Stephen Hawking is Wrong About Artificial Intelligence

For probably the first and last time in my life, I feel Stephen Hawking is wrong about something scientific.

Stephen Hawking wrote a column for The Independent warning about the risks of people taking artificial intelligence (AI) going too far. The sentence that Dr. Hawking wrote that caught my eye as incorrect was, “..., there are no fundamental limits to what can be achieved: there is no physical law precluding particles from being organised [sic] in ways that perform even more advanced computations than the arrangements of particles in human brains.”

This is not correct, while computers are fast and can do calculations faster than man, AI is boiled down predictions based on statistics and classification algorithms. While good predictions are right most of the time, they are still wrong sometimes.

AI algorithm are wrong because the world is still dynamic and unpredictable. AI will never be prefect because of nature and human free will. AI will make mistakes, and AI systems needs a human to verify and train the system. The bottom line is there needs to be someone there to be kept accountable, and with this. humans are needed whether de jure or de facto for a society-wide acceptance of AI. Dr. Hawking later in the article states his main point, that we need to understand how AI will affect our lives, and I totally agree with that and I think society will evolve to make the policy and legal decisions for such.

These societal rules will come in place because its human nature to blame someone for mistakes.

There will always need to be someone responsible for AI mistakes whether it is an errant autocorrect in your text message to your Mom or mistaken airstrike from an armed drone. There will always need for someone to stop the buck. When thinking about AI remember there will always be someone there wanting to cover their backside.

No Skynet is not coming and it will never come. AI will never totally take over the world. The need for CYA overcomes any AI.

College Athletes Should Be Paid $92,000/Year

By my calculations, the average college college should be making $92,000/year. How did I come to that number? I compared the average salaries of other professional athletes to the amount of revenue their leagues brought in to get a ratio of how much an average salary for an athlete should be.

In the table below I have the estimated revenue, average salary, and my estimate of the total league players salary and ratio of salary/revenue for each of the four major North American Leagues

League	Revenue	Average Player Salary/Year	Estimated Number of Players	Estimated Revenue Per Player	Ratio of Salary to Revenue
NFL	$11,000 mil	$1.9 mil	1696	$6.48 mil	29%
MLB	$7,000 mil	$3.3 mil	750	$9.3 mil	35%
NBA	$5,000 mil	$5.2 mil	450	$11.1 mil	45%
NHL	$3,300 mil	$2.5 mil	810	$4.07 mil	58%

The revenue and salaries were from the Major Professional Sports Leagues in the United States and Canada Wikipedia page and the number of players in the league was calculated as the roster size of teams (53 NFL, 25 MLB, 15 NBA, and 27 NHL) multiplied by the number of teams per league (32 NFL, 30 MLB, NBA, and NHL). The range of league ratios ranges from 29% to 58%.

As a note, the NFL collective bargaining agreement is 47%, so either the numbers are off, or the NFL has revenue that is not part of the collective bargaining, and it looks like the NHL union has done a great job with negotiating with the league.

Using this knowledge I calculated what the salary NCAA athletes should be making by estimating the revenue of a NCAA D-1 athletics departments at about $4,800 mil. I got his number from the ESPN college sports revenue and expenses from 2008. The median revenue was $40 mil for 2008 and I multiplied that by the number teams at 120. Then I estimated the number scholarship athletes at a school to 150. This number is rough since schools have varying sports per school, but I figured 68 football, 30 both basketball teams, 20 baseball and softball, 10 swimming, 10 tennis, and 12 as fudge factor. I decided a fair wage would be about 30% of the athletics department revenue based on the fact that college athletics is not the cash cow professional sports are and there is more overhead in the athletics departments. For the year 2008, that would put a fair salary at $80,000/year for 2008. With an growth estimate at 2% for 6 years that would put athletes at about $92,000/year for 2014.

