Artificial intelligence: not there yet

NASA didn’t reinvent MATH to put a man on the moon. The math was already known. How to use the math and equations was unknown. The same with autonomous vehicles. The math is there…it’s the programming and using that math that’s not there. You seem to be just quoting phrases from a google search without really understanding what you’re talking about.

bravo, Mike, thank you!
you really made the point!!

I’m not debating that 1 + 2 still equals 3, just to agree on the basics, ok?

before the coders will start to busily bang on their keyboards, somebody has to invent “the math” or “algorithms” which these millions of lines of code programmers would have to implement (with many bugs and shortcuts in the middle too, BTW)

did you note the quotes on “the math” ?
this is to make sure you do not go back to pre-K grade argument on arithmetics

do you have anything to say on the substance although?

you seem to be much more “advanced” in critique of the messenger than on saying anything meaningful on the neural networks topic

what fundamental research do you think will overcome the current shortcomings, which I think are still fundamental?

It’s to debate total nonsense. You talk in extreme generalities that you read in a google page. You talk about computer software like you know anything about it and yet it’s obvious to someone who’s been working as a software engineer for decades that you don’t.

Good Luck. You live in your reality. I’ll live in mine.

2 Likes

Fifty years ago when I was doing graduate work in applied statistics, my dissertation advisor, who was close to retirement, told me that the future would be in artificial intelligence. About 18 years later, I attended a week long training session on a military base on neural networks. I remember that one problem that was presented was to train the neural network to classify the target image picked up by infrared devices as either a tank, a truck, or a Jeep.
I realized that from less than perfect images from the infrared inputs, the output was the classification of the target as the tank, truck, or Jeep.
I was involved in predicting the success of students in the beginning computer science course. I had inputs of SAT scores, high school grade point average, high school class rank, grades in the prerequisite calculus courses, and gender. I had been using a logistic regression statistical model to predict whether or not a student would be successful. I compared the results with a back propagation neural network model. The neural network model was much more accurate in classifying the success or failure of the students in the computer science course. The downside of the neural network model was the time it took to “train” the network. The computer with the Intel 286 chip that was available to me at the time took almost a week running 24 hours a day on only that task before the network converged on a solution. The next time I did the experiment, I had a computer with the Intel 486 processor chip and the network converged in about half an hour.
Before I retired ten years ago, the neural network was recognized as a valuable statistics tool. My advisor from 50.years ago really saw into the future. If I were starting my career today, it would be in artificial intelligence.

Besides my AI class with Lisp, I’ve only done rudimentary work with AI. It’s a field you have to dedicate yourself to for many years to be proficient at it.

My early work on AI was on Vax/VMS systems. Tough to put one of those in a car…maybe a large bus with a real good AC unit. The biggest problem with early AI was processor speed. One of my classes at SU with Lisp was creating a program for Chess. It could actually play fairly well, but took hours to analyze all the possible moves - Even on the IBM 370 mainframe.

But with faster and smaller chips (processers these days run over 1,000,000 times faster then that state-of-the-art IBM 370 mainframe from the mid 70’s. That same chess program now takes less then 1 second to do the same analysis. AI has made huge strides.

Driverless Autonomous Semi’s are on the road TODAY. About 2,000 autonomous vehicles with a backup driver are on the road. Over 5,000 worldwide. There are over 54 Million vehicles with some autonomous features (lane assist, backup assist, adaptive cruise control…etc…etc) on the road today.

And I personally really have come to appreciate the advancement in technology. Hoping our 2017s make it another ten years and looking forward to the next car where I can plug in a destination and take a nap!

