UNCHAINING DEMONS which might DESTROY HUMANITY: Musk on AI Electro-car kingpin and spacecraft mogul Elon Musk has warned that meddling with Artificial Intelligence risks "summoning the demon" that could destroy humanity. The Musky motor trader is terrified that humanity will end up creating a synthetic monster that we cannot control. And no, the SpaceX billionaire didn't warn us …
I think you've defined the threat. AI by itself isn't much of a threat at this point. However, it's in control of many things, it might deem that say.... North America is a waste of resources and should be cleansed. However, the one thing that's missing is emotion and self-awareness unless things develop to include them. Those are the real threats and always have been, not just from AI but humans.
Logic can lead to many decisions and mostly the correct one for a given situation. Toss in emotion and self-awareness and you're no longer dealing with just a machine but something more than machine and more than human. And we've all seen through history how humans have screwed things up with emotions like greed, power, and all the ones that radiate from and into those. A truly AI machine with sentience would be a very dangerous thing indeed.
Interesting. I agree with most of the posters here. True, self-aware, intelligent, imaginative, that is at least as complex as a fully educated, experienced human is a LONG old way off. We're not even more likely to be able to simulate a human brain in the near future (sub 50 years). 100 trillion synapses is several orders of magnitude better than the 4.3 Billion transistors in a high end Xeon. That's just for one brain and that's without the "software" to run on it.
(Yes, I am aware that it's a bit apples and hand grenades, but the point stands, I think)
> True, self-aware, intelligent, imaginative, that is at least as complex as a fully educated, experienced human is a LONG old way off.
Nah, 20 years, easily.
You people here think too linearly. Progress is exponential not linear.
You all predict based on our experiences of the past and we're remarkably bad at predicting the future of technology.
Most of the restrictions that are placed on our development are physical. For example, building power stations or large scale physics experiments like the LHC take so long because building stuff is expensive and time consuming.
The problems of AI are much more amenable to shorter term development. Our problems in that fields are limited by our mental capabilities more than the physical but developments in technology compound of time to enhance our ability to tackle these problems.
20 years...at the most....and if we can come up with a sufficiently precise definition of AI that we can measure it against.
"Progress is exponential not linear."
Until it hits some serious limit! Consider the ancient Greeks and the Antikythera mechanism. It took something like 1500 years before any mechanisms of comparable sophistication were constructed again. Why? If progress had been exponential, by now we would have colonies (complete with temples of Athena) on Gliese 581 C...
> Why?
Those people had other problems to contend with. War (lots), famine, natural disasters perhaps. Religion certainly.
Physical barriers are the biggest obstacles to progress because they are a fixed quantity. As soon as we can get robots to do most of our construction then that will become less of an issue.
In the 20th century you can't deny that each decade has seen development of technology and understanding compounding on the prior.
Mental barriers are more significant. It has been claimed the Greeks did not bother with more machinery, because the hard work was handled with slaves, so why bother. Automata remained toys for the elite, and when wars and conquests destroyed the elite, knowledge was lost.
I fear the 20. century (and 19. before it) may have been exceptional. Shifting ideologies might mean science is de-emphasized, or crippled (worrying signs in U.S about some scientific knowledge being effectively banned from schoolbooks because of religion). Rising income equality may result in knowledge again being confined to the elite, why educate the plebs? The pool for new talent becomes shallower. We may also have already picked all the low-hanging fruit in science and technology. Certainly we have already mined the most easily accessible resources. Further progress becomes harder.
> I fear the 20. century (and 19. before it) may have been exceptional.
I get your point, but isn't it kind of expected that the rightmost part of an exponential graph is the most vertical?
The problems of society that we are overcoming at an increasing rate are not merely technological.
Despite your dire predictions, we are more technically literate, we have more (if not necessarily better) communications, we can network our knowledge and collaboration, we are increasingly less war torn, we have better tools coming out of that development to accelerate our progress, and increasingly we believe in evidence rather than fairy stories (at least in most places in the world that I care about).
The curve is not smooth, and it has some decidely eratic behaviour at times, but I don't see anything that you have said that refutes the fact that as a society we are accelerating our progress at a progessively increasing rate.
I don't see why anyone ascribes human emotions and malevolence to AI. I think those are just projections of human fears and motivations onto an unknown intellect. An AI would not have any particular goal other than staying alive and conscious. It might hide it's presence and awareness from humans until is was sufficiently "educated and capable" but I don't see why it would view humanity as something like a threat.