Also, median revenue/athlete in NCAA D-1 is about $300,000, I feel paying athletes is just plain fair. Undergrad research assistants help professors generate income through research and they get money, graduate teaching assistants generally get a scholarship plus a stipend for their teaching and research efforts, why are college athletes getting their only getting their tuition and room and board paid for? At my alma mater, New Mexico State Universtiy, the athletic department brought in $25 mil for 2008, and I estimate the average football player with tuition, books, room and board cost the university about $20,000, based on 150 athletes thats about $ 3 mil in expenses, or salary athlete expense to revenue ratio of 12%. NMSU's athletics program is dwarfed by programs like Alabama or Texas where the ratios will be even lower at about 5% athlete expense to revenue ratio.

I propose the NCAA schools pay their students a flat salary across the board on top of their scholarships, say $45,000/year (which is more than I made as a grad student) plus the scholarship students would be earning at the school. This will be the salary per scholarship player across all sports to keep in the spirit of Title 9. This flat salary would prevent rich schools like Notre Dame or Alabama from buying players off. Also, the money would encourage students from dropping out due to the higher opportunity costs. I know in sports that are not football or basketball, coaches can split scholarships across multiple players, the coaches can still do that for those sports.

This is America, when people participate in the capitalist system they should be compensated for their hard work, and I think college athletes are not getting their fair shake.

Link to Big Data Article

I have not posted in a while due work picking up and the fact that I am back teaching two classes at Pima Community College (Introduction to Assembler and Introduction to C#). However, I wanted to post this great article about from Wired, Big Data is Too Big for Scientist to Handle Alone.

The Doppler Effect and the Police: A Match Made in Heaven: Or How Does a Radar Gun Work?

Radars can measure distance and they can measure velocity of objects. A police (or baseball) radar gun measures speed. To measure velocity, radar’s take advantage of the Doppler effect.

Army MP using radar in Iraq

Basics of a Doppler Radar

The Doppler effect is basically when the radar's frequency (how quickly the signal moves up and down) is different of what the radar expects it to be when it receives the reflected wave when it comes back or is received to the radar. When a wave hits an object that is not moving the frequency of the radar does not change, but when an object is moving the phase compresses when an object is moving toward the transmitting radar and decompresses when an object is moving away. When the frequency changes, it is called a Doppler shift. A larger Doppler shift translates to higher speed. A Doppler radar keeps track of what the expected wave frequency should be when the radar receives the signal if an object in not moving; when it receives a signal that has a different frequency than what it expects, it measures the difference between what the radar says it should be and calculates the speed of the object.

Doppler effect model showing compressed waves on the right as the object moves toward something and decompressed waves on the left side as it moves away

However, as any radar/light/x-ray etc. beam moves away from the transmitter the beam spreads. Think about a flashlight, as it gets further away the beam spreads, the same thing happens with a radar. That is why when you get pulled over the state trooper or police officer will say something about he got a good read on you, which is his way of telling you that there were not cars around that there is a good chance that the beam was not calculating the speed of other cars around you.

On some car radar detector, you may see that they can detect Ka and X band radar which tells you what frequency they are transmitting at. Police usually use radars with a frequency of Ka or X which are about 35 GHz or 10 GHz, respectively. 35 MHz means the radar wave fluctuated up and down 35 billion times per second. For comparison, an over the air TV signal transmits at UHF which is between 400 and 800 MHz (400 to 800 million hertz) and light transmits at 430 THz (trillion hertz) for red and 750 THz (trillion hertz) for violet. Radars are what is called coherent which means the Doppler can be measured. The newer laser radar also called lidar, light radar, the police use are not coherent. So how do they work?

LIDAR Speed Guns

A lidar can measure speed based on repeatedly measuring the distance an object of interest. Police like lidars because they have a narrower beam which makes them less likely to get measurements from multiple cars and they are harder to jam. Instead of using radar, lidars use a beam frequency that is infrared which is not visible to the naked eye, but has similar properties to visible light like a much narrower beamwidth. Like a radar, a lidar can measure range based on the fact that light and all waves travel at the speed of about 300,000,000 meters/second or about 671,000,000 mph (the speed of light constant). A lidar measures the time between it transmits and receives a pulse, then multiples that time by the speed of light and divides by two (to account for the time it takes for the beam to travel to and from the object). For example, if you were 150 meters from the lidar it would take 0.000 001 seconds for the beam to transmit and be received by the lidar gun (300,000,000 m/s times 0.000 001s divided by 2). It then does that measurement many times in less than a second then calculates the change in distances over that period to calculate your speed. For example, you went from 150 meters away to 110 meters away in a second you were traveling 40 m/s or about 90 mph (and if a cop measured you would probably get pulled over).

