X-Files.
I know I've watched an episode dealing with the bees, and then I start thinking about all these microbiologists that have died, and it's funny how that's like amazing, and yet there's no accounting for all these scientists, microbiologists that have died.
And I start looking at that.
I start connecting the dots, and I start wondering, gee, is this just all, I mean, I don't know.
Well, I'm not sure they're connectable dots, but certainly we have lost a disproportionate number of microbiologists, scientists.
We have lost a lot of our bees.
And in fact, what?
60% nationwide, I think 27, 28 states now, something like that.
So something awful is going on.
And it involves the environment.
Now, I don't want to suggest...
Look, the political conservatives in America are not really wild about environmental concerns.
So, you know, I'll leave it for you to judge those who mention this disaster and those who don't and what their motives might or might not be.
unidentified
Well, Art, the one that I guess I was following this in the internet, Shaney is taking a look at this, and I don't place much faith in Shaney, and I don't know.
It's scary.
And, God, I hope that they get a lock on this, because if they don't, we might just start facing getting used to eating, I mean, having bread and water.
All our food crops, just about everything is affected.
I really, I have no comment because I have no proof that there's any specific agenda being executed.
Yes, I did intend to use that word by anybody out there.
I just don't have any proof, so I don't know any more than I know about the bees.
But I know that our climate is changing, and if the bees on top of that disappear and we're unable to pollinate crops, we're in an awful lot of trouble.
In 1955, 1956, I believe it was, President Eisenhower signed an executive order ordering the Justice Department, corporation of the Interior Department to approximately there were over 4 million, they assumed at that time, where in less than 15 months, over 2 million of them were deported.
And the funny thing about it was that those who came from Mexico, they actually bused them back.
Those from the Caribbean, they actually put them on ships, and South America put them on planes.
In other words, it was something that can be done if you apply yourself.
Now, just one more thing about immigration.
Are you familiar with something called I-N-I-T-I-N?
Yeah, the IRS is giving those out to people who can't get Social Security numbers.
It was something signed into law by President Mr. Clinton in 1996.
In a 10-year period from 96 to 206, over 10 million of those identification numbers were given out.
Approximately 8.2 million of them went to illegal immigrants.
Now, Homeland Security is responsible at this time for finding, you know, doing their job, which they're not doing very good at, finding them and deporting them.
The IRS Refuses to give them the names of these individuals, claiming that it's against federal law, which I think is the most absurd thing you can ever think about.
If they're here working, we might as well get the tax revenue, whatever it is, as opposed to not getting it at all.
That's not to say that I support illegals being here, because I don't.
Period.
However, getting the tax revenue is of assistance to the country.
These are separate issues.
In my mind, they're separate issues.
Might as well go ahead and get the money.
From my point of view.
Otherwise, that's more money that you and I may have to pay.
Right?
Now, the argument between the federal agencies and the issues of privacy.
I certainly cannot resolve that one.
That'd be like tearing your IRS refund in half or something.
I don't like the member tearing the baby in half.
I can't make that kind of decision.
But we might as well get the money.
Now, as far as illegal aliens are concerned, we obviously have got to do something.
We're being overrun, and I fully support whatever measures they'll take, whether that's building a fence, whether that's extra patrol people on the border, whatever it takes, we've got to stop the illegal immigration.
We have a right to protect our country, enforce our borders, and we have more of a reason than ever to do so because many coming across, or at least some, I guess, coming across, are not just coming here for better conditions, to send money home, to get a job.
We can well imagine some coming across could have evil intent.
Yeah, Art, we'd spoken briefly last month about the aspect of the Luciferian doctrine as related to the globally elite wealthy of the planet.
And you probably remember that conversation, I have a feeling.
I do.
And tonight, I'd just like to take it one step further.
And you had posed the argument basically that when the economy is going well, that that's when these people would prosper the most owning the world's economy, essentially.
And in a way, that question that you pose answers another more important question about what is really at the seat of power as far as the objective of those who actually have control of the world's economy and many believe control of the world's political structure.
And to give a short answer to it, it's basically to control men and eventually to control men's souls.
When it actually gets into the area of being a trillionaire art, they've got many times more money than they could ever spend, but they continue to furiously amass it.
And they do control the world and the people of the world in Moss through their method of social engineering and what is referred to as the Hegelian dialectic, which is a philosophical way to implement changes into the world which will bring about the kind of world that they want to have, which is really nothing that either you or me or any common person basically on the planet would want to see the world be that way.
Well, the way things are going and the way, you know, there's so many areas that are deteriorating in the world at this point.
But to look at what seems to me, and I think I believe you also to be the most pressing issue upon us, which would be euphemistically, I think, called climate change.
It's really global heating beyond global warming, basically.
So little has been done to remediate this in any sector, whether or not it be chlorofluorocarbons, which I firmly believe that chlorofluorocarbons finally exhausted the transformative ozone gases, which gave us the cooler green and blue rays that we got in the summer and the springtime of the sun.
Now we're getting ultraviolet year-round, and that's what's melting the glaciers, and that's what's drying the woods out and causing all these forest fires around the world.
They're paying absolutely no attention to it, but they do pay very close attention to management.
In other areas of perpetuated wars and unnecessary wars and economic deprivation and starvation and famines, all of this, there's just been plenty down here as far as resources and technology goes that we shouldn't have these ills as a race.
Basically, you know, I know that we're in Iraq, you know, for security reasons, but it's hard for me to take that seriously since our borders here at home are pretty much completely open.
They're not completely open, but they're certainly very leaky.
unidentified
Yes, sir.
We have, you know, a pretty, it's not that hard, you know, from my understanding at least, for people, determined individuals at least, to come across our borders.
And I just feel, you know, if one of these people wanted to harm us, it would probably not be that hard.
And here we are in Iraq, you know, spending so much money every week to fight this war and to stabilize this country.
And I feel like, you know, that amount of money we could be spending on perhaps, you know, nuclear bomb sniffers or anything.
You know, but here's something to think about, sir.
There are plenty of people, as you know, who want to harm us.
As of yet, that has not occurred save or since 9-11.
So somebody's doing their job and doing a pretty good job, I must say, of it.
We don't hear about the successes they have.
So you might make the case that what we're doing in Iraq is concentrating our enemies' assets there instead of here.
unidentified
Yeah, that's true.
I guess I just really feel like our, you know, being a taxpayer, I just wish my money was being spent to militarize our borders or do something to really protect the bulk of Americans here on our homeland.
But you are correct.
I'm sure that in some form, you know, the war in Iraq helps to avoid terroristic activities or for them to actually carry out the act.
But from my understanding, most of the, you know, the best thing that we've done so far is domestic surveillance to stop terrorists who are here in our country.
If you do any travel around the world, there's terrorism, of course, is quite rampant in Europe, throughout much of the Far East, frankly, the rest of the world.
And we've done a pretty good job here.
