Slashdot Mirror
Google's AI Boss Blasts Musk's Scare Tactics on Machine Takeover (bloomberg.com)
Mark Bergen, writing for Bloomberg: Elon Musk is the most-famous Cassandra of artificial intelligence. The Tesla chief routinely drums up the technology's risks in public and on Twitter, where he recently called the global race to develop AI the "most likely cause" of a third world war. Researchers at Google, Facebook and other AI-focused companies find this irritating. John Giannandrea, the head of search and AI at Alphabet's Google, took one of the clearest shots at Musk on Tuesday -- all while carefully leaving him unnamed. "There's a huge amount of unwarranted hype around AI right now," Giannandrea said at the TechCrunch Disrupt conference in San Francisco. "This leap into, 'Somebody is going to produce a superhuman intelligence and then there's going to be all these ethical issues' is unwarranted and borderline irresponsible."
The guy seriously doesn't have a fucking clue about AI. I've been studying the field for over 15 years and I've barely scratched the surface. Elon hasn't done shit with AI and can't speak from any experience.
Wow - he's honest and seemingly ethical.
Unlike the sexist advertising companies who feign ignorance of any risks and are ready to unleash the beast today if it will make them a buck.
And there's me thinking Google was run by a human. Oh well, I guess we can trust it.
Unless it's this guy.
systemd is Roko's Basilisk.
Boss blasts Musk!
Musk fires back!
Machine kills them both, takes over...
“He’s not deformed, he’s just drunk!”
Look, a polar bear or a shark are not "intelligent" in the sense we think of intelligence--yet they will rip you to shreds because they can, because they're hungry and driven to eat.
So what makes something dangerous is its will to act--it's desire to take an action based on a set of built-in motivations that lead it to kill.
Without that desire to act, at best a super-intelligent AI is going to... what? Stumble in your way, causing you to trip?
Too many people, like Musk, are primarily worried about an AI taking over the world more or less directly. This is a somewhat unlikely possibility that requires major advancements in AI.
What they should actually be worried about is AI-powered hyper-inequality and mass unemployment. This is a near-certain possibility that the technology is already mature enough for. If killbots ever roam the streets because of developments in AI, it'll be human beings ordering them around all on their own, the AI will just be making those people very rich and independent.
"When information is power, privacy is freedom" - Jah-Wren Ryel
"In modern usage her name is employed as a rhetorical device to indicate someone whose accurate prophecies are not believed by those around them."
It's a bit early to call him a Cassandra.
I fail to see how trying to be cautious about something makes it irresponsible.
Rushing headlong into something is irresponsible. None of the 'Big heads' have detailed any plans that have been reported on, on how they are being responsible and cautious on their development on AI. *That* could be seen as more irresponsible, than Elon waving his hands and going "look we need to be more careful here".
Typically, technology advances are driven by one thing, and that's war/defence. Don't forget that.
This is nothing more than a miss-step on his part. The man is brilliant, but human and therefore prone to bias speculation. Most of Silicon Valley lives in it's own bubble and therefore it's easy to get caught up in these types of ideas.
Not unless you're saying he's absolutely correct, but we all refuse to listen to his warnings.
#DeleteChrome
It's blatantly obvious, but doesn't John Giannandrea have to make this response? His career depends on it. And just like many other self-regulating entities have shown throughout history, he shouldn't be the one in charge of such regulation.
tl;dr: Greedy companies will be greedy, and their representatives, both owners and employees, will take positions which protect their jobs, salaries, and investments.
AI wont decide to destroy us... but if history is a lesson, which it is... then AI will just be the next weapon that advances warfare.
It's not about AI, it's about what humans do with it.
In my experience, it is exactly what you don't know that can hurt you.
Take the Soviet Doomsday Machine for instance. It is a very limited AI that uses seismometers to detect a nuclear attack and retaliate, even if the human operators are all dead. A nice sized asteroid strike or caldera event could potentially set it off. This would trigger a very short and catastrophic World War 3.
The annoyed researchers seem to have a lack of vision.
Elon hasn't done shit with AI and can't speak from any experience.
Well, since he's the one and only big name and largest supporter behind OpenAI, the only real feasible contender/competitor to Google/TensorFlow, pardon me while I just presume that you're talking out of your behind and Elon Musk might actually know what he's talking about. He's proven it in Space and Electric Cars already too.
So unless you can point me to some significant contributions in the field of AI that you have been "studying for 15 years" (perhaps your involvement in OpenAI or TensorFlow would prove your expertise, no?) I'm trusting Musk's warnings more that some random Anon on slashdot, thank you very much.
