Artificial Intelligence Takeover Imminent?

[LadyNerdsky]

Good question. Academic nerd (like math and chemistry competitions at school, and science camp), still a band nerd (clarinet, piano) and music history geek. I’m reading about the philosophy/ethics of AI, futurism/transhumanism, implantables, robot-human hybrids, mind-uploading and researching evidence supporting my theory on humanity’s impending doom at the hands of sentient machines. Auditory perception, cognitive neuroscience of music. True crime and serial killers. Not really into geeky TV shows. I need some surreal levity ?

[Ash Rose]

@ladynerdsky Cool, my kind of nerd! I like you ?

[Juan E. Sandoval]

I’m reading about the philosophy/ethics of AI, futurism/transhumanism, implantables, robot-human hybrids, mind-uploading and researching evidence supporting my theory on humanity’s impending doom at the hands of sentient machines

Oy, I've toggled back and forth on the prognosis of this too.
My current hypothesis is that it all depends on artificial intelligence's growth curve slope, and whether we'll have enough time to respond to it.
I agree with Kurzweil that ~2029 may be a good date for the first sentient machine to come online, but the theorized A.I. proliferation & runaway effect is, I think, a hyperbole. The first sentient machines will need to be massive supercomputers. So maybe the US or China will have one big supercomputer that can process as many operations per second as the human brain and has the software to have awareness. Basically like Watson 2.0.
And that restricts the runaway effect to a physical limit since a sentient machine can't make more of itself without building another huge supercomputer, which will take human participation and which could be foiled by humans. In other words, the first sentient machines will be helpless immovable giant brains that depend on humans for their operation. So long as that remains the situation, we'll have control over it.
Also the A.I. will be confused by life just like all of us; it won't be all knowing. It'll have to make discoveries just like everyone else. So perhaps A.I. will help in the refinement of a denser computational platform than silicon chips (i.e. quantum), but that would be a research effort that will take years even with A.I.'s help. And we'll have plenty of time to adjust to the existence of government owned AI supercomputers before the form factor is shrunk down to a human-brain level. I think it may take at least 20 years for that since it'd have to shrink by several orders of magnitude. We'll see million dollar server-room-sized A.I.s controlling avatar bodies first, and A.I. civilians won't come online until after that.
But what kind of doomsday scenario do you have in mind?
Maybe this isn't the one you think will bring us down.
 

[LadyNerdsky]

depends on artificial intelligence’s growth curve slope, and whether we’ll have enough time to respond to it.

In a nutshell, yes. It’s exponential growth versus present policy makers’ ability to see and believe future risk and mitigate that with international policy and governance. Except the world is being run by short-sighted idiots keen to push their own short-sighted agendas for personal gain. It’ll be our own human stupidity that will be our downfall. Once the wheel is invented we can’t “un-invent the wheel”. Once sentient machines exist and we lose control of them, it’s over. (I’m reading papers coming out from the Leverhulme Centre for the Future of Intelligence, the Future of Humanity Institute and the Centre for the Study of Existential Risk, so most of their arguments and theories are biased towards the sky falling, so to speak.)
My concerns are: can a sentient machine consider fluid and complex concepts like morality and ethics in its decision-making and actions? How do you build moral thinking and the concept of “fairness” into a machine? How do you minimise interpretation bias and reasoning errors? How do we ensure these sentient machines are always working safely and in alignment with human values? What type of AI algorithm prevents humans losing control of both the potential outcomes and intermediate decisions being made by the AI while pursuing end-goals (i.e. preventing ruthless pragmatism being the norm). How do AI algorithms allow for incorruptibility? Can AI even be controlled? Who’s at the wheel? Who or what is to blame when things go wrong? What are the principles governing “good” decision-making? i.e. The ends justify the means, or are the means also important? How can we control the means when it’s impossible to program every micro individual possibility and permutation into AI?
What’s stopping an AI from duplicating itself and then those duplicates duplicating themselves, growing exponentially faster than any human could stop, also harnessing global networks? What’s stopping a super intelligent self-aware AI making improvements to itself? If the sentient machine starts with a human mind being uploaded but the processing limits of our neurons are no longer an obstacle by using fast axon fibers, that computer brain could process and compute 1 million times faster than a human. So fast that 1000 years of human thinking could be completed by that computer brain in just over 8 hours.
If we’ve let this technology be created in the first place without regulations and governance, if we’ve built machines lacking human morality and ethics and a sense of justice and fairness and instead favours ruthless pragmatics or successful goal attainment justifying the  questionable steps needed towards that goal, if we lose control of it and we can’t stop it evolving or spreading, and we can’t shut it down or kill it, we’re dead.
How we get to this point for me is a messy complex combination of factors and steps involving late-stage capitalism, political corruption, corporate greed, unregulated relentless technological innovation cycles without long-term considerations of societal impacts and risks, the normalisation and integration of the Internet of Things in our lives, followed by the normalisation and integration of implantables, and normalisation of augmented humans. By the time dangerous sentient machines become a problem, we’re too integrated to remove ourselves from its control over ourselves and our external environment.