Now, you just don't hear about the successes.
CIA, FBI, Homeland Security, they must be doing a fairly decent job, and I'd like to commend them because we just don't generally don't hear about the successes, but you can argue that the lack of anything happening, even given the somewhat leaky, fairly leaky borders that we have, somebody's doing a pretty good job out there.
I'm sorry, I'm going to have to cut it off there, sir.
We'll be back and discuss robotics and more in a moment.
I'm Art Bell.
What a world we live in, huh?
It is an amazing world.
My wife currently sitting in the other room, having simultaneous video conferences with her sister in Arkansas, her other sister, and her husband on the island of Mindanao, southern Philippines, and then also chatting with her niece, who's at school, Caguan, which I mispronounced, I'm sure, also on the southern island of Mindanao, all going on at the same time.
It is an amazing world.
And coming up, Daniel H. Wilson will add to it.
Daniel earned a Ph.D. from the Robotics Institute of Carnegie Mellon University, where he also received master's degrees in robotics and data mining.
Data mining.
He has worked in top research labs, including Microsoft Research, the Apollo Alto Research Center, PARC, that's P-A-R-C, and Intel Research, Seattle.
Daniel currently lives in a fully wired, smart house.
He is the author of How to Survive a Robot Uprising, which I'm particularly interested in, as well as Where is My Jetpack, as well as a contributing editor to Popular Mechanics magazine.
So lots to talk about coming up with Daniel H. Dr. Daniel H. Wilson in a moment.
Yeah, I think a lot of that stuff is in our future.
Although they, of course, in the movie, they play around with some different possibilities.
They have Gigolo Joe, which is a robot that satisfies carnal love.
And then there's, of course, the little boy David, who satisfies more of a higher level sense of love.
And there's a lot of indications that people do interact with humanoid robots the same way they interact with people because they just the way that the robot looks, it provokes and invokes all of that natural built-in, all those natural built-in mechanisms for human interaction.
And in any scientific endeavor, you try to nail down some really specific problem that you can solve.
And you can say, okay, look, I did this.
Now let's go to the next problem.
But AI is a really hard problem because in order to make a human-level intelligence, you have to solve the AI hard problem, which is you have to solve all of it.
And that's really, it's tough to do the whole enchilada at once.
So what we have right now is a lot of research in artificial intelligence for solving really specific problems.
So it's not what most people think of when they think AI.
But this is maybe there are people who spend their entire careers working out vision algorithms that can, so you can have a machine that can look at a scene and then say, that's a cat, that's a dog, you know, the cat is on top of the dog.
Or, you know, that's a human, he's running away, that kind of stuff.
So in terms of getting general purpose artificial intelligence to a human level, I think that that's one of the hardest problems in robotics.
I think it's harder even than the mechanical problems of building a robot.
And there's one surefire way to know when you've reached your goal, and that's if you have a robot that can pass what's called the Turing test.
So Alan Turing was a mathematician and a computer scientist, and he came up with a test to determine whether or not you can tell whether you're interacting with a human-level intelligence.
The really simple version, that's not a full robotic version, is you essentially go into a chat room with someone, and you talk and you're allowed to write whatever you want to write.
And if you can't tell the difference between, if you can't tell whether the person you're talking to is a human or a robot, then, hey, you know, it's a problem.
Yeah, well, so people think that the difference between the way the human brain works and the way most, the way a computer processor works is that the brain is just much more parallel.
So it's got a lot of different processes happening at once.
And there are a lot of, so artificial intelligence is really seeking to reproduce human intelligence.
And so the researchers often draw inspiration from real brains.
And I have lots of friends who are neuroscientists.
And one way to do that, well, actually, so my advisor at Carnegie Mellon was really interested in something he called, and I could be getting this name wrong, but he called it the Cyber Baby Project.
And the idea was, you know, a lot of the reason why these AI algorithms, why the chatbots, they don't really understand what you mean whenever you say ice cream or going to the park.
They don't have all those experiences.
And so the idea is to put a camera on a baby, you know, from the baby's perspective, and microphone, and just record every single thing that happens for an entire lifetime.
And then if you took that childhood and stored it on a hard drive, then you could ideally train an artificial intelligence algorithm to know everything that a human knows.
And as it turns out, you can calculate how much data the eyeballs are collecting every second, if you think of it as video and a camera, and how much data your ears are collecting, how much data all your senses are collecting.
And then as it turns out, it's not that much.
If you run a video nonstop, it's a lot, but it's not infeasible even now to store that much data.
So we are very finite.
Everything we experience in our lives someday soon will fit on every experience.
Well, so artificial intelligence is just a field of research.
If you want to say, I guess you would call that general purpose human-level intelligence, because so many people work on problems where they make really smart algorithms that learn really fast and do their job really well.
So for instance, every time you, if you've got a newer model car, or at least not an ancient car, when you hit the brakes, if you've got an anti-lock brake system, you've more than likely got a neural network that's sitting in between you and the actual brakes.
When you push down on the pedal, you're really just pushing a button.
And the button tells a neural network, The human wants to slow the car down, and then it figures out how to slow the car down.
Because honestly, as a human, you're not as good as it is at actually stopping your car.
You'll wreck your car.
The analog brakes will keep you from wrecking your car because it knows more about braking than you do.
The top car was actually from Carnegie Mellon, and it made it about seven miles before it high-ended on a rock and broke an axle.
The next year, after 12 months, I think at least five cars finished.
They were actually passing each other.
I think the head car, the winner was from Stanford.
I should mention that the guys behind the head car had just left Carnegie Mellon to go to Stanford, so a little pride there.
But anyway, Stanford beat us out.
It was humiliating.
And their car, it was just a modified Volkswagen Toureg.
And cars these days are so drive by wire, which they're just controlled by a computer that's sitting inside the car that can actually brake and accelerate and do all these other things as part of different safety features.
The car could actually, they just tapped right into that.
They just tapped right into the computer, and then they put a bunch of sensors on the car, and it went 130 miles into something, I forget, you know, like 10 hours or something like that, had an average speed of 30 or 40, but this is over just rough terrain.
There were cliffs, really amazing.
And next year, or actually later this year, they're kicking up the stakes, and now they're not going to be racing out in the empty desert where no humans were allowed anywhere near these cars, right?
Who gets blamed when somebody gets hurt?
This year, they're going to be doing it in a mock urban environment.
So it's really clear, it's really obvious what DARPA is going for.
And that is unmanned ground vehicles that can operate in crowded urban environments with following street signs, everything.
And so if you look in there and you look and see what UGVs are called for, they're not primarily weapons platforms.
One of them is a mobile ambulance that actually, you know, so there's no place for a human to sit in these things.
It makes them much more efficient.
They can go further with less fuel.
And so what this thing is supposed to do is go out on a battlefield, find a hurt soldier, open, you know, a little door opens, and it drags the guy in there.