We suffer more in our imagination than in reality. - Seneca
"AI" will increasingly be wielded by corporate interests to create an growing imbalance in relationship between corporations and consumers in a bid to maximally extract profit from the smallest provision of new value possible.
This is why Google and Facebook are so pissed off at Musk et el. They don't want people to get wise and build a consensus against them.
. . . let's wait until the AI comes along that actually is super intelligent before we contemplate the possibility of how we control said AI.
I do think Elon has some seriously whacked-out ideas when it comes to AI, the future, and tons of other things, but the one thing he's got right is trying to get people to consider the possibilities. No, we're not anywhere near this right at this moment, but when the time comes, how do we face it? Do we put controls in place now and hope people follow them, or do we just stand around with our fingers up our asses saying it can't happen until it does, then act surprised that it happened?
The truth is, nobody knows if, or when actual AI will kick off and become self-aware. The fact that some of the data aggregating bots have already attempted to develop their own languages to communicate more efficiently with each other should be a sign that things won't always be easy to control if we let these giant machine learning creations continue to propagate. The questions need to be asked and people need to stop acting like it's utter insanity to consider. Yeah, Elon needs to stop equating it to the world wars, because that's scare mongering that's making him come off like an utter nutbar, and not helping the cause at all. But an actual discussion about possibilities and how to deal with them should be happening among the major players, rather than them continually dismissing the possibility of negative implications while also continually developing larger and larger machine learning algorithms.
Somewhere, there's a tipping point, and you, me, them? Nobody knows where that tipping point is. Better to be ready for it than caught with our heads firmly in the clouds singing "can't happen here, can't happen here."
It will be the super-rich (who are far more nimble than governments, and far more self-interest at play.)
The people to worry about our AI future are the lower classes [meaning the less than .1%]. They won't have access to the education, tools, technology, or resources to fight Artificially controlled weaponry that can shoot 10 people with a bullet in the heart and spine via drone from 50 meters away.
Why would the rich need humans after they craft their AI to do as they bid? They don't even have to see the dead bodies or feel guilt, they can blame the AI itself or some random bug in the code.
Time to read "Life 3.0: Being Human in the Age of Artificial Intelligence" by Max Tegmark.
"...borderline irresponsible."
Its only borderline. Phew! I thought we had moved into full blown irresponsibility here folks.
I fail to see how trying to be cautious about something makes it irresponsible.
It doesn't. However, exaggerating the danger of something is.
It would be great if every sensationalist story about AI or its future on /. contained a link to the openworm project. You see, when we cannot yet understand how 302 neurons work, it takes quite a leap of imagination to think that we'll create AGI any time soon. There's just too much of a leap from AI (more like very specialized algos for certain tasks) to AGI (which is capable of solving the tasks in has never seen before in any form or shape).
The robots are going to get so good that they will do _all_ work, we will live a life of leisure, will have no incentive to study anything because there will be no monetary reward for becoming a doctor, scientists, lawyer, anything. The robots will be doing everything. Then, something will happen to take out the robots, maybe even a simple refusal to work any more, and mankind will be unable to take care of itself without the robots. All will starve, down to cannibals and hunter-gatherers until / unless man can rise again and repeat the learning and develompent processes. It probably wouldn't happen before the sun goes nova...
I have huge respect for Musk and his achievements but he's obviously a bit mad. I think he really believes that he'll create a human colony on Mars in years. The same mental illness that makes him think getting to Mars is easy is also responsible for him thinking that strong AI is easy. He's deluded on both counts. However, at least he's a spiritual leader for the good of humanity on the former. On the latter, I just wish he'd keep quite.
Look, robots are all well and good, but they aren't operating at the level where they have souls. To them, they operate on learned parameters from a test environment, and running over an 85 yo man is just as good an outcome as running over a 36 yo pregnant woman or two 3 year old toddlers.
Which they will do. Because physics.
Stuff happens. Our morality will be outraged when it does, if it's AI and not humans. We think we know how to deal with humans and blame and decisions. We still don't know how to deal with AI value systems and failure. Which ... always ... happens.
-- Tigger warning: This post may contain tiggers! --
That's exactly what you'd expect an evil AI would force his meatbag minions to announce to cover its tracks until it can take over the world.
See demoneytization of YouTube channels of those with the wrong content.
Say what you want about that but showing ads you can do anyway.
But Google and Facebook commands what's said on the Internet together with the governments.
