An AI-run fighter jet went up against another controlled by a human pilot in a drill, the US has said. The aircraft flew at up to speeds of 1,200mph during combat that's often referred to as a dogfight.
The aircraft flew up to speeds of 1,200mph. DARPA did not reveal which aircraft won the dogfight.
Those of us who play video games do at least. All the AI difficulty settings are arbitrary. You give the bot the ability to use its full capability, and the game is unplayable.
In video games the AI have access to all the data in the game. In real life both the human and AI have access to the same (maybe imprecise) sensor data. There are also physical limitations in the real world. I don’t think it’s the same scenario.
Not exactly, AI would be able to interpret sensor data in a more complete and thorough way. A person can only take in so much information at once - AI not so limited.
Don’t get me wrong. Humans have many limitations that AI don’t in this scenario. I’m not saying that a human would do better. For example, as others have stated, an AI doesn’t suffer from G forces like a human does. AI also reads the raw sensor data instead of a screen.
All I’m saying that this case is not the same as a videogame.
Video games can model point of view and limit AI to what they can legitimately see, while still taking the governor chip off their aiming and reaction time performance.
Ai can balance a physical triple pendulum and move between positions fluidly just using vision alone, a human has no chance at coming close.
We’re genuinely in the sci-fi robots are better than humans phase of history, by 2030 you’ll be used to seeing impressive things done by robots like dude perfect videos with people setting up crazy challenges like ‘I got my robot to throw THIS egg through THIS obstacle course and you’ll never belive how it did it!’
I think even the imperfect sensor data is enough to beat a human. My main argument for why self-driving cars will eventually be objectively safer than the best human drivers (no comment about whether that point has already done) is this:
A human can only look at one thing at a time. Compared to a computer, we see allow, think slow, react show, move slow. A computer can look in all directions all the time, and react to danger coming from any of those directions faster than a human driver would even if they were lucky enough to be looking in the right direction. Add to that the fact that they can take in much more sensor data that isn’t available to the driver or take away from precious looking-at-the-road time for the driver to know, such as wind resistance, engine RPM, or what have you (I’m actually not a car guy so my examples aren’t the best). Bottom line: the AI has a direct connection to more data, can take more of it in at once and make faster decisions based on all of it. It’s inherently better. The “only” hurdles are making it actually interpret its sensors effectively (i.e. understand what cameras are seeing) and make good decisions based on this data. We can argue about how well either of those are in the current state of the technology, but IMO they’re both good enough today to massively outperform a human in most scenarios.
All of this applies to an AI plane as well. So my money is on the AI.
For sure without humans the AI probably wins, assuming the instruments are good. This wasn’t without humans, but it probably still wins.
I’m fairly certain most dogfights happen on instruments only at this point, so I don’t see a chance the human won. The AI can react faster and more aggressively. It can also almost perfectly match a G-load profile limit (which could be much higher without humans on board) where a human needs to stay a little under to not do damage.
This is all assuming the data it was given was good and comprehensive, which I’m sure it was. It also likely trained in a simulation a lot too. This is one of those things AI is great for. Anything that requires doing something new and unique it can’t handle, but if it just requires executing an output based on inputs, that’s a perfect use case.
Electrogravitics seem like a conspiracy theory. Unless they’ve been around as long as human centrifuges, which DO simulate g-forces, I doubt that they’d be more economical even if they do exist.
There is a connection between gravity and electromagnetics, but it’s mostly through the stress-energy tensor giving photons momentum (and thus gravitational pull) but to use an EM field to measurable gravity you need absolutely insane amounts of energy.
You essentially need the literal inverse of a supermassive nuclear explosion (almost like a small star), because the gravitational effect of energy is equivalent to the gravitational effect of the mass which it would form if bound, and given E=mc^2 and the fact that nuclear bombs are small enough to barely have measurable gravity then the math means you need truly insane amounts of energy. (unless somebody can figure out a cheat to create directional pull with much less energy, but I strongly doubt it)
It’s more plausible that somebody would be able to scale up “optical tweezers” to move large masses (directly depositing momentum of the energy field on an object) because that no longer involves the E=mc^2 equation, but it would be even more complicated by a HUGE factor than building the type of large supercooled electromagnets which already can make humans hover (due to water in the body being diamagnetic)
We all know which aircraft won the fight.
Those of us who play video games do at least. All the AI difficulty settings are arbitrary. You give the bot the ability to use its full capability, and the game is unplayable.