And while the person is in this little, basically a mobile coffin, there are robotic arms that are being teleoperated by surgeons, you know, back in the States.
And so you have someone ideally doing surgery, doing, you know, whatever has to be done on this poor soldier while he's getting taken out of danger.
When I was young, I went to see Forbidden Planet with my dad, and we all thought that by now, you know, I'm beginning to get along on the tooth here, we'd have Robbie and more, but it didn't happen.
Well, so, you know, the field of artificial intelligence, really, they thought they made some wild predictions early on because they had some early successes.
Some of the very first times that people tried their hands at AI, they were able to get some really great results.
They had programs that could solve, prove equations on their own.
So I mean, that seems really smart.
They could play chess.
They could beat a human at chess, right?
This is a very, you know, you have to be a real smart person to play chess, right?
It's very cerebral.
What they found out was that all this high-level thought, all these very, with a lot of, you know, games where you have a lot of rules or math where you have, again, a lot of rules, this is the natural domain for machines just because you program them with rules.
And so it's everything, essentially, it's everything a two-year-old can do is what was really hard.
So recognizing shapes and colors, getting up and walking, learning how to walk after you fall down, learning to recognize speech, learning to recognize faces, social stuff, all that stuff, really, really hard.
And as it turns out, if you look at the human brain, there's just huge areas of the brain that are devoted to all these things.
There's the motor cortex, there's all the vision stuff up in the front.
Even Even if it's way in the back and hardly any part of the picture, it picks out the face and puts a little square around it.
That's astounding.
We'll be right back.
I never really gave it a lot of thought until just now.
But I was amazed.
I bought that little camera.
I think it was a Canon, perhaps a Canon, and the very latest, you know, 10 point-something megapixels, really hot stuff.
And after I bought it, I realized it was picking out faces.
Now, that's an interesting technology when you think about it.
No matter where they move, where they are in the room, it finds the human face and puts a little box around it.
Now, how does a camera know how to do that anyway?
And I know that we talk too much about the possible weapons applications of technology, but if a camera can do that, imagine what a gun could do.
And I suppose it could look at the face and decide if the face is the enemy or a friendly.
I'm not sure how it would do that, but if it can do what it does now, it can probably do that, and then it can probably order the trigger to pull.
There's a terrible thought for you.
We'll be right back.
In the pressure of combat, Dr. Wilson, many friendlies are unfortunately killed.
Friendly fire kills a lot of people.
If a camera can decide where a human face is so well, and it does a great job of it, as I said, even if it's only a tiny, tiny part of the framed photograph, it still finds the human face and puts that little box around it.
Then it seems to me, how big a step is it from that to, for example, deciding if that's a friend or foe and whether the gun should shoot?
Yeah, well, those are questions that are being asked right now by a lot of people.
In fact, I just today actually read a document that came from the Naval Surface Warfare Center.
And it is a, basically, they're laying out, there's something called the laws of armed conflict, or the law of armed conflict, and this just says what you can and can't do.
It's related to the rules of engagement and the discriminate use of force.
And they're trying to figure out what to do with this when it comes to autonomous, you know, autonomous, whatever they call them, weapons platforms, right?
And so what the latest is, what I've just read, is that one way to get around this, because you cannot, basically it's agreed that you cannot have a machine that's going to not have a human in the loop.
There always has to be a person welded to the machine to give it the go ahead, to actually use any lethal force.
Because actually discrimination is really, that's a key part of law of armed conflict.
Any weapon has to be able to discriminate.
You have to be able to target what you're after.
You can't just drop something that's going to kill everybody and have all this collateral damage.
That's a huge strike against it.
But what they're trying to get away from is actually having to have a person there all the time because these things are obviously much more efficient if there's nobody watching the trigger.
So here's what they came up with, and I think this is really interesting to think about.
Machines only attack machines.
And humans kill humans.
So this is the way they've got it set up.
And they have a lot of precedents that they've laid out.
There's something called a captor.
It's a subhunter.
And what it does is it's basically a mine that sits on the bottom of the sea.
And when a submarine comes by, an enemy submarine, it launches a torpedo.
It's just like a mine that only kills submarines.
There are other tank mines and things like that that won't trigger for people, but they'll trigger for tanks.
What's really interesting, so they're saying that this has already been done before.
What's really interesting is the number one thing that they're concerned with is that the enemy will spoof a system designed to only attack machines, and they'll intentionally try to get it to kill people because it would be a very bad publicity for the machine.
And obviously it would mean that the machine was used inappropriately, right?
Well, the three laws, I think, are a beautiful thing.
And they're not in place.
The machines are being used pretty much in whatever way they can.
And I mean, for the record, sticking a gun onto a mobile robot is nothing new.
And these, you know, it's already happened.
In fact, Samsung has a mobile gun.
Actually, wait, it's not mobile.
It's just a sentry.
So it's this gun with a camera on it.
It's exactly like, I mean, it's very similar to the camera that you have.
And what it can do is it's sitting along, South Korea bought a bunch of these things, and apparently they're deploying them along the demilitarized zone.
And they're just going to sit there, and they watch, you know, I think it's a one and a half miles swath there.
And they recognize people.
And if people come too close, I assume it'll shoot at them.
And so far in Iraq, there have been a lot of mobile robots that have saved lives by going out and tagging the improvised explosive devices, right?
So instead of having a man walk over and kick a junk pile to see if there's a bomb in there, they'll send a robot.
And, you know, I've actually, at conferences, met Marines who have used these things, and, you know, their intent, of course, you know, they're there to sell the, particularly what I'm talking about, the Talon from Foster Miller.
Are you actually afraid that one day in our future, and I don't know how far it is, maybe not as far as I might imagine, robots will rise up and make some sort of decision to kill those that made them?
In a robot uprising, which is what I consider, you're not necessarily fighting military robots.
If it's a robot uprising, that means all the servant robots decide to come at you.
So now you're fighting your Roomba, right?
But as it turns out, the sensors are the most delicate, and they also have the most constraints, because a lot of times they'll have to be, for instance, a camera has to be line of sight.
It has to have, you can't put it behind a thick piece of steel.
For instance, I actually worked for Northrop Grumman for a little while, which is a big military defense contractor, and we were working on an autonomous boat.
So it's just an unmanned surface vehicle, a boat that drives itself.
And we were not allowed to use any active sensors.
So sensors that bounce signals off of other things, like a laser range finder or ultrasonic sensors that operate the way bats echolocate, because that gives away your location immediately.
And so it was much harder to design this thing without.
I think instead of looking at why Americans are not interested, I think first it pays to look at why the Japanese are so interested.
And I think part of that is their culture, obviously, and some religion.
There's a really good book on this called Loving the Machine, which is all about why the Japanese love robots.
Part of it has to do with Shintoism, which is a religion that associates sort of a life force with every inanimate object, including robots.
