How close are we to having a J.A.R.V.I.S type AI???
- Thread starter Joerib
- Start date
LOL!
There's the question, which do you think will come first? True AI or fusion?
What's meant by "true AI"? And we already have fusion, it just takes more power than we get from it.
The technology for conversing with computers is already here, albeit imperfect. Things like Siri for iPhone show where things are headed. The most practical use for "AI" will be having your computer or phone assist you with various tasks.
I don't think the day is too far off when you could have a conversation like this:
"I need a flight from JFK to LAX on September third. Any time is fine."
"Please wait while I see what's available. There is a flight departing at 7:30AM and arriving at 10:51AM via jetBlue. The price is $1,057. Is this acceptable?"
"Yes, please book it."
"Your flight has been booked using the credit card on file."
Stuff like that is actually pretty trivial in terms of the language recognition part. The difficult part is mining large amounts of data to pick out what's most relevant to the user, but our techniques for that are always improving.
The technology for conversing with computers is already here, albeit imperfect. Things like Siri for iPhone show where things are headed. The most practical use for "AI" will be having your computer or phone assist you with various tasks.
I don't think the day is too far off when you could have a conversation like this:
"I need a flight from JFK to LAX on September third. Any time is fine."
"Please wait while I see what's available. There is a flight departing at 7:30AM and arriving at 10:51AM via jetBlue. The price is $1,057. Is this acceptable?"
"Yes, please book it."
"Your flight has been booked using the credit card on file."
Stuff like that is actually pretty trivial in terms of the language recognition part. The difficult part is mining large amounts of data to pick out what's most relevant to the user, but our techniques for that are always improving.
True AI. A machine aware of itself, that actually "thinks", rather than merely calculates. I've got no doubt we will come up with machines that are more and more efficient sorters of information. But that's the easy part. Now, A.I. as depicted in Iron Man? It'll come, I'm pretty certain, and maybe even sooner than we think.
And yeah, I would say until we can actuall USE fusion power, we don't really have it. Not being cute, but that's kinda the point.
And yeah, I would say until we can actuall USE fusion power, we don't really have it. Not being cute, but that's kinda the point.
Last edited:
I know that we will discover true artificial intelligence one day, but I think it will be only after we discover some true natural intelligence first.
Do we really want to have "artificial" intelligences? What purpose would they serve? What's the point in having a self-aware machine? Would it be more efficient, or useful?
Something like Siri, as Maxwell House pointed out, is definitely the precursor to many forthcoming pseudo-intelligent programs that will assist us, but I don't think making something like Siri completely self-aware and independently intelligent is necessary or useful.
Something like Siri, as Maxwell House pointed out, is definitely the precursor to many forthcoming pseudo-intelligent programs that will assist us, but I don't think making something like Siri completely self-aware and independently intelligent is necessary or useful.
I don't think we are anywhere near having "True AI," as in a self-aware computer. We don't even know what makes biological lifeforms become self-aware, much less how to simulate it on a computer.
I read an interesting article on io9, which proposed a rather optimistic timetable of the future. One thing it did posit was the development of true AI. And of course, they made military robots out of some of them. Two nations got into conflict with each other, and the first units that met on the battlefield were groups of these AI warbots. They looked each other over, then pretty much decided not to fight. They said, "This is illogical. Fuck off." The AIs en masse pretty much segregated themselves from humanity, built their own spaceflight capability and launched themselves off into the solar system to settle down and build their own society.
Given that (from what I've read) our current state-of-the-art computer hardware has an IQ about on par with a concussed cockroach that wasn't even terribly bright for a cockroach to begin with...I think it'd take a quantum leap forward in data storage and processing to get anywhere near the potential for a sentient digital life-form.
I say "sentient" because that's what we really mean when we say "intelligent" in this context. Intelligence just denotes a pronounced ability to reason and process and comprehend information. It's not the same thing as being self-aware. For example: GERTY from 'Moon' was intelligent, while HAL 9000 was sentient (and sapient). Huge difference.
If it does happen though, I suspect it'll be gradual and accidental. No sudden Skynet style awakening but an incremental development over time. Even so it's hard to comprehend how such a being would react or even if it could perceive and comprehend our world at all. Tron may not have been far off the mark in terms of a miniature universe within a computer.
I say "sentient" because that's what we really mean when we say "intelligent" in this context. Intelligence just denotes a pronounced ability to reason and process and comprehend information. It's not the same thing as being self-aware. For example: GERTY from 'Moon' was intelligent, while HAL 9000 was sentient (and sapient). Huge difference.
If it does happen though, I suspect it'll be gradual and accidental. No sudden Skynet style awakening but an incremental development over time. Even so it's hard to comprehend how such a being would react or even if it could perceive and comprehend our world at all. Tron may not have been far off the mark in terms of a miniature universe within a computer.