In video games the AI have access to all the data in the game. In real life both the human and AI have access to the same (maybe imprecise) sensor data. There are also physical limitations in the real world. I don’t think it’s the same scenario.
Not exactly, AI would be able to interpret sensor data in a more complete and thorough way. A person can only take in so much information at once - AI not so limited.
Don’t get me wrong. Humans have many limitations that AI don’t in this scenario. I’m not saying that a human would do better. For example, as others have stated, an AI doesn’t suffer from G forces like a human does. AI also reads the raw sensor data instead of a screen.
All I’m saying that this case is not the same as a videogame.
Video games can model point of view and limit AI to what they can legitimately see, while still taking the governor chip off their aiming and reaction time performance.
While true, I wonder how many games actually do this.
Ai can balance a physical triple pendulum and move between positions fluidly just using vision alone, a human has no chance at coming close.
We’re genuinely in the sci-fi robots are better than humans phase of history, by 2030 you’ll be used to seeing impressive things done by robots like dude perfect videos with people setting up crazy challenges like ‘I got my robot to throw THIS egg through THIS obstacle course and you’ll never belive how it did it!’
I think even the imperfect sensor data is enough to beat a human. My main argument for why self-driving cars will eventually be objectively safer than the best human drivers (no comment about whether that point has already done) is this:
A human can only look at one thing at a time. Compared to a computer, we see allow, think slow, react show, move slow. A computer can look in all directions all the time, and react to danger coming from any of those directions faster than a human driver would even if they were lucky enough to be looking in the right direction. Add to that the fact that they can take in much more sensor data that isn’t available to the driver or take away from precious looking-at-the-road time for the driver to know, such as wind resistance, engine RPM, or what have you (I’m actually not a car guy so my examples aren’t the best). Bottom line: the AI has a direct connection to more data, can take more of it in at once and make faster decisions based on all of it. It’s inherently better. The “only” hurdles are making it actually interpret its sensors effectively (i.e. understand what cameras are seeing) and make good decisions based on this data. We can argue about how well either of those are in the current state of the technology, but IMO they’re both good enough today to massively outperform a human in most scenarios.
All of this applies to an AI plane as well. So my money is on the AI.
Plus they had humans on board the AI jet. I imagine it could pull some crazy insane Gs without the human pushing the engineering to the red line.
For sure without humans the AI probably wins, assuming the instruments are good. This wasn’t without humans, but it probably still wins.
I’m fairly certain most dogfights happen on instruments only at this point, so I don’t see a chance the human won. The AI can react faster and more aggressively. It can also almost perfectly match a G-load profile limit (which could be much higher without humans on board) where a human needs to stay a little under to not do damage.
This is all assuming the data it was given was good and comprehensive, which I’m sure it was. It also likely trained in a simulation a lot too. This is one of those things AI is great for. Anything that requires doing something new and unique it can’t handle, but if it just requires executing an output based on inputs, that’s a perfect use case.
Removed by mod
What if the pilot has a stroke?
Yeah but a camera lead falling out isn’t normal
What if we invent artificial gravity just so we can simulate pilot orientation and g forces while they sit still in a simulator?
We already have g-force simulation.
No we have g-force production. Until we release those electrogravitics from the top secret labs we can’t actually simulate g forces.
Electrogravitics seem like a conspiracy theory. Unless they’ve been around as long as human centrifuges, which DO simulate g-forces, I doubt that they’d be more economical even if they do exist.
There is a connection between gravity and electromagnetics, but it’s mostly through the stress-energy tensor giving photons momentum (and thus gravitational pull) but to use an EM field to measurable gravity you need absolutely insane amounts of energy.
You essentially need the literal inverse of a supermassive nuclear explosion (almost like a small star), because the gravitational effect of energy is equivalent to the gravitational effect of the mass which it would form if bound, and given E=mc^2 and the fact that nuclear bombs are small enough to barely have measurable gravity then the math means you need truly insane amounts of energy. (unless somebody can figure out a cheat to create directional pull with much less energy, but I strongly doubt it)
It’s more plausible that somebody would be able to scale up “optical tweezers” to move large masses (directly depositing momentum of the energy field on an object) because that no longer involves the E=mc^2 equation, but it would be even more complicated by a HUGE factor than building the type of large supercooled electromagnets which already can make humans hover (due to water in the body being diamagnetic)
No centrifuges create g-forces. The forces you feel in a centrifuge are actual g-forces.
Why do we need “authentic” g-forces to be “created”? As you’ve said, people already feel g-forces in centrifuges.
Removed by mod
Nonsense hypotheticals are not whataboutism.
Thank you