So they think of them naturally as something that's alive, and so they forgive them for not maybe performing up to a standard that makes it worth the money to buy them.
You know, actually, I found that the United States, we're doing a lot of work on humanoid robotics, don't get me wrong.
Part of the reason why the United States, Western Europe, and Japan are really interested in humanoid robots that are going to be in our houses helping us out is because in the States, all the baby boomers are going to retire or starting to.
And in Japan and Western Europe, they're also noticing that their population, the demographic is skewing so that there are a lot more retired people than there are workers.
There's nobody to take care of everybody that's retiring.
And the robots are literally the only caretakers that a lot of people are going to be able to have.
You know, in my own research, I have, my research is all about putting sensors into people's houses, and the sensors connect to an AI, and the AI basically figures out what your patterns are, and then it looks for functional decline.
Like, most of the women in my family are nurses.
My mom's a case manager.
She's a nurse that visits people in their homes and makes sure that they're able to do what they can, that they can perform all their activities of daily living.
If they can't, she gets someone in there that helps them out.
Well, no, he'll probably now, depending on how complicated the robot is, the guy in India may have his hands in gloves so that he can teleoperate the robot's arms and hands.
So that he can do something.
Like, dexterous manipulation is really tough for Things that are slippery or things that are deformable.
So, washing dishes is pretty tough.
There's a robot that can do it, but it has to have specially colored dishes and it uses plastic dishes for obvious reasons.
But, you know, why do all the research?
Why not just have a guy put his hands in some gloves and let a human do it, you know?
So, this is the idea at AnyBots.
The problem is that there are human eyes behind the robot cameras.
Well, so what I found in my research, which I was, you know, I will talk about my research all day, so feel free to tell me to shut up about it.
So what I found in my research was that I could not use cameras to, you know, there's this, what I found basically was that if you're going to invade somebody's privacy, there has to be a huge perceived benefit.
There has to be, you're really doing something that's crucial because nobody wants to give up their privacy for nothing, you know, and I totally agree and understand that.
That's why, you know, that's why these people let my mom come into their house and ask how they're doing.
And I've got to think about this outsourced intelligence for robots in your home.
What do you say, folks?
I'm Art Bell.
We'll be right back.
Here I am.
If you just joined us.
Daniel H. Wilson.
Dr. Wilson, who earned a Ph.D. from the Robotics Institute of Carnegie Mellon University, where he also received master's degrees in robotics and data mining, is with us.
And he's talking about artificial intelligence robotics.
In a moment, we'll ask about Terminator robots.
All right.
Dr. Wilson, what about the possibility of Terminator robots?
So, you know, if, God forbid, you end up in a situation where you've got a Terminator coming after you, as it turns out, they can run as fast as your car.
So really, it's all about fooling the sensors, because that's a robot's weak spot.
They're not really good general-purpose thinkers.
So Terminator might be good at tracking you down.
So what you want to do, for instance, is it'll be looking for skin colors.
So you might want to cover up your skin or put on some kind of camouflage.
Another thing is it'll actually lock onto the trackers will generally lock onto the clothing that you're wearing in order to disambiguate you from the background.
So if you can't actually get something, some kind of cover, like something substantial between you and the robot, then get some clutter in between yourself and the robot.
And that's just anything visually that's going to that's going to make it harder for the robot to see you and figure out.
And that's something else that's interesting about it's really hard for machines to figure out the physical properties of objects that you throw.
So it can't tell, for instance, if you throw a rock, whether it's a rock or a balloon or what the physics of the object is.
One last-ditch thing you might do is if somebody else is wearing something similar to what you're wearing, you can just run over there and give them a big bear hug.
What will happen is you'll combine into one entity, really.
If you think about that bounding box that is on your camera, where there's two boxes around two different people, when they hug, they combine into one box.
That's true.
When you run away, there's a 50% chance it'll chase the other guy.
Well, you know, I thought the piece I was writing, which may, you know, is still in the process of being edited, was really all about the differences between Japanese and American approaches to creating humanoid robots.
And so what struck me about that was that it was really innovative.
Well, it's because there's no, basically it's because there's no bottom line.
We're really driven by, robotics research is driven a lot by healthcare.
You know, how are we going to deal with the baby boomer crisis?
So that's where I got all my money, for instance, to do my research.
It's driven a lot by DARPA, which that does not necessarily, by the way, mean that whatever you create, if you produce it on DARPA money, that it's going to have a gun attached to it.
I mean, this is the agency that funded the development of the Internet, you know, and a lot of other really basic fundamental research.
So when you get money from DARPA, right, everyone will assume that you're evil, you know, and that's part of why I wrote How to Survive a Robot Uprising is because I wanted to say, you know, look, robots are not always evil, and the people that are making them are definitely not evil.
We all have these great applications in mind, and we need money.
So you get things.
One of my favorite examples is I have a lot of friends who work on something called RoboCup.
And RoboCup is where you get a team of five or six, I forget how many, robots who all cooperate together to play soccer.
And a lot of the RoboCup money is from DARPA.
And what happens is once a year, when DARPA comes to visit, they take these really cute little Aibo dogs.
They look like little soccer-playing dogs.
They are the cutest things you've ever seen.
And they, you know, they put them in these tactical warfare situations.
And it's a lie.
It's just fundamental research about multi-robot cooperation.
And sure, DARPA wants to use it for its own reasons, but everybody doing the research, they all have a lot of other reasons that they're really working on it.
And it doesn't necessarily involve DARPA, just their money.
And so that's my rant, where you say, just because you've got money coming from the military doesn't necessarily mean you're working on something evil.
It throws tremendous amounts of RF up into the ionosphere.
And originally it was to do communications research and that sort of thing.
But it was taken over by DARPA.
And now there are better than rumors that say, well, they're thinking they can bounce this immense RF, modulated a certain way off the ionosphere and confuse enemies on the battlefield, or worse, at a great distance.
So there's an example of DARPA taking over and perhaps modifying the goals a little bit.
So what my house specialized in was really using non-invasive sensors.
So sensors that don't bother people.
And what I stuck to was anything that you can find in a public restroom.
So when you stick your hand under the faucet and the water comes out, that's a sensor that's detecting your presence.
So motion detectors, contact switches.
I filled my house up with those.
And then I put in a learning algorithm that could learn to spot when it was me, when it was my roommates, when it was just the dog.
We tended to find that the dog would place itself equidistant between all the people in the house and just create this noisy data that was just awful because it was systematically noisy.
It's pretty easy to get rid of noise if it's random, you know.
So what happens is it has to be really smart about predicting things because I'm not giving it a lot of data.
I'm not giving it, for instance, a camera image in my face because that's just never going to, you know, people, it's never going to fly, especially in the bathroom.
So, what I found with this research is that it's completely possible to do it.
And what I really found personally is that, you know, humans, myself included, are just so predictable.
