AI...A Warning from Elon Musk

louis · 2021-08-28 18:27:01

An excellent video about Musk's views on AI and perhaps an explanation of why he is so keen to relocate humanity to Mars. Maybe doing that will give us a short breathing space.

https://www.youtube.com/watch?v=AQNBO6WbQo8

SpaceNut · 2021-08-28 19:25:50

Artificial intelligence but we know its got to be broken to make it work

louis · 2021-08-28 20:33:56

That makes no sense unless you want to explain.

Musk's essential point is that AI is an independent force from humans...unless humans actively constrain it in many, many ways it will be independent of us and do things it wants to do, not things we want it to do.

SpaceNut wrote:

Artificial intelligence but we know its got to be broken to make it work

SpaceNut · 2021-08-28 20:45:31

A computer program does not think for itself it continues step by step through code until it gets to the next decision point in the coding.

Breaking the code means that it must come up with its own decision point which makes it think as its no longer just moving step by step with in the coding instruction.

tahanson43206 · 2021-08-29 00:54:43

For SpaceNut ... the step-by-step computer code you're describing is NOT what AI is about.

If you have time (and I realize you may not) there are resources available to help you begin to understand the rudiments of neural networks and related branches of AI ...

You are certainly right to remember that many decades ago, AI research started out with the idea that step-by-step programming might be how AI would come into being. Tremendous achievements have been made with the step-by-step approach, but modern AI development is stretching beyond that basic level.

(th)

SpaceNut · 2021-08-29 08:25:13

You could think of neuro nets as the subroutines of computer code that is in pools awaiting use but with computers the addressing of those commands are as limited as to the computers instruction set that it is using.
The instruction set can not be modified in hardware but can be created by new subroutines. Its the creation of new subroutines that means we are now artificial as those command sets are learned and acquired for use as they are not written by a programmer but by the machine.

louis · 2021-08-29 08:49:47

Thanks for the explanation.

I have never believed computers, even the most advanced, can "think" for themselves. They aren't conscious. As John Searle, the philosopher, reminded us long ago now with his "Chinese code" theory.

But they can be provided with enough coding to make decisions independently of human beings (or situation-specific code) and yet make good decisions. This is precisely what a driverless car does. The car's computer is combining codes to make decisions, but without situation-specific code. Perhaps you agree in your comments about decision points.

The thing about the decision points is that the computer can store decisions and their outcomes and make decisions about whether they would repeat or modify their actions, given the outcomes.

SpaceNut wrote:

A computer program does not think for itself it continues step by step through code until it gets to the next decision point in the coding.
Breaking the code means that it must come up with its own decision point which makes it think as its no longer just moving step by step with in the coding instruction.

SpaceNut · 2021-08-29 18:16:31

Sensors are a hardware subroutine is started and ends when the last command is executed based on sensor input.

Thats a higher level when there is no routine for it to execute from that sensor input which it must come up with but thats not really choice as its got no sense of right or wrong out comes for the decision that it might make.

Learned out come is statistics and not really learning as its not trying to right a correction routine to control the out come.

tahanson43206 · 2021-08-29 19:45:02

For Louis ... modern neural networks have nothing to do with step-by-step programming that originated in the dawn of the computer age.

Here is a web site that offers a simplified explanation of the technology, and gives a bit of history.

https://www.freecodecamp.org/news/deep- … n-english/

(th)

SpaceNut · 2021-08-30 19:36:56

What makes the net is the hardwired links between each of the routines that are running on seperate processing units where the same input is supplied to multiple computing where the output is directed to the end result computers that make the actions happen.

louis · 2021-08-30 19:49:27

I never used the phrase "step by step programming" so I don't know why you're suggesting I did.

I don't think you understand the concept. The key word in the article is algorithms. It's all algorithms = code. Yes there's more subtle interaction between calculation points, but it's still all code. No "neural" network would work without code.

tahanson43206 wrote:

For Louis ... modern neural networks have nothing to do with step-by-step programming that originated in the dawn of the computer age.
Here is a web site that offers a simplified explanation of the technology, and gives a bit of history.
https://www.freecodecamp.org/news/deep- … n-english/
(th)

kbd512 · 2021-08-30 20:25:46

Every computer program, no matter how sophisticated, is ultimately constrained by its ability to evaluate input and produce a meaningful output- either to the program itself, or to the person who wrote the program.

New Mars Forums

Announcement

#1 2021-08-28 18:27:01

AI...A Warning from Elon Musk

#2 2021-08-28 19:25:50

Re: AI...A Warning from Elon Musk

#3 2021-08-28 20:33:56

Re: AI...A Warning from Elon Musk

#4 2021-08-28 20:45:31

Re: AI...A Warning from Elon Musk

#5 2021-08-29 00:54:43

Re: AI...A Warning from Elon Musk

#6 2021-08-29 08:25:13

Re: AI...A Warning from Elon Musk

#7 2021-08-29 08:49:47

Re: AI...A Warning from Elon Musk

#8 2021-08-29 18:16:31

Re: AI...A Warning from Elon Musk

#9 2021-08-29 19:45:02

Re: AI...A Warning from Elon Musk

#10 2021-08-30 19:36:56

Re: AI...A Warning from Elon Musk

#11 2021-08-30 19:49:27

Re: AI...A Warning from Elon Musk

#12 2021-08-30 20:25:46

Re: AI...A Warning from Elon Musk

Board footer