New Artificial Intelligence System Beats Human In Aerial Combat Simulation

[BON PLAN]

Yup. I think so. They need to develop a fighter for that. That would take most of the time. Integrating the fighter parts with the software will also take a long time plus a lot of tests.

It's easyer thant to developp a fighter with a himan pilot : no support of a pilot = less weight and space. No G limit to 9G. less drag because no canopy. A single motor is less problematic : in case of a failure no risk your pilot be captured or killed by ennemy...

Main problem : safety in no war operation. If your computered plane shoot down a liner because some lines of the software are corrupted.....

 

[Kshatriya87]

It's easyer thant to developp a fighter with a himan pilot : no support of a pilot = less weight and space. No G limit to 9G. less drag because no canopy. A single motor is less problematic : in case of a failure no risk your pilot be captured or killed by ennemy...

Main problem : safety in no war operation. If your computered plane shoot down a liner because some lines of the software are corrupted.....

The problem can be sorted out easily. The code/program should be carefully broken up in different configurations for different exercises. Before flying the mission, the fighter AI should be given authorization to execute only the necessary part of the program. Say - Execute Code 1 or Code 2 etc. A firewall can be put in place so that the AI does not execute other codes.

Anyways, fail safe switch should always be there like I said before which either shuts down the AI or gives the human pilot at some base the control of the fighter.

 

[DingDong]

Human Brain is not a computer, it is something else (still not well understood). Current computing models have serious limitations which don't effect the Human Brain.

Computer Models like Artificial Neural Network, Fuzzy Logic, Genetic Algorithm are much simplified versions of the corresponding real-world observations (read my point about the Simulated Reality).

An intelligent being cannot create an agent which is more intelligent than himself.
1. If we can create something with equal intelligence like ours then that means we understand with 100% accuracy how our brain functions.
2. If we cannot create something which has equal intelligence like ours, we cannot create an higher intelligence either.

The Simulated Reality created by an artificial being will be much simpler compared to the real world in which the creator lives.
An artificial universe like Matrix created by the Human beings will have much simpler physical laws governing it than our universe, for two reasons:
1. Our universe cannot contain something (even a simulation) which is more complex than itself.
2. We do not understand all the physical laws correctly and we never will.

 

[Kshatriya87]

Human Brain is not a computer, it is something else (still not well understood). Current computing models have serious limitations which don't effect the Human Brain.

Computer Models like Artificial Neural Network, Fuzzy Logic, Genetic Algorithm are much simplified versions of the corresponding real-world observations (read my point about the Simulated Reality).

An intelligent being cannot create an agent which is more intelligent than himself.
1. If we can create something with equal intelligence like ours then that means we understand with 100% accuracy how our brain functions.
2. If we cannot create something which has equal intelligence like ours, we cannot create an higher intelligence either.

The Simulated Reality created by an artificial being will be much simpler compared to the real world in which the creator lives.
An artificial universe like Matrix created by the Human beings will have much simpler physical laws governing it than our universe, for two reasons:
1. Our universe cannot contain something (even a simulation) which is more complex than itself.
2. We do not understand all the physical laws correctly and we never will.

No. But a human brain can develop something that is smart, can learn, develop and become smarter.

E.g. Human brain itself. It is not that developed when it is younger. As a person learns, brain develops. It calculates, reacts accordingly. An AI can similarly be developed which can carry on learning and growing.

 

[Bharat Ek Khoj]

Interesting debate. Keep it coming !!!

 

[BON PLAN]

An intelligent being cannot create an agent which is more intelligent than himself.

The new software (as the famous which beat GO world champion) are able to learn from them self. "game" after game they become stronger.
Even among humans, some are more intelligent than some others. Why a computer, with huge calculation power can't beat a human? I'm sure it is or it will be. But I'm not able to prove it (as anybody else in the other side !)

 

[BON PLAN]

I don't see AI able to write poesy, or a Shakespear like theatre.
But a war machine, where speed is of the most importance.... We are smashed i'm afraid.

 

[ezsasa]

How about incorporating the AI into a electro-mechanical robot which also does the manual functions of a fighter pilot ?

Advantage : interoperability of the fighter jet between humans and robot. And no need to developed a whole new aircraft to fit a pilot less AI.

I know i am getting into terminator territory, but still let's see where it goes.

 

[Razor]

0. An intelligent being cannot create an agent which is more intelligent than himself.
1. If we can create something with equal intelligence like ours then that means we understand with 100% accuracy how our brain functions.
2. If we cannot create something which has equal intelligence like ours, we cannot create an higher intelligence either.

I am not sure intelligence is that well defined that we can quantify it. So terms like "more intelligent", may infact be meaningless, at least as of now. Besides, intelligence is not necessarily a requirement for survival. Luck & adaptation are more important factors.