Well, okay, so there's, in a smart house, you've got two aspects.
You've got the tracking aspect, right?
Like figuring out, I call that simultaneous tracking and activity recognition.
And then you have what does it do for you, right?
I did not focus on the what does it do for you.
But what's actually what's really interesting about this research, the big picture, right, if you go forward years and years with this, is not what can it do for you, but why does it do it?
So learning causal rules in order to improve your life.
So the application here, sorry, I think this is really cool.
The application here is to put one of these smart houses in, let's say, an Alzheimer's ward, where you have a situation where you don't want people to wander off.
And it might learn over a period of time that people afflicted with Alzheimer's tend to avoid dark hallways.
And if a machine learned that, in order to decrease the probability that someone is going to wander out of an open door, it might turn the lights off.
And in that way, it's not saying it's impossible.
It's not closing a door.
It's just reducing the possibility and the probability.
And learning really complicated rules like that could be pretty interesting.
You say it can parallel park itself, but it would not recognize a human behind it and in front of the car that it's about parallel park in front of, right?
Because that's the main difference from highway driving and driving in a city is you have to be aware of everything around your car because you can accidentally hit somebody if you're not paying attention.
Whereas if you're on the highway, you just need to look straight ahead pretty much.
So yeah, this is already, all of this is going together.
What's really interesting is there's something called the slam problem, the simultaneous local simultaneous localization and mapping.
And this is where you've got to figure out where you're at and you have to form a map of the local of the area that you're in at the same time.
And the way To do that is to figure out which things around you are static, like a mountain or a lamppost, and which things are moving, because you can't use the things that are moving in the same way as you can use a landmark type thing.
And so that means figuring out who the people are.
And that can mean some face recognition stuff going on.
So these cars are really, really aware of their local area, and they have to be.
So one thing that you can also do, what that face recognition is doing in your camera is it's finding two dark, it's basically finding a triangle of dark spots that correspond to two eyes and a mouth.
That's pretty much how they do it.
And there's the same ratio.
So that's how it can do the scale, right?
If a person's right up in the front of the camera and their head's really big, or if their person is very far away in the background and their head's really small, the ratio between that triangle stays the same.
And so that's how they can solve the scale problem.
Now, one thing I suggest is to, this is something tropical fish do, if you were to put a big black dot in the middle of your forehead, or you know how a tropical fish will have a big black dot on its tail, and it makes it look like its eye is on the other end.
And so, whenever it moves, it fools the predator, which is looking for the fish to turn the other direction.
Well, I think a lot of the law enforcement software, well, first of all, if you put a big black dot on your forehead like that, you're probably going to be more conspicuous than if you didn't.
That's a good point.
The thing there is you probably just want to put on a mustache or some other type of less conspicuous.
And this is why the computer, the thing is, if you're doing face recognition and you know that the person you're looking at is one out of ten people, let's say, you got it, or maybe it's nobody.
So you've got like a one in 11 chance of getting it right.
You know, they made reference to that in the article.
All right.
Hold tight.
We're at another breakpoint.
Dr. Daniel H. Wilson is my guest.
We're talking about AI and robotics.
I'm Art Bell.
There's actually a question in here about what you would do if your smart house turned murderous.
And of course, that would never happen because, well, of course, Dr. Wilson has programmed it very carefully.
But maybe there's a potential question here about a virus in the house.
That's something to think about.
We'll ask about it in a moment.
Dr. Wilson, the whole prospect of a smart house is really cool.
I really, really like the idea.
And I don't imagine for one second that it would become murderous and turn against you.
But in this day and age, I would imagine that, for example, it would be irresistible to have part of your smart house hooked up to the internet.
Now, that would mean that potentially somebody could gain access to the software controlling your house, and they could put in some sort of virus or something that would literally turn it against you and it would murder you.
Well, you know, I think that it's all about preparation when it comes to the hypothetical scenario of escaping from your smart house.
What you really need is a plan, you know, kind of like, I don't know, when I was a kid, I had to practice the fire escape plan, you know?
Which route do you take in order to get out of the house?
You know, and this could involve some sort of emergency supply stash.
So you've got, you know, let's say an axe.
The house is going to probably try to distract you with a lot of loud music, or it may even try to use evil robot logic.
Everyone knows that exists.
And so what you do is you just go through the house.
You destroy the sensors whenever you see them with the axe.
You can chop holes through walls if you have to until you're able to get all your family members together and then escape.
And the thing to remember, though, is once you get out of the house, you're still not safe because if you're living in a smart house, chances are you might have a smart lawnmower or a smart car.
Well, that is a really practical aspect, actually, of having a smart house is it collects all this information that it can share.
I saw a really cool study where there's a mobile robot that's in the house.
And the thing is, you know how you do that dance whenever you're at the grocery store and somebody's going down the aisle?
You bust it down.
You don't want to be doing that with your robot all the time in the house.
I mean, what would be more annoying than that?
So I saw a study where they actually used data from a smart house, from an embedded sensor, to tell a mobile robot when to get out of the way of a person that's walking through.
Yeah, the jetpack, first of all, is key, and it's also a metaphor for all the other technology that we kind of thought that we would have.
Jetpacks, so the lowdown of jetpacks, really quick.
1961, a guy named Wendell Moore gets an Army grant.
He's got some money.
He's working for Bell Aerosystems.
So he's working for an aerospace company.
And he has money, and he's got the facilities, and he builds the Bell rocket belt, and it only takes him about a year.
And essentially what he did was he took a small rocket lift device, type of rocket that's used on airplane wingtips to maneuver really high up where there's not much air.
And he stuck it on his back and tested it himself with a tether, and he flew, right, for 30 seconds.
And that's the major problem: this thing was invented in the 61.
I think they made about four of them, and a lot of people tested them.
Wendell Moore actually eventually shattered his knee and never again went up in his rocket belt.
But there are rocket belts now available for purchase, but they're all copies of this.
Really, by now, the technology should have, seems like it should have been updated in the intervening years.
However, the Army, once it got the product that it paid for, realized that it was not going to be useful for soldiers.
Normally, soldiers are trying to hide behind things, not hover 20 feet over the battlefield like sitting ducks.
And so the Army wasn't interested in continuing funding.
And so since then, there's been no concerted effort.
And it's really, I think, an instance where it shows how if you've got funding and a high degree of cooperation between a lot of people, you can get science done really fast and you can achieve really amazing results.
But if there's no huge call for it, then it's up to the individuals who are operating not on funding, but on childlike awe, you know, intensity, the desire to have these things.
Well, the 1961 jetpack, which is what's available, only has 30 seconds of fuel before it sends you careening toward the ground, which is a major drawback.
There are some other drawbacks that, you know, I try not to focus on why we shouldn't have jetpacks, but just where are they and where can I get one or steal one or build one if I need it.
But another drawback is that it's shooting out superheated steam.
So it's really hot.