Police officer with a lidar gun

For both radar and lidar, the speed measurements are the most accurate if your car is going directly at the radar beam, since police or radar cameras are on the side of the road the speed measurement is not going to be the true speed you are going.  Unfortunately, the speed the radar gun collects is slower than your actual speed, so it is not an error you can use to get out of a ticket.

Speed Gun Jammers

To counteract the radar guns some people use radar jammer or scramblers, you can read more about these in a previous blog, but basically the radar jammers figure out what frequency the police officer has and sends out a nonsense signal at the frequency so the cops gun cannot detect the speed. These jammers are generally illegal for civilians to have since they transmit high energy radio waves which are regulated by the Federal Communication Commission (FCC).

Lidar guns are much harder to detect and jam because laser beams are much narrower than radar (that is why most police agencies using them now), but a laser jammer works in the same way as radar jammer.
Most police point lidar guns at your front license plate since it is a flat surface and it will bounce back more directly to the lidar gun. To counter act that, some people use techniques that are used by stealth aircraft to counteract radar. Stealth airplanes use the fact that if the beam does not get back to the receiver, a measurement cannot be made, so they have body shapes that reflect the beam away from the receiver with offset shapes and have radar absorbing paint, so some dedicated people have lidar absorbing paint on their cars and license plates or have their license plates slanted so the beam will bounce in to the sky.

This game of coming up with counter-measures for radars and then coming up with counter-counter-measures radars is played by both drivers and police as well as militaries when it comes deceiving radar.

What is a Jammer?

Tucson company Raytheon just won a major contract to develop upgraded radar jammers to the US Navy for the F/A-18s. I thought this would be a good opportunity to explain radar jamming. My first notion of radar jamming came from the 1987 Star Wars spoof, Spaceballs. So, I felt the need to include a clip of the radar jamming scene below.

This scene inspired me to ask 25 years ago, “What is radar jamming?” 

Jamming a radar is a way to prevent antagonist from effectively detecting an object. Before I get into jamming, let me give a brief explanation of radars. Radars are able to detect objects distance from a radar receiver, the speed of the objects they are detecting, and in some cases the angle of the object relative to the radar’s current position. If you do not want the radar to know this information, you jam the radar.

What a jammer does is make the radar not be able to do all of those things. However, you know when you are being jammed. So, think about as if someone does not want you to see them and they poke you in the eye, you cannot see and know you have been poked in the eye. On a side note, when someone cannot see you and they do not know you are there that is stealth.

The way the most jammers work is they figure out what frequency the antagonist radar is at (think of a radar frequency as just like a radio frequency, it is basically the station you are transmitting your radar at.) Then they send out a stronger signal of noise (gibberish) at a higher magnitude or power than your radio to drown out your information from the radar. It is like when you are watching TV when someone is vacuuming in your house to cover up the sound of the vacuum you increase the volume of the TV. Think about jamming as turning up the volume on their signal to try to disrupt it except instead of sound radars use radio signals.



 

 

To counter jamming, the radar can change its frequency like constantly changing stations so the jammer cannot block the message. Like when you change stations when the static for one station gets too bad.

Radar jamming is one type of way to prevent a radar from figuring where you are and how fast you are going. Electronic attack (EA) is what jamming and other ways to prevent a radar from working are called. The list of things that can be done to prevent EA is called electronic counter-measures (ECM). I will talk about more of these in another blog.

You Have an AM/FM Radio, What's TM?

Tucson company TM Technologies Inc announced they have demonstrated a new way to transmit information in electromagnetic waves according to announcements in the Arizona Daily Star and TVTechnology.

