When I worked on autonomous vehicles I realized that if I really wanted to figure out how to get the vehicle to end up where I wanted, I had to model the actual processes involved. It wasn't actually that difficult. I literally just generated a set of curves that would start with the wheels of a vehicle pointed straight ahead, and tracing the path the car would follow if the steering wheel started being turned at a reasonable rate to a maximum fixed angle, holding it there for some amount of time, then straightening the wheel again. This curve wasn't a simple shape at all, but it didn't matter. It gave the most accurate results. Once the family of curves was generated, I could just select the one that best fit the situation.
The unstructured input attack surface problem is indeed troublesome. AI right now is a bit gullible, but as systems evolve they will become more robust. However, even humans are vulnerable to the input given to us.
We might be speed running memetic warfare here.
The Monty Python skit about the deadly joke might be more realistic than I thought. Defense against this deserves some serious contemplation.
The thing that blows my mind is: say you start filling the plane with pi. Pi has been proven to contain every finite sequence. That means that somewhere in the plane is a full physics simulation of YOU in the room you are in right now.
Does that you exist any less fully because its not currently in the memory of a computer being evaluated?
Depending on the infinite grid filling scheme even these properties may not be sufficient to guarantee that every two dimensional pattern is initially generated because the grid is two-dimensional, but the number property is "one-dimensional". A spiral pattern for example may always make it line up in a way such that certain 2d patterns are never generated.
Since it's not provable with pi, then we'd have to do a more circuitous proof of every finite pattern occurring. Inspired by Champernowne's constant, I propose a Pontifier Pattern that is simple, inefficient, but provably contains every finite pattern.
Starting at the origin, mark off rows of squares. the Nth row would contain NxN^2 squares of size n x n. Each square would be filled in left to right reading order with successive binary numbers with the most significant digit at the top left.
Somewhere in that pattern is the physics simulation of you reading this comment :)
Yes, sounds like it! Though I'm thinking that the relative arrangement of patterns would also make a difference. I wonder if such a thing as "all (infinitely many) possible arrangements of all patterns" can exist
My mother is currently dying in the hospital with breathing problems. I mentioned this to her earlier today... I thought this would have been much farther along than it is. Hurry up with the medical tech already!
I remember reading somewhere that because Ternary computing is inherently reversible, that from an information theoretic point of view that ternary computations have a lower theoretical bound on energy usage, and as such could be a way to bypass heat dissipation problems in chips built with ultra-high density, large size, and high computational load.
I wasn't knowledgeable enough to evaluate that claim at the time, and I'm still not.
I recently saw another video about a high accuracy 3d positioning stage. The differences and similarities were very interesting. For instance, both used rigid rods with ball joints for accuracy, but wildly different encoders and testing methods.
reply