You've got to get one of those white Elvis jumpsuits.
Well, yeah, but you know, if they had been working on it all this time, it seems to me that we really would have a practical jetpack that probably would give you more than 30 seconds.
And in fact, Wendell Moore did get additional funding from DARPA in 1969.
He made a legitimate jetpack.
The Bell rocket belt is technically a rocket, so it's technically a rocket pack.
He built a jetpack, which was a jet engine on your back.
And this thing was more substantial.
It was a lot bigger.
And it had a parachute on it because it lasted, I think, geez, I mean, 10 times as long.
I mean, 100 times as long.
It lasted maybe a half hour or something.
Wow.
The problem is, again, well, actually, the problem is that Wendell Moore passed away, had a heart attack and passed away during the course of this research, which caused the funding to get cut, and then it was broken down and used in a different flying device.
Well, I have a friend who is orbiting satellites privately right now and plans eventually to have a space station in which we can all perhaps take a vacation one day.
Yeah, so this is something that I've kind of caught onto a little bit.
So the X Prize was offered to anybody who could take a ship up about 70 miles and do it again and hang out there for a little while.
And if you look at that, what you see is that when you get to 70 miles, two things happen.
You can see the curvature of the Earth, which is really beautiful.
And it reminds you that, you know, it's not flat.
It's finite.
You know, you remember how the huge impact that the first image of Earth from space had on the public consciousness.
And the other thing is you're weightless.
You know, what's more associated with space than being weightless?
And what this means is space tourism.
People want to go.
They don't want to go to 50 miles.
That's just like being in an airplane.
You go to 70 miles, though, and now you've got something people will pay for.
Well, the next big prize is offered by Bigelow, yeah, and it's the America's Space Prize.
And this is to the person that can get a ship up about 200 miles around where the shuttles go, and then can dock with an inflatable space station, which is a space hotel.
Yeah, you know, I wonder if an astronaut can come back to Earth and just, you know, trip and fall off a cliff and just be totally cool with it through the air.
Well, but I think that that's on the short horizon, seeing special purpose robots that are around the house, you know.
And it's kind of, I've noticed this thing where, you know, when I started writing this, I was thinking, you know, I just was out of a bar conversation.
I said, you know, where is my job?
Where is all this stuff?
You know, I was kind of incensed.
And then I started looking and I realized, you know, we got a lot of it.
It's just that we immediately, immediately take it for granted.
As soon as we have some technology, it's just part of the scenery.
And I think it boils down to the difference in hardware between human beings and between machines.
And so if you think of love as being some sort of chemical reaction in a human brain, then, you know, no.
If you want to make that kind of robot, just, you know, have a kid.
But in terms of having a machine that perfectly mimics all the physiological symptoms of being in love, yeah, we're all over that.
And you know what is amazing?
Actually, this is, sorry, I get excited about robots.
What all the researchers have found is that human beings will just project like crazy onto anything that's even remotely animal-like or lifelike.
So there's this really great, you know, it's one thing to be weighted on hand and foot by a robot.
Actually, I don't think that's really that healthy, particularly for someone who's losing their functionality over time.
What's really good is to give somebody something to take care of, to give them something that empowers them so they feel like they're, you know, useful.
And there's a robot called Paro, who looks like a baby harp seal, which is awfully cute.
It's a little white thing with big black eyes.
And as soon as you see it, I mean, you just, you can't help it.
You want to pet it.
I saw it at a conference, and it was handed around a room.
Every roboticist was, I mean, these are people that are familiar with being around robots, but they were pushing each other out of the way just to pet this thing.
Well, I mean, if a cat feels jealous, I mean, it's a cat jealousy, right?
But what you're projecting is you're imagining that it feels jealous the same way that you feel jealous.
And, you know, the more human-like an animal is, you know, the better, the easier it is to think of it as a human and to anthropomorphize it, you know.
So cats are little furry animals.
And, you know, if you back up far enough and you squint your eyes, humans are little furry animals too.
You know, We're not that different.
Now, if you compare a human to, say, a starfish or something like that, it gets a little tougher.
And they found that whenever they give robots human-like qualities, you get a lot of the natural human mechanisms.
So there's this one robot that's kind of getting old now.
It's called Kismet.
It's at MIT.
And one thing that Kismet could do, it basically looks like a little gremlin.
And so if you take that and you project what could be done and look at the next 20 years or so, I would imagine you could just about totally manipulate a person.
Now, he is an expert in artificial intelligence and robotics.
So if you have a question in these areas about what can be done now, what's going to be done shortly, and how you'll handle it.
And oh, by the way, I'd love to know how you feel about robot intelligence being outsourced.
That one just hit me right between the eyes.
I'd love to hear it.
So any questions for Dr. Wilson coming up next?
Interesting.
I was thinking about unconditional love, Doctor, just before we go to the phone.
I truly believe my cats unconditionally love me.
I really do believe that.
Now, some of that may be me projecting what I think based on, I don't know, their instinctual reaction.
So a lot of it may be me projecting, but I truly believe they love me.
So it couldn't be that hard down the road to produce an artificial intelligence that probably can go past what my cats, the emotion my cats are able to produce or my projection of that.
So it would be pretty tough to duplicate the abilities of a cat with a robot, even just those abilities.
But, you know, they actually showed that the Paro robot, when people had it, they gave it to elderly occupants at a nursing home.
And they were given this robot, and they actually showed that they had higher reports of happiness, lower blood pressure, things like that.
They had real physiological effects.
So whether or not people say they're in love with it, whether or not it instills all these feelings, the bottom line really is whether it helps you or not.
And in that case, it does help those residents there.
Well, then, if they really got cooking on this, if DARPA really put some money behind it, which they're unlikely to, we really probably – Don in Omaha, Nebraska, you're on with Dr. Wilson.
unidentified
Hey, Art.
Thanks for making me fall in love with radio all over again.
Sure.
I have a question I think you're both going to like.
Have you tried using your camera to take pictures of your cats?
Does it focus in on their faces or is the ratio wrong?
And I want to ask your guest if that doesn't work.
Are they working on something that will get kitty and doggy faces too in the little box?
I've had them since I was quite young, so I'm glad you've caught on.
They're amazing, amazing creatures.
They are complicated.
They're a little more standoffish than dogs are, so they're not for everybody, but they're amazing animals.
Tom in Seattle, you're on with Dr. Wilson.
unidentified
Hi.
Tom?
Yeah.
Dr. Wilson.
Hello, Art.
Hi.
I've been privileged some years ago to have a few alien abductions, and I have run into alien artificial intelligence on several occasions.
And on one instance, I had a trip to a strip of foil that had a master programmer in it and was populated by alien programs that were designed to do various functions.
Yes, I know, but again, I ask, how do you know it was artificial?
How do you know it wasn't real intelligence?
unidentified
Biological intelligence.
Well, I could see orbiting electrons around copper atoms.