0. But if we take the general sense of the idea, "more intelligent" I'm sure you'd agree Einstein's father isn't "more intelligent" than Einstein.
1. Einstein's father was not able to understand Einstein with 100% accuracy, but he did make him, using certain techniques.
2. {0}+{1}={2}

The Simulated Reality created by an artificial being will be much simpler compared to the real world in which the creator lives.
An artificial universe like Matrix created by the Human beings will have much simpler physical laws governing it than our universe, for two reasons:
1. Our universe cannot contain something (even a simulation) which is more complex than itself.
2. We do not understand all the physical laws correctly and we never will.

Okay.

 

[DingDong]

The new software (as the famous which beat GO world champion) are able to learn from them self. "game" after game they become stronger.
Even among humans, some are more intelligent than some others. Why a computer, with huge calculation power can't beat a human? I'm sure it is or it will be. But I'm not able to prove it (as anybody else in the other side !)

Read "Automata" and "Computability" theories these are basic Computer Science courses which tell us "what a computer can do and what it can't". The problems known and unknown to human can be classified under three categories based on time and effort it takes to compute them:
A. Polynomial time
B. Intractable
C. Not computable

Only problems of types (A) and (B) can be computed by the man made computers.

Not going into theoretical treatment, all the programs and software we deal with in daily life are built to solve problems of type (A).

Intractable problems are those which can be computed theoretically but require massive (may grow to infinite) computing power, e.g. breaking an AES 256 encrypted message is equivalent to solving an intractable problem.

But, could you imagine that there are practical problems known to us human beings are actually not computable? It doesn't even matter how fast or slow or accurate a computer is.

We have built rockets which can fly and reach outer-space while we can't. But we can't deny the fact that Human body is considerably more complex compared to a rocket, that is why we can create rockets but we cannot create a human from basic chemicals.

Just because computers can bean humans in certain tasks doesn't mean that computers are more powerful. Human brain can do what a computer can and much more.

 

[Bharat Ek Khoj]

Why a computer, with huge calculation power can't beat a human? I'm sure it is or it will be.

When it does, just unplug it

 

[A chauhan]

Curiosity is the key, if you manage to make machine "self curious" to know and learn new things, you'll succeed in AI.

 

[Kshatriya87]

How about incorporating the AI into a electro-mechanical robot which also does the manual functions of a fighter pilot ?

Advantage : interoperability of the fighter jet between humans and robot. And no need to developed a whole new aircraft to fit a pilot less AI.

I know i am getting into terminator territory, but still let's see where it goes.

I'm not sure about that. A robots movements are still not as agile as humans. If you develop a robot who works, walks, acts exactly or better than a human, you have achieved one of the ultimate goals of AI. Not to mention they would cost you almost as much as a fighter plane.

Sent from my Nexus 5X using Tapatalk

 

[Kshatriya87]

Read "Automata" and "Computability" theories these are basic Computer Science courses which tell us "what a computer can do and what it can't". The problems known and unknown to human can be classified under three categories based on time and effort it takes to compute them:
A. Polynomial time
B. Intractable
C. Not computable

Only problems of types (A) and (B) can be computed by the man made computers.

Not going into theoretical treatment, all the programs and software we deal with in daily life are built to solve problems of type (A).

Intractable problems are those which can be computed theoretically but require massive (may grow to infinite) computing power, e.g. breaking an AES 256 encrypted message is equivalent to solving an intractable problem.

But, could you imagine that there are practical problems known to us human beings are actually not computable? It doesn't even matter how fast or slow or accurate a computer is.

We have built rockets which can fly and reach outer-space while we can't. But we can't deny the fact that Human body is considerably more complex compared to a rocket, that is why we can create rockets but we cannot create a human from basic chemicals.

Just because computers can bean humans in certain tasks doesn't mean that computers are more powerful. Human brain can do what a computer can and much more.

No one is denying that human body is complex. But we are talking about an AI here. That's just means the brain.

As you said, there are some problems which are not computable. You can teach the AI for that as well and hence make it more intelligent than a single human brain.

Example. An AI can have advanced computing for all computable problems in life. For all non computing problems, the answer is collective consciousness. You can feed in the experiences, decisions, way of thinking and decision making, perspective etc. Aspects of a number of intelligent and good human beings.

Like I can agree that a human cannot build something that is more intelligent than him if he uses only his brain to do it. But what if he uses more? Isn't that why people collaborate to make things?

You can feed collective conciousness of people like Einstein, Stephen hawking, Abdul kalam etc. In the AI and that will make it more intelligent than yourself.

Sent from my Nexus 5X using Tapatalk

 

[ezsasa]

I'm not sure about that. A robots movements are still not as agile as humans. If you develop a robot who works, walks, acts exactly or better than a human, you have achieved one of the ultimate goals of AI. Not to mention they would cost you almost as much as a fighter plane.