Last edited:
The thing most people overlook is that machines don't NEED to be aware of themselves as such. They're designed to accomplish a limited number of tasks and the "smartness" of AIs depends on how many tasks they're designed to accomplish. But a swiss army knife doesn't become self aware just because you equip it with a million different tools, or a billion, or ten billion, or five thousand plus a mirror.
Sentience is an emergent property of organic thought engines, so "true AI" as you put it would be something that intentionally emulates the biological operating system, either physically or logically. We don't have a particular reason to do that yet, so even twenty years is a pretty liberal estimate.
I'd bet fifty dollars somebody at DARPA already has one, and now they're just debugging it and figuring out how to run it on a computer smaller than an office desk.
But we DO use it. We have the ability to use high-energy fusion reactions to level entire cities with a touch of a button. What we haven't figured out is how to POWER cities with fusion energy.
Then again, we can't power cities with gunpowder either but that doesn't mean we don't have gunpowder technology.
I'ld say Fusion will come first. State of the art in AI is roughly as advanced as when man first discovered fire. Yes, we can do quite a few impressive things with the AIs we've got right now but the majority of the "Artificial Intelligence" still comes from us, human beings, translating our discoveries and know-how into computer code. In order for there to be "True AI", the AI needs to be able to learn stuff from its environment and be able to reason about the stuff its learnt. I can tell you there isn't a single AI right now that even comes remotely close to being able to achieve this.
The best performing AIs that we have right now generally come in the form of a mathematical formula with a number of unknown variables that need to be solved in order for it to do a specific task. By guesstimating the values for these unknown variables from various inputs (sensor readings, data, indicators), the AI "learns" how to solve that task. In order for that AI to do a different task, the variables need to be wiped and a new set of variable guesstimates need to be computed.
Just look at how well Apple's SIRI is at deciphering human speech, or how well Google's language translation stuff works.
The best performing AIs that we have right now generally come in the form of a mathematical formula with a number of unknown variables that need to be solved in order for it to do a specific task. By guesstimating the values for these unknown variables from various inputs (sensor readings, data, indicators), the AI "learns" how to solve that task. In order for that AI to do a different task, the variables need to be wiped and a new set of variable guesstimates need to be computed.
Just look at how well Apple's SIRI is at deciphering human speech, or how well Google's language translation stuff works.
For a 'true AI' the first thing we need to get right is Quantum Computing... without that first step there isn't a viable way to create a true artificial intelligence...
Basic 'AI's' are reasonably possible using todays technology... having your computer help you with basic tasks, answer questions and such, that sort of technology is within reach...
but to actually have a computer that would pass a Turing Test... that's gonna require some heavy hardware above Quantum Computing levels...
M
Basic 'AI's' are reasonably possible using todays technology... having your computer help you with basic tasks, answer questions and such, that sort of technology is within reach...
but to actually have a computer that would pass a Turing Test... that's gonna require some heavy hardware above Quantum Computing levels...
M
^The turing test wouldn't need quantum computing to be passed. It's not an issue of speed, but programming.
In fact, the turing test is not really a good test for "true AI", being that it is only concerned with whether a computer's responses can fool someone into thinking it's human.
In fact, the turing test is not really a good test for "true AI", being that it is only concerned with whether a computer's responses can fool someone into thinking it's human.
I know a lot of humans who fail the turing test (on either side).
The kind of "true AI" you're talking about is possible with ordinary solid-state electronics. The reason we haven't achieved it yet is because we haven't figured out how to configure a hardware device that operates on similar principles to a human (or at least animal) brain. Some of the latest neural nets come close, but not nearly close enough.
The thing you've got to ask yourself is why would a machine NEED to think like a human in order to function effectively? It would be cool to be able to do it and all, just so we know we can do it, but it wouldn't be any better at a particular job than a well-trained human in the same position. If you want to create a machine that would SURPASS human capabilities, you would have to approach the programming task from an entirely different direction and the resulting AI would not think anything like a human being.
After all, what sort of artificial personality would you program into a robot vacuum cleaner? I doubt a Sentient Roomba is going to waste a lot of processing power listening to symphonies, painting pictures and exploring the deeper meaning of its existence. More likely it'll be something about as intelligent as a German Shepherd whose only consistent personality drive involves an insatiable craving for pet dander and the feeling that 120VAC is really soothing.
The thing you've got to ask yourself is why would a machine NEED to think like a human in order to function effectively? It would be cool to be able to do it and all, just so we know we can do it, but it wouldn't be any better at a particular job than a well-trained human in the same position. If you want to create a machine that would SURPASS human capabilities, you would have to approach the programming task from an entirely different direction and the resulting AI would not think anything like a human being.
After all, what sort of artificial personality would you program into a robot vacuum cleaner? I doubt a Sentient Roomba is going to waste a lot of processing power listening to symphonies, painting pictures and exploring the deeper meaning of its existence. More likely it'll be something about as intelligent as a German Shepherd whose only consistent personality drive involves an insatiable craving for pet dander and the feeling that 120VAC is really soothing.
Similar threads