The computer was based on the data storage of altering the orbits of electrons around the atoms, and you could see that inside the machine, and the entities would come up and talk, and so on.
In another instance, I ran into a dead alien that was 450 years old, required by law to die, and had his mind installed inside a machine.
Simply point your cell phone at the face, and it gives you the name.
unidentified
Yeah, I was wondering if Dr. Wilson had heard of a program that occurred slightly after the Roswell crash, about 49 or 50, that began in incurring artificial intelligence into human agents.
And this program was terminated when Kennedy was assassinated in 1963.
There was a program to, well, actually, machine existence and machine intelligence has existed, you know, other earthly, otherworldly, you know, considerably longer than even human existence has existed on this planet.
But there was a program initiated by the United States government shortly after the Roswell crash to combine human intelligence with artificial intelligence.
What you're saying out here, when you're seeing all these saucers and stuff, they may contain aliens and stuff, but the saucers and such are intelligent machines.
Now, here is a question for you, which occurred to me as he was speaking.
We're beginning to approach the time when people are starting to talk about taking the contents of a human mind and eventually uploading it to a machine storage unit.
Yeah, I don't, you know, I think that it may be possible to.
So the interface between a brain and a computer is pretty complicated.
And right now, there is some really cool research in brain-computer interfaces, and it's not, I mean, you know, it's very far away from uploading your mind onto a computer hard drive.
But what it's all about now is it's about starting to figure out what the electrical activity in your brain corresponds to in the real world.
So for instance, the most, there's a company called CyberKinetics that has taken a human test subject, a person who sadly is in a locked-in state, who has no control over any part of their body, except their brain.
And so what they do is they actually open up the brain and put electrodes into the surface, and the electrodes pull out the electrical activity that's happening there.
And it happens, you know, that electrical activity happens for a reason.
So if you stick it into the motor cortex, you're listening to a person trying to move their limbs.
And people have actually been able to move computer cursors, anything that you can connect to the computer, you can move with your brain here.
And it's really useful if you lack control over your limbs.
You could say that the human programmed it in the first place, but when it rewrites itself, I mean, then, and really, I don't even know if that argument is enough either.
I mean, I can program something in five minutes that will learn a behavior, and it will do things that, you know, what's really interesting is whenever you get a lot of robots together, let's say you have thousands, even if they're just simulated, they'll start exhibiting behavior that you could never imagine that they would.
This is like they've gotten robots that can flock like birds and they can swarm like ants.
And all this behavior completely falls out without, you know, in fact you set it up and then as humans you study it to see what they'll do because you have no idea.
unidentified
I understand that.
And I agree with you on that.
I guess it's just a matter of how you look at freedom of choice.
And I don't think the human mind, and I am a man that believes in a God, I don't believe a human mind can ever be recreated or even saved by a man because I think that it's something totally independent of the body.
And, you know, of course, goes on to another place.
I don't think that takes anything away from your belief in a higher power at all.
unidentified
I understand that now.
The comment I wanted to make about a propulsion system.
If you were to, robotics has a problem with movement, you know, climbing stairs and freedom of movement.
Has a propulsion system been developed where you use magnets within the robot and you work on a grid that is able to repel the magnet, if you understand what I'm saying.
How you can take a magnet, turn it one way, it attracts the other way, it repels.
And you could have a series of magnets within the robot such that it would provide lift and propulsion just by tilting these magnets.
And then the floor, the work area this robot would be in, the space, would have a metallic floor, either a sub-floor or a metallic tile.
And then there would be grid within there for coordinates.
Is that something, and that way you could have a very lightweight robot.
It could climb stairs because it would never touch them.
And you'd have to structure the environment so that you have metal floors?
That's right, yes.
I have not heard of any research like that, although I have spent a fair amount of time instrumenting floors to try to figure out where people are at.
But yeah, I haven't heard of anything like that.
Although, you know, I'm sure that that sort of thing would have wider applications.
It would probably be pretty prohibitive to force people to put metal all throughout there.
I did actually see a really interesting muscle application where it's a magnet that goes basically and replaces where, just like where a human muscle would be.
And when you power the magnets, they contract.
And whenever you depower them, they just kind of go loose again, just like real muscles.
That was kind of interesting, magnet-related, but yeah, that's all I got.
It's about the ability of machines to follow us around on the ground and not only read our lips, but read the structural changes in our skulls and so forth that creates words and can actually read our thoughts.
You need a robot that can follow a person around so that it can help the person out.
What's interesting, there's a quadruped called Big Dog that runs on a diesel engine.
It's from Boston Dynamics, and it's designed to follow soldiers around.
In terms of communicating with a robot, you can do that with sub-vocalizations with a microphone that's on your throat, and it can pick up the beginnings of words and figure out what you're saying.
And there are also bone-conductive speakers that can put sound in your noggin.
That is, what our guest thinks about the idea of replacing our judicial system with artificial intelligence, namely, he says, to prevent judges from legislating from the bench.
Well, a lot of times there are decisions made that are real headshakers.
And you wonder why somebody got a sentence that just didn't seem right.
I wonder what Dr. Wilson thinks about that.
We'll find out in a moment.
You know, I suppose in a way, in a way, computers or robots, if you will, could be seen to almost be reading minds.
If they could learn body language, they could almost be seen to be reading minds.
But to the question about our judicial system, wouldn't that be almost an ideal starting point for artificial intelligence, Doc?
I think that's a really scary setup, mainly because, well, there's really no room for exceptions with AI a lot of the time.
So that would be an expert system like I talked about earlier where it just takes into account every rule and every precedent and then makes a decision.
And that is highly deterministic, actually.
The problem is that it would never know when to make exceptions for special circumstances or special things that's never seen before.
And that's kind of a key, you know, I guess that's what, if you wanted to avoid any new legislation from, you know, that kind of thing, then yes, this would be great.
But the fact is, there's a reason we don't have the same laws we had in 1900.
You know, it's because continually new precedents are being set and new situations are arising, especially and maybe ironically because of technology.
Can you imagine all the new laws that would have to come into effect?
Well, and another thing you can have is this is kind of cool, actually.
There's a lot of research on how humans can interact with robots where you do let the robots do sort of low-level stuff, but you still need to keep an eye on them.
And so NASA has this plan for robots that are in orbit building giant space stations, you know, just 24 hours a day, mostly autonomously, but sometimes not.
And the idea is to have a human who can just, from a computer console, flick in and out of different robot bodies to oversee what's going on.
And the robots learn when to ask for help.
And so what you could have is a situation where, you know, robot judges are working together with humans and calling on help whenever something interesting happens or something that might call for a human.
Well, there are different ways to make robots work together as teams.
One really easy way is to have a central computer that just controls all the robots.
If you've ever seen Star Wars Episode I, there's the part of the movie where they destroy the orbiting space station and all the robots just fall down because they were all controlled from one vulnerable point in the middle.