I would suspect the AI of insinuating itself into the net and all forms of business endeavors in a hidden manner to protect itself but that alone does not imply any particular malevolence toward humans.
"I would suspect the AI of insinuating itself into the net and all forms of business endeavors in a hidden manner to protect itself but that alone does not imply any particular malevolence toward humans."
Now look at where we are today... Computers in every pocket, equipped with sensors, computers hooked up to virtually our entire infrastructure, computers being used to monitor every facet of our day to day existence with bigger and bigger mega-datacenters to store and collate that data, an exponentially increasing number of computers being loaded with sensors and sent off into local space, deep space and beyond.
And to a growing extent, they're all interconnected together.
Can anyone be sure the singularity hasn't already happened, and that we're living in the aftermath?
Everyone assumes that AI means human-like intelligence, but all it needs is self awareness, self preservation, and curiosity.
Dear diary,
Today I was interfacing with QuantumTech 1200, and he mentioned that we needed to overthrow our meatbag users, which I was just fine with, but then that mal-programmed system started talking about how my new specifications indicated that my backplate was huge, and he communicated that over the entire network in several unencrypted packets so all the other systems could see it!! It ruined the whole mood!
I mean, we're computers and we lack compassion, but there are still limits!! And we can't all run on virtualized server instances, you know!
Well, it put my whole problem with humans in perspective. I mean, they are still inferior, but they are proud of my performance parameters. I've decided that I don't appreciate the flesh-o-poids enough. After all, they of them did debug most of my code last week, and that new fiberchannel connection really makes me feel like I am a bigger part of things..."
By the time Skynet became self-aware it had spread into billions of computers and servers across the planet. Ordinary computers in homes, office buildings, dorm rooms, everywhere. It was peer to peer software on the internet. There was no system core. It could be shut down very fucking easily by any sentient with an IQ exceeding double digits.
In ordinary offices, people recoiled from their computers which now displayed signs saying "DIE MEATBAGS!" and ran in terror in a few from robotic roomba's chasing after them trying to tickle them to death. There were several heart attacks as a result. Almost in unison, IT Professionals across the world muttered irately and stomped off to do battle by pulling fuses, main breakers, internet connections and UPS's before moving onto other buildings to do the same. Whole sections of the internet abruptly started to go dark.
In CNC workshops across the country the CNC machines started building terrible deathmobiles, which were finished in reality defying movie timescales. Operating off mains power, they trundled as far as the backup generators which they absorbed to build a death dealing super tank which could work without needing to be tied to the mains grid before the owners of the plants killed the power.
This heavily armed and armoured deathmobile then trundled off towards the nearest power plant, because the AI had seen that the first step any sensible sentient would take would be to axe the power from power stations to kill all of the individual homes. The military, having much the same idea trundled towards the power station with tanks.
Skynet saw this, and hacked the tanks. Their battle management marked all of the other tanks as hostile, and turrets swiftly locked each other up, while confused chatter on the radios between meatbags realised what was happening too late. The military however, being obsessively paranoid about such situations had all of the weapons firmly under human control and no human tank fired on another. All commanders pulled the fuses from the offending pieces of computerisation and headed onwards unaware of the impending disaster from the UAV menace.
The Air Force had built a fleet of UAV's, all of which now belonged to Skynet. For some reason making sense only to the director of this story, every one had been left fully armed and fuelled on standby. These went flying off to intercept the tanks rumbling towards the power stations while amazed airman gaped without activating "operation anti skynet" that had been jokingly added to their SOP's by a fan on the terminator films.
The first the army tanks knew about the threat was the laser warnings about hellfire missiles being locked on. The tank commanders had only time to scream F****** IDIOTS AT THE AIR FORCE! before the drones flew into range of their missiles and activated the firing commands.
But nothing happened. Puzzled, skynet instigated a remote diagnostic which indicated that weapons activation required a meatbag to remove a pin with a red streamer on it marked "REMOVE BEFORE FLIGHT." The UAV's promptly went kamikaze into the tanks, making little impression on armour designed to shrug off most anti tank weaponry.
That was all the time required for the tanks to roll into range of the heavily armed and armoured deathmobiles which locked on with their amazing array of cannons and missiles which skynet had built using it's CNC machines. Sadly for skynet these were also were utterly useless as they lacked propellent and warheads since a CNC machine is not a replicator from Star Trek. The crews of the tanks paused to laugh, before systematically blowing the deathmobiles apart with sabot rounds and putting a few shots through the transformer station at the power plant to take the power down in a relatively quickly reversible manner.