Sent from my Nexus 5X using Tapatalk

This particular robot we are talking need not do many of the things that humans need to do(example walking). It just needs to replicate the manual functions that a pilot is required to do in the cockpit.

More importantly it need not even look human.

Fundamentally it is required to do the following manual functions:

1) Interpret the multiple cockpit displays for AI to process.
2) Handle the stick
3) Handle foot pedals(if any).

Expectations from AI can be as follows:
1) Understand and implement mission objectives
2) Process visual data and give feedback to manual controls
3) Capability to have multiple redundancies both for logical controls and for continuous operation.
4) Ability to anticipate and react to change in threat perception during the course of the mission.

 

[ezsasa]

No one is denying that human body is complex. But we are talking about an AI here. That's just means the brain.

As you said, there are some problems which are not computable. You can teach the AI for that as well and hence make it more intelligent than a single human brain.

Example. An AI can have advanced computing for all computable problems in life. For all non computing problems, the answer is collective consciousness. You can feed in the experiences, decisions, way of thinking and decision making, perspective etc. Aspects of a number of intelligent and good human beings.

Like I can agree that a human cannot build something that is more intelligent than him if he uses only his brain to do it. But what if he uses more? Isn't that why people collaborate to make things?

You can feed collective conciousness of people like Einstein, Stephen hawking, Abdul kalam etc. In the AI and that will make it more intelligent than yourself.

Sent from my Nexus 5X using Tapatalk

We also need to make a assumption that a pilot AI need not have intelligence, it needs to pick a best possible solution to a problem from a library of solutions pre-fed into the AI's database in least amount of reactive time.

The problems that are presented in the flight envelope to a pilot will definitely not be a infinite number. even a million possible scenarios will not be a big number for a good AI.

Have a look at this site, https://www.wolframalpha.com/.
This is not AI but a logical engine which gives answers from a pre-fed library of solutions. you can ask many things within the scope of its functionality.

 

[Screambowl]

No. But a human brain can develop something that is smart, can learn, develop and become smarter.

E.g. Human brain itself. It is not that developed when it is younger. As a person learns, brain develops. It calculates, reacts accordingly. An AI can similarly be developed which can carry on learning and growing.

because human brain has a mind , AI will just have a brain without mind.

 

[AnantS]

The new software (as the famous which beat GO world champion) are able to learn from them self. "game" after game they become stronger.
Even among humans, some are more intelligent than some others. Why a computer, with huge calculation power can't beat a human? I'm sure it is or it will be. But I'm not able to prove it (as anybody else in the other side !)

If defeating computer is your aim, start learning/practice with single focus of defeating it. Eventually you too will learn to defeat it.

You do have some people in world who calculate faster than calculator, do you? Now let ,me interject Hindu POV. *You* are much more powerful than *Your* brain and outlast your bodies. Question is who are *You*?

 

[AnantS]

because human brain has a mind , AI will just have a brain without mind.

Question is who created AI? Its Human mind isnt it? Hasn't the logic been designed by humans?

 

[DingDong]

No one is denying that human body is complex. But we are talking about an AI here. That's just means the brain.

As you said, there are some problems which are not computable. You can teach the AI for that as well and hence make it more intelligent than a single human brain.

Example. An AI can have advanced computing for all computable problems in life. For all non computing problems, the answer is collective consciousness. You can feed in the experiences, decisions, way of thinking and decision making, perspective etc. Aspects of a number of intelligent and good human beings.

Like I can agree that a human cannot build something that is more intelligent than him if he uses only his brain to do it. But what if he uses more? Isn't that why people collaborate to make things?

You can feed collective conciousness of people like Einstein, Stephen hawking, Abdul kalam etc. In the AI and that will make it more intelligent than yourself.

Sent from my Nexus 5X using Tapatalk

An AI running on a computer is bound by the rules and limitations of computing. Biological brain is not a computer, it has been wired differently and hence is not bound by the limitations of a computing device. Forget about Human, an AI running on a computer cannot even match the brain of a cockroach.

It is almost unlikely that the computers as we know them will ever become conscious because of their inherent limitations. This is another matter that computers can do faster and accurate calculations compared to a Human brain, but then the biological brain isn't designed for this purpose.

To create something which is capable of matching Human Intelligence, we will first need to create something which may mimic a Human Brain.

because human brain has a mind , AI will just have a brain without mind.

You reminded me of this:
https://en.wikipedia.org/wiki/Theory_of_mind
https://en.wikipedia.org/wiki/Theory_of_mind_in_animals

Not only the AI but also many animals lack the "mind". Many AI experts have postulated that an AI created by the Human will not be able to grow a mind of it's own.