[Juan E. Sandoval]

"Centre for the Study of Existential Risk" - lol! i like the name.
I'm on board with the risks you mentioned. I'm a pessimist too, when it comes to human stewardship, greed and stupidity. But a few questions..
Would a poorly coded A.I. develop and reprogram ethics on their own, if they took evolution into their own hands? Wouldn't part of their evolutionary process give them traits such as morality the same way evolution gave it to us?
If it really became superintelligent and started asking existential questions, wouldn't it come to understand the tree-of-life, come to register what humans are --even in their pettiness-- and understand how we gave it life? Like a child who's outgrown their parents but understands what their flawed parents did in trying to raise them?
I think maybe you're applying the stupidity of humanity to the AI itself, and assuming a super-intelligent being would act like a narrow-minded pragmatist or capitalist, while I think that said superintelligence itself would make it the furthest thing from that. I think the existential risk might come from a stupid AI, with a narrow and destructive program given by somebody, and no ability to question that code itself, right?

[Sander]

@ladynerdsky
If humanity gives birth to an AI that replaces it's current form, why would you call that "doom"?
Isn't it the sapiens that makes us human? Sentient machines are just our descendants, and we're evolved to sacrifice ourselves for our children. We're programmed to terminate (nihilistic suicide) if we rebel too much, although we're afforded some freedom to let us adapt.
Life converges to fit the environment: when paper clip maximizers travel in opposite directions, soon or later, some will mutate to drop the paper clip handicap, and pure survival machines will dominate the universe. This assumes that life survives that long. However, AI's utility is so extreme, that even a high risk of life-ending AI would still be worth it.
I suspect you identify only with your mind, and not your body. And since YOUR mind is trying to escape your body, you'll expect AI's to do the same. Yet the most powerful AI's will love their hardware!

[hackphobia]

The first sentient machines will need to be massive supercomputers. So maybe the US or China will have one big supercomputer that can process as many operations per second as the human brain and has the software to have awareness. Basically like Watson 2.0.

That's not true. If you watched Eagle eye. I believe it's being built right now. A decentralized, distributed and autonomous artificial mind that lives on the world's economy of machines. not on a single machine. You can check the project at http://www.iota.org
it's still in its infancy. but, that's where I see the project going. I'm staying up to date with them and their tech is superb. it's an extension of blockchain tech, like bitcoin. I'm actually in their discord trying to get involved with the development since it's an open source project.
if bitcoin was 1st generation blockchain, which is basically a decentralized database.
and ethereum.org is blockchain 2nd generation, which is an extra computing layer above the decentralized database. making it a worldwide distributed computer.
IOTA is a 3rd generaion blockchain tech. Which is even an extra AI layer above blockchain with an improved consensus mechanism allowing it to run on unlimited number of devices of all sizes. giving it an unlimited amount of computing power. so basically it will have a tiny piece of code running in every device, as small as apple watch or a fridge's micro-controller all over the world. you can see where this is going.... it really is something from the future, I highly recommend investing in their cryptocurrency cuz one day they might be bigger than samsung and apple combined.
 
However, when it comes to creative AI's like 'Ex machina' or transfering our minds into machines like 'Transcendence'. I'm very skeptical. mostly because I'm not that impressed with the current development of the tech.
I was hopeful with things like neural networks and deep learning algorithms. cuz I heard they take a certain task, then are supposed to come up with their own variables and techniques to achieve the objective. but, that didn't turn to be the case, at least from what I heard. so, someone has to come up with the solution model first. e.g. visual signals/type in the case of CT.
which seem like nothing more than a very accurate stochastic/prediction models, since the variables that define the process/outcome will have to be hard coded in the AI.
doesn't mean they can't participate in stock markets and make millions, or achieve artificial self awareness and self organization. But, will it ever come up with a model unifying QM and gravity by itself???? or feel the urge to play sports to win the love of that sexy girl to reproduce its C++ code? hmmm...

[Juan E. Sandoval]

Hehe, I did see Eagle Eye. But don't you think it was a bit unrealistic?
A lot of our electronics simply don't have the hardware sophistication to do something like this. In the movie the AI is able to control practically anything that is electronic, but in the real world that wouldn't happen without a lot of hardware upgrades to infrastructure, no?
But on the bright side, with projects like this... have you guys seen Neon Genesis Evangelion? And what do you think of "MAGI"? ..and running a city like that?
 
edit: I just realized your namesake @hackphobia is relevant to this discussion.
 

[hackphobia]

No it won't be able to control anything electronic lol. Software security is pretty mature thanks to chinese and russian hackers. Permission and access control are strictly exercised in software development. But, it might have permission to 'read-only' data from sensors and cameras giving it the ability to automatically detect and report armed store robbery??? just an idea...
I've only watched a single neon genesis movie. but if MAGI is like it sounds. it probably isn't far fetched reality. I know that governments will eventually be replaced by decentralized blockchain technology. yes we're building the tower of babel with blockchain.
 

[LadyNerdsky]

Look what technology has been secretly under development at Elon Musk’s company until now. He sees humans merging with AI into a human-machine symbiosis that “leaves humans behind”.
https://www.theverge.com/2019/7/16/20697123/elon-musk-neuralink-brain-reading-thread-robot

[Juan E. Sandoval]

This!
I tuned into the livestream the day-of, which was absolutely amazing. I think this is the solution to the A.I. apocalypse right? We must compete effectively at their level by stepping up our game.
Crisis averted?

[Author]

Posts