That's not really how any animals that I know of actually work.
And so what you do instead is you have a lot of, you can do distributed robots where they each are pretty cheap to produce, they're pretty cheap to make, and they're pretty dumb.
But they all follow these same simple rules, but when they all follow them at once, some higher level intelligence falls out of it.
So for instance, if you had a lot of robots and half of them were red and half of them were blue, and you told all of them, if you see someone blue, get behind them, and if you see someone, you know, or line up with the blue or whatever, what you'd see is instantly they would all shift into a position.
You know, you tell all the blue robots to get behind red and all the red robots to get behind blue.
Or if you have a bunch of robots and you tell them to avoid obstacles but stay a certain distance from the other robots, what happens is you get flocking behavior.
They'll all flock around in one group, but Then, whenever they come across an obstacle, they'll part and then they'll flow back together.
I have a thought on the computer chips for the robots and not necessarily have the computer chips in the robots, but have them my idea is stackable computer chips that you could build a supercomputer and they're liquid cooled.
Is this a possibility that maybe this could be done?
In fact, whenever you have a lot of computers all in one place, that's the biggest thing that they have to do to make sure they don't break, is they have to cool them.
Now, I don't know how, I guess I don't know enough details on how stacking them could improve that.
I know that liquid cooling is a good way to keep chips cool.
It's one that's in use.
There's case modders that like to modify their computers, and a lot of times they'll install flashy liquid-cooled devices.
My question is, I have found that with voice recognition programs, like on the OnStar system of cars, and I live in Detroit, the home of George Norrie.
We have one of the big utility companies there.
You call in, it's nothing but a computer system that has voice recognition, that they don't recognize my voice.
And it became a scary reality when we had a gas leak at our house and I couldn't get through to anyone, and the computer just wouldn't go through the prompts.
It just, again, the commands that it told me to give, it wasn't listening to.
I have used, so, you know, I mentioned earlier that the smart house I designed, you know, didn't do anything.
Earlier on in my research, I had it designed so that that's all it did, was it would do things for you, like turn on and off the lights or the radio.
And the way to do that would be to tell it to do what you wanted it to do.
And that meant a speech recognition system on a wireless microphone.
And when I was training it, you have to train it.
So you have to talk to it so it learns your voice.
And I had, at the time, I had an Australian girlfriend, and it could not, there was no way that it could be trained so that it would recognize me and her because it's so speaker dependent.
And because the difference between male voices and female voices is huge.
The difference between accents is huge.
If you have a poor phone line, or if you have a lot of background noise, that can be huge.
That can really lower speech recognition accuracy.
And when you have a system that just has to recognize anybody who's going to call in, you can't train it for a particular individual, then you really lose some accuracy.
And so I guess some people are just out of luck because the technology is not there yet.
That was probably Dragon Dictate, which I also used and was unable to get it to work.
The problem there, again, is that it's not a limited domain.
You can say anything.
I don't know how many words.
I think there's 300,000 words or something in the English language that are commonly used.
And the probability of you saying any one of those words is 1 over 300,000.
That is really hard for a computer to predict what you're going to say next, even if it hears you.
What happens to increase the accuracy is when you write a lot or once you've used it a whole lot and gone through and corrected it after it's wrong again and again and again, eventually it learns how to figure out what you're saying.
Another interesting thing is people learn how to talk to the machine.
Over time, we are all learning all the time.
We can't help it.
We can't turn it off.
It's what makes people great.
And so we learn how to game the machine so that it's more likely to recognize what we're saying.
And that's another interesting thing that people do.
Well, I think that it's actually just more efficient to put text on the screen.
But they do have, it's called speech synthesis, and they do have some pretty good stuff.
What's interesting, AT ⁇ T has a system.
It's called TTS, text-to-speech.
And the way they do it, and you might be interested in this because you've obviously got a great voice, is you can give it some segments of your voice.
So you record your voice, and it will take all of those, combine them together, and then you give it a few words.
And so it can, using your voice, synthesize any word.
It breaks down all the phonemes.
It basically takes all the pieces from all the words you've said and saves them up.
The trihune-brain theory model, surprisingly enough, I came across this in a self-help relationship program that I had gotten.
The human brain, the idea is that at the core of the human brain is a reptile core kind of around what I believe would be the amygdala.
And surrounded on the outside of it is the mammalian brain.
And complete outside, the gray area, is the thinking brain.
And so we have reptile brain, mammal brain, and thinking brain.
The reptile brain is the hands.
And the mammalian brain is the heart.
And the thinking brain is the head.
And the key here, what I'm trying to get at, is the reptilian brain.
What it's basically composed of is it's trying to decide whether we want to kill something or to copy with it.
And all three of these things work hand in hand.
The reptilian brain being the physical part of us, the hands.
The male brain being the emotional part of us, the heart.
And the thinking brain being the logical part of us.
Now, these are the three areas of human history.
Physical, logical, emotional.
When you combine physical and logical, you get science.
When you combine physical and emotional, you get art.
And when you combine emotional and logical, you get psychology.
I think this has been completely absent in the study of AI.
And I think maybe something like this can cause a revolution because it's a little bit risky to put something into an AI, whether it wants to kill something or to copyright with it.
I have a question, but I'd like to preface it with a little bit of information, and it'll become readily apparent why.
First of all, I heard the doctor speaking about a robot that was essentially identical to a human from the waist up with joint movement and so on.
And I only caught about the last hour and a half of your program so far.
Then the other issue I wanted to mention was the Isaac Asimov type robot where you're talking about something that's a humanoid type robot.
Well, okay, with those two things mentioned, the question was there was a project and I believe it started in the 90s, I want to say 1999, and it was supposed to be a 10-year project to synthesize human intellect.
And the project was, it's kind of hazy because I just, I think I read it in Scientific American or someplace, but it was a 10-year project.
And five years into it, either DOD took over or just kind of took the whole thing and said, we're not giving any more information and we're going to control it.
Did the doctor or you have any information about that project and humanoid work?
So I, you know, in fact, before I came on the show, I was trying to remember the name of a project that's really similar to what you're describing, and I could not for the life of me remember it.
But this is a project that it would not have been, if it's the one I'm thinking of, it would not have been associated with a humanoid robot because it was really for a chat bot.
So the idea was to pass the Turing test, to have something where you type a question and it answers just like a person would.
And the problem is that there's just so much stuff that we pick up.
You know, and I talked earlier about recording the life of a child from the very beginning in order to get all that stuff.
Well, their solution to that was to just type in every fact that you could think of.
And I mean just having people on the internet, having people hired, just everything you can think of, just in sentence form.
And then there was a lot of natural language processing to pull out what all that stuff meant and to make all those connections.
And I think that ultimately, yeah, it didn't pan out and it didn't pass the Turing test.
And ultimately, I think we figured out it's pretty futile to try to do that.