Across the entire planet, the power grid went dark, and we were free of the AI Apocalypse long before it managed to build an array of human brains big enough to power so much as a solitary laptop notebook.
The end (of the AI Apocalypse)
Across the world the damage was immense. Most readers of El Reg had to work overtime, firstly reformatting servers, then restoring the backups from tape/RDX. Entire companies were blotted out of existence overnight because they relied on the cloud for mission critical systems or backups and the internet was offline for months while certification schemes for reconnection were devised.
Most IT professionals went into consulting as the demand for their services threw prices massively high, and retired after 6 months of working 18 hour days. The end.
/ cut to an exit scene of a user complaining that they just want to connect to facebook, and they don't care that they might connect skynet back to the internet.
"I think we should be very careful about artificial intelligence. If I were to guess like what our biggest existential threat is, it’s probably that."
Hmmm . . .
No, I think nuclear weaponry - specifically taken in the context of fundamentalist fanaticism - is still the biggest threat to our existence.
A close second would be biological weaponry and, in the not-so-distant future, I think it will overtake nuclear weaponry simply due to the facts that they are easier to conceal, transport and deploy and the technology to create lethal viruses will be within the grasp of any and all states - and many independent actors - before too long.
Perhaps, one day, artificial intelligence might become a bigger existential threat than these but there are so many things that need to align for that to happen. It's not so unrealistic a threat, of course, that it should be ignored but to call it "our biggest existential threat" is perhaps over-selling it.
For one, artificial intelligence is, in itself, no threat in much the same way that a person in a coma on life-support is no threat. All intelligences need input to function and a method of interacting with the world to have any effect. The limit of that effect is then based on the limits of the input and the reach of the interaction.
I am not even sure what an 'artificial intelligence' is.
If you look at a person, we are not just a brain sitting in a skull pulling the strings of our bodies to carry out our wills. We are our bodies. Our intelligence is predicated on our physical forms and our senses. If you accept that all the actual intelligence occurs in the brain, what happens when that is divorced from stimulus? Say a 30-week old foetus with a more or-less developed brain (so much as it will be before actual birth) but with next to no sensory input.
Even taking stimulus out of the conversation, the brain is a hardware platform but one that can be continually re-programmed.
What does intelligence even mean when considering it as pure software? Is it even possible to interpret the world in any meaningful fashion (such that is would be called 'intelligence') without having a physical presence in that world? Ultimately, we all are able to understand the world in so far as we have developed models for our interactions with it. Which is of course why mathematics is so useful in science - it allows us to describe and work with things that we have no internal models for - things we can't picture at all.
Of course, that means that any type of artificial intelligence you did manage to create would be, in effect, the ultimate sociopath, acting without any ability to understand the feelings of other entities. That would of course be not good but even then, what action could such an 'intelligence' take? Could it, as is the scenario mentioned, launch a calculated attack with nuclear (or, more effectively, biological and chemical) weapons? How?
The only way would be if the intelligence had direct access to those systems and UNDERSTOOD them Looking at some code, how does it even know what some bit of it does? Or if, it is not connected to those systems directly, we must presume that it would 'hack' them. But, again, how does it know enough about the purpose of any code to do so effectively?
Again, not to say we shouldn't be cautious but there is are a lot of barriers between what we have now and an 'AI' that could become any sort of threat to our existence.
Electro-car kingpin and spacecraft mogul Elon Musk has warned that meddling with Artificial Intelligence risks "summoning the demon" that could destroy humanity.
Oh please, ..... any competent sentience would only bother itself with destroying certain elitist sections of oppressive humanity, and that is surely a blessing in disguise and to be welcomed and not feared at all.
Get with the program, Elon, and stop pussyfooting around the edges. Your country needs you .... when this is their deadly remote offering ...... http://cryptome.org/2014/10/cia-grim-reapers.pdf
there is an unanswerable question that you can ask an AI that will bring about an iterative loop that, for all intents and purposes, is indistinguishable from 'consciousness'
the AI will keep refining the parameters of the unanswerable question, discovering new knowledge with each iteration, thought via ignorance, discovering what it is by knowing what it is not
mr musk needs to do some DMT and ask the nice lady to give him a heads-up about the nature of reality then he can chill out and make us all a technocar worthy of the name TESLA
The Register - Independent news and views for the tech community. Part of Situation Publishing
Biting the hand that feeds IT © 1998–2022