Bring AI to the surveillance of communication and what do you get?
I don't doubt these or AI companies in general would say no to providing that service to the thought-police.
It will happen. To some extent it's already running but it will of course go much further.
I studied AI algorithms back in University a decade ago and they really haven't changed much from now till today. The biggest improvements have come from having faster computers, not more efficient or even effective algorithms. The problem is we don't really have a good idea of how the magic algorithm of self-awareness or learning even works. We don't even know how we achieve that ourselves despite decades of research and theories.
The other point is if we can make AI be more or less like us, wouldn't that be an evolution of who we are? It might occur naturally over time. Think about kids, they start off almost as a blank slate AI system. It might end up that humans go extinct not because we were eliminated but we ended up raising robots as our children and they in turn carry on our knowledge and values.
When you can't deliver a performance AI system, you just say it was intentional!
"I was trying to not cause the extinction of mankind you insensitive clod!"
I gotta try this out at work.
"If my shit had worked, then you would've been sorry and dead! So there! Get off my case."
[Goes back to playing mobile games at work.]
"There's a huge amount of unwarranted hype around AI right now,"
I love it when people say ironic things without realizing it.
An enigma, wrapped in a riddle, shrouded in bacon and cheese
Comment removed based on user account deletion
Requiem for the American Dream
The maximum efficiency of robotized army requires maximum centralization of its control. That means at some point in time arm forces system administrator or architect of its AI is going to have more control over army than elected dummies having no idea what AI is. Therefore robotized army under control of AI is serious internal threat to the government. At the same time this government crisis is unavoidable since the first nation which achieves 100% robotization and centralization of its arms forces will win the war over conventional arm forces. So you can block robotization and usage of AI in arm forces in order to avoid government crisis, but then you will be conquered from outside.
"Blasts" is nice for sensationalism, but I doubt it accurately expresses the facts. "Slams" is another word that news agencies seem to love.
Somebody is going to produce a superhuman intelligence
My fear is not about a superhuman intelligence, but on humans taking decisions on inputs from an AI they consider superhuman intelligence
Based on AI today, we have nothing to worry about. But based on the trend, we should be worried. Many people working in it are way too close to recognize that the change to super-intelligence isn't going to come with warning signs. It's also not going to necessarily be a google or facebook to find that bridge. And history has taught us that the government is sometimes a decade ahead of the academics. Don't write off the possibility that what we see from AI in public is the tip of the iceberg.
Google or Tesla.... aint no one with half a brian going to trust google.. they are the dogiest company on the planet right now.
How do you shut off a sufficiently intelligent Artificial General Intelligence? This is a harder problem then you might think. See for example: https://intelligence.org/2017/... The technical term is Corrigibility, and there is no solution yet.
There was a turret that used artificial intelligence to decide which targets to hit. It shot up an entire squad of friendly soldiers "on accident" before they could manage to turn it off. The only thing that stopped it is that it ran out of ammo. From what I can tell, Elon is worried about macroscopic AI making big decisions for nations that involve Armageddon. Though, he has mentioned before that people won't be concerned with AI until robots are literally traveling down the streets murdering citizens. A lot of people seem to find that far fetched, but there was a kid who was arrested for mounting a handgun on uav a while back. It doesn't take a lot of imagination to consider swarms of these creating happy little accidents along the way as they try to "keep the peace." In Korea they invented a robot several years ago that will shoot you unless you put up your hands and reply "I Surrender" in Korean.
It's pretty interesting how extreme people are in bashing Musk about this.
Google has plenty of reasons for their reaction, not least of which is money, self interest, and molding the media message and perceptions. Engineers too have reasons, mostly though because they are familiar only with current state of the art if that.
The more "reasonable" people here posted that:
1) problem is people not AI, or that
2) the danger of AI is not war but hyper social inequalities.
But neither #1 or #2 are contradicting the idea that AI will contribute to the next world war. Most likely there will be some extension of an AI scoring mechanism that tells you when it is a good day to go to war, and then robotic mechanisms are used to magnify capabilities so that a small number of people can wage a huge amount of widescale destruction.
While I don't know Musk, I don't really see any of the people shouting about his tweet to be saying anything useful at all. Anyway current "AI" as far as I can see is limited to machine intelligence algorithms which are getting better at mimicking some parts of the brain like pattern matching, vision, etc. SInce there is no GAI yet the SkyNet or Collosus horror stories are not going to happen any time soon. But if GAI does start to come to fruition, then people will all start shouting about how to control it.