@MikeInNH “Besides my AI class with Lisp, I’ve only done rudimentary work with AI. It’s a field you have to dedicate yourself to for many years to be proficient at it”.
You are one up on me in AI. I had no course in AI. Such courses didn’t exist when I was working on my degrees. I was doing some work in back propagation neural networks and finally was given a load reduction for one semester to do research. Unfortunately, three weeks into the semester, one of my colleagues died. He was teaching a graduate course in computer simulation. There were about 15 students in the class. I took over the class so that the class wouldn’t be dropped and the students wouldn’t lose credit hours. I had to learn visual SLAM over a weekend. When I retired back in 2011, it was in.my 10 year plan to continue research in neural networks. Ten years have past and I have had too much fun doing other things.
AI is an exciting field with mind boggling applications. The self driving vehicles is just one of many applications.

I was drafted in the Army right out of High-School and spent 5 months in Nam before they started pulling us out. I then spent another 1.5 years in the Army to get better GI education benefits. So when I started college very few colleges offered degrees in Computer Science. Either it was an EE major with classes in Computer Engineering or a Math Major with classes in Computer Science. But even then they offered AI classes in Lisp. Half way through my degree at SU (as a part time student) they started offering a degree in Computer Science so I switched from a Math major to CS major. But I got my MS in Applied Math at BU later in life.

One guy I worked with at Digital did his Doctoral Thesis at Brown on AI (as applied to Computer Networks). I read his paper and barely understood it.

I’ve heard of vSLAM…but never worked with it.

As I said earlier…if I was just starting out as an engineer I’d be applying to one of the Autonomous driving technology companies here in the Boston area. Sounds like a real fun field to be in right now.

@MikeInNH I would bet your degree in applied mathematics was really an asset as you studied computer science.
I would still like to take courses in AI. However, I am almost 80 years old and though I think I still have all my marbles, they roll too slowly for me to compete with the younger generation of computer scientists. I am happy to sit back and witness all the new applications of AI.

The success of NASA’s tiny Ingenuity helicopter on Mars is a clear demonstration of how AI can work to perfection. Because it would take 11 minutes for course correction signals to reach that craft from the Earth, it had to contain AI which allowed it to avoid obstacles and to distinguish the best place to land.

It worked perfectly.

I watched that on 60 minutes yesterday. Even an experience flyer couldn’t manually fly the remote on earth in a chamber that had the same Atmosphere pressure as Mars…

Reminds me of a couple of speeding jokes:

One time we were driving through a construction zone and the sign said, SPEED LIMIT 35 AHEAD. And, since there were four of us in the car, we were through there in no time.

And the other:

Police officer scans his radar and finds a car doing 22 MPH over the speed limit, and pulls them over. When he get to the car he is surprised to see it is a group of four very senior guys, maybe the youngest is 90…

To the driver he asks: “Do you know the speed limit along here?”
Driver: “Sure officer, it’s right there on the sign: 57.”
Officer: “Uh…no… that is the route number, not the speed limit. Here it is 35 mph.”

As he is writing up the ticket he glances around at the other passengers, and they all seem agitated and nervous. Being a bit suspicious now, he asks them “You fellows all OK in there?” And one pipes up: “Well we are now, but we had just come off from Route 109…”

6 Likes

We have an 895 in Maryland. I would not recommend triple digit speeds through the Baltimore Harbor Tunnel.

1 Like

Even 195 to BWI is a tad fast.

2 Likes

I live near both I-170 and Olive Blvd., also known as Route 340. (The maximum speed limit on Olive is 35 MPH, BTW.) I’d be in trouble if I tried substituting the route number for the speed limit in either case.

1 Like

I frequently drive on I-270 where many drivers seem to take it as a challenge to attempt a like speed to the designation.

One recent drive on I-270 between Oakville and Page Ave. I was keeping pace with traffic averaging 80-85 mph in the right hand “slow” lanes while those more ambitious zipped by on the left at easily close to 100 mph in some cases. Then a few of those idiots cut across four lanes with no signal from far left to exit right at the last minute. Made for an “interesting” drive up to St. Ann. :roll_eyes::flushed:

I imagine so. The general opinion frequently seems to be that whatever’s posted is a suggestion rather than a rule.

2 Likes

Kind of like stop signs and red lights :wink:

2 Likes