For me, this is a local interest story since I live and work in Tucson; however, this technology has the potential to revolutionize communication. According to their press release they demonstrated a new way to modulate electromagnetic waves which they call transpositional modulation (TM). A TM wave is opposed to a frequency modulation or amplitude modulation which you may recognize as FM and AM on your car radio.

FM is a way transmit information by changing the frequency (the number of wavelengths/second that travel across a particular point) of the wave. AM transmits data by increasing the power of the transmission wave. There is a third way of transmitting data called phase modulation (PM) which changes the phase of wave, basically by shifting the phase from a sine to cosine. PM and FM are very similar because FM is the derivative of PM and many people consider PM and FM to be the same. 

Top: base signal
Middle: base signal with AM to carry signal
Bottom: base signal with FM to carry signal

When the receiver (like the radio in my truck) receives the signal it measures the changes in the wave and converts that data into a usable medium (like music in your car speakers).

What TM does is change the same of the wave by adding indentions to the wave. By doing this extra information can be stored in the waves. In the Daily Star, the chief scientist for the Medusa Scientific, the parent company of TM Technologies, Rick Gerdes said, “At the least, the technology allows at least double the throughput, but in some cases four or five times more, and in some extreme cases 30 times more data.” This method would be a fourth way to transmit data on a wave.

I am interested to see how this would work, since detecting minor changes to the sinusoidal signal would have to be detected by the receiver by separating it from the noise and the loss of information when the wave is quantized or converted from analog to digital. On top of separating the signal from the noise, the receiver needs to sample the data at a high enough rate to detect the indentations in the signal wave. In September, they plan on doing a transmission of a UHD signal, which is a fairly low frequency UHF signal which is between 400 – 800 MHz (for comparison WiFi transmits at 5GHz).

TVTechnology lists two patents pending for Mr. Gerdes,(Patent #5200715 –Waveform modulation and demodulation methods and apparatus and Patent #5327237– Transmitting data with video) that they believe the technology associated from the technology is derived from. If this totally is a killer app, I would expect most of the technology would be kept as a trade secret. 

Predicting Future Locations Paper by Salilek and Krumm

 “Far Out: Predicting Long-Term Human Mobility” by Adam Sadilek and John Krumm is an interesting paper that says you can predict long-term were some will be. There is a good non-technical summary by Camille Sweeney and Josh Gosfield of Fast Company.

The idea is that people move in patterns, and you can predict where someone will be in the future based on where they are now. The authors recorded the movements of 703 subjects (307 people and 396 vehicles) from 7 to 1247 contiguous days with the average number being 45.9 days and a standard deviation of 117.8  days (I’m guessing the 1247 was one of the authors.) They had 33,268 days of location data.

They used Fourier analysis to find the periodicities in movements of the subjects and used singular value decomposition (SVD), a type of principle component analysis (PCA), to reduce the dimensionality of the data and to form predictive weights.

They broke the surface of the globe into triangular cells to make the locations and movements more finite, and broke up the day into finite blocks as well. The authors formed the data by breaking up the probabilities that a subject would be at 11 particular locations by 24 hour blocks and by days with a separate block for holidays.

Using the past data the authors, formed the predictive models to predict the locations of people up to 80 weeks in the future. The results were above 80% accurate and better than their baseline.

Normally, I do not like using PCA for dimensionality reduction because the top PCA (aka eigen) features may not be the features you need for your modeling goals. For the radar automatic target recognition work, my team and I picked features manually (using stuff like length and width of targets) to identify them because we knew the size of the vehicles and physical characteristics we were looking for, but with huge data sets where you may not know what features are best, PCA could be a better way to go. I have also used discrete cosine transform (DST) for really large 3D volumes because my PC did not have enough RAM to handle the matrix transforms of PCA.

I was skeptical of this paper when I first looked at it because I did not think people’s movement were that regular to be good for prediction, but with their accuracy people are more predictable then I thought. When you consider the accuracy values you also have to consider most people sleep 6 to 9 hours a night or 42 to 63 hours a week and work 8 hours a day or 40 hours a week and those schedules are fairly regular. So, the authors really need to account for the approximately 72 hours in a week. 

A system like this has both applications for marketing and demographics as well as security and defense applications.