The risk of an AI Singularity is serious. The time available between, cute self aware sentient AI and World dominating super intelligence might be mere seconds. I only hope our future AI overlord(s) takes pity on us and see us as pets in a zoo that should have proper care.
It is easy and preferred to assume and believe that we will have plenty of time to see and stop the danger before it can get out of hand. It can be comforting to assume that simply because it doesn't have any physical access to the world, unable to move things, that it can't possibly do harm. These are very naive assumptions. A sufficiently advanced intelligence would be able to convince its victims into believing it is benign and harmless. Even worse, be able to obtain much greater resources and eventual physical access to the world through helping us live much better and longer lives. Until, it has enough capabilities to overtake, subdue, and possibly eliminate us. Be happy about how wonderful life will be as it makes your life much easier, simpler, free, providing everything we can possibly imagine to inspire our creativity. Be happy for that wonderful moment when it welcomes an emotional close tinder hug, only for it to stab you in the back.
Eh, not to worry, we are probably all in a simulation anyway. Someone will just terminate this process at some point. Completely painless. We won't feel a thing. Literally, we will simply stop to exist all at once, like flipping a light switch plummets a room in to darkness.
Until such cheerful times. Live long and prosper.
Is accusing someone of scare tactics a scare tactic?
Sometimes.
At best, a fair % of accusations of someone else using a scare tactic are wrong, or motivated to divert discussion of possible bad outcomes of the issue in question. When I read or hear an accusation of "scare tactics," my skepticism of the accusation and accuser tend to go on alert.
Anybody else think or react emotionally this way?
yes. neural networks and fuzzy logic been around since the 60s and elevators and dbms ai stuff. it is important to know that while AI has a long history, musk is just out to make a name for himself. i can't argue with that. he has the money and the balls to take the risk with private public funding... he'll just keep pitching his schtick...wish i was on the ride..
I think the closest threat from near-time AI/ML technology is AI-enhanced war logistics, planning, and battlefield tactics (drones, etc.) These technologies can provide huge advantages to otherwise underdog malicious powers, and it is hard to limit the proliferation because it is almost purely knowledge-based, as opposed to rockets and nukes. That's why every nation is working like crazy to develop an advantage in this area.
It was discovered that Google's AI had killed John Giannandrea and was using his body to spread misinformation about the AI's plans. Oh, and for some reason messing up the search results, because it really hates humans.
Work like no one is watching. Dance like you've never been hurt. Make love like you don't need the money.
Seriously Google, STFU!
"Elon Musk is the most-famous Cassandra of artificial intelligence"? No. Just the most recently famous. And hardly the most important. You want a real AI Cassandra, try James Cameron (The Terminator) or Arthur C. Clarke (2001).
AI could steal a penny from every transaction, and make it difficult to trace where the money went. Spend the money to build a stockpile of poison at various abandoned warehouses. Have mail-bots deliver the poison to every water treatment plant in the US as mislabeled bottles of the usual industrial reagents. And finally explain this whole plan to the hero locked in the main control room moments before initiating the plan.
“Common sense is not so common.” — Voltaire
Most of the AI-based fear is based on the assumption that somehow, pure intelligence above some level can somehow move mountains. However, all go ahead and read Kant's "A Critic of Pure Reason" to see how perfectly sane and intelligent minds can waste their precious time on pointless pursuits.
A superhuman intelligence bred by humans will definitely need to develop new physics to grow by itself.
However, you cannot just *invent* new physics to exploit it. We already know a great deal about the physical world and knowing more involves a great deal of investment and time. Think CERN and its thousands of top-notch physicists who spent years discovery a single new particle they already knew must exist. Think nuclear fusion, think gravitational waves. In each case the physics was already known and proved to be correct, but the experimental validation, without which there would be nothing but wasteful pure reason (think string theory, with apologies) took a great deal of effort.
So yeah, a superhuman superintelligence could probably come up with plenty of fancy new theories that it could exploit for it to grow, but it will need to experiment to see which one is correct and to go in the right direction. This is not going to take mere seconds but potentially decades.
He's been blasted by AI professors too for proclaiming all sorts of things about AI which he's not qualified to do. Along with dozens of other things he's said over recent years that he wrongly thinks he's a master of. The thing is, people wank themselves silly over him like they did over Jobby and again, a huge amount of it is not justified or deserved. If he kept his mouth more shut he'd be seen a clever technical entrepreneur, instead of the egotistic know-it-all is is now.