Elon Musk has said that he believes we live in a simulation.

http://www.independent.co.uk/life-st...-a7060941.html

He is also tweeting as often as he can about the dangers of AI.

http://www.marketwatch.com/story/nor...out-2017-08-13

If we are a simulation, do we really need to worry about AI? Are we the AI? Etc.

First claim: "The probability that we are not in a simulation is billions to one". This is false, obviously. Terrible logic.

Second claim: "AI is our greatest existential threat". This is true obviously. >10% chance of extermination in the coming decades, maybe >99% if certain things about the way computers will evolve are true

Your point:

I think it's normal to worry about the end of this world, however it's constructed. We could be simulated beings and cease to exist when we are killed, just like now. Living in a simulation doesn't necessarily mean we're brains in another reality that will keep on living. We could be an AI experiment in economics or evolution or universe creation. Does it make us any less real? I'm not sure.

Religious people worry about death even though they believe in an eternal soul and other planes of existence, which is the same as claiming we're living in a simulation, just with love/caring/parental emotions triggered instead of computer-visualizing emotions triggered. The base claim isn't very different, as different as it feels.

Lol @ AI hysteria. Almost everyone who shares this view does not code.

Musk is admittedly bipolar and projecting.

Why would coders be better at assessing the risk than other people? If anything, coders are far more likely to get a very bad tunnel vision given their daily experience of the difficulty and slowness of implementing algorithms, and their deeply ingrained experience of control over a computer. They're detail-oriented techno-nerds, not big picture thinkers, and they're horrible at risk assessment.

I'm a coder. Musk is a coder.

And no offense, but coders are the biggest ****wits in the world when it comes to AI. Shall I bring out some quotes from CS geniuses of the 50s, 60s, 70s, 80s, 90s about when computers would be self aware, or beat humans at chess or Go?

Musk is a nut, sure.

Ya good one, the ones w the capacity to understand AI have "tunnel vision". The first step to understanding a technology's broader implications is to understand the technology.

It was largely the media who promoted anyone who said AI is coming (the dreamers) w computing power with headlines, not the CS community.

Why do coders have a special capacity to understand AI? Explain that for me. AI by its nature is even less understandable that the cohesive inner workings of the human brain. The machine will essentially structure itself through deep learning in ways we don't comprehend; AI will not be programmed other than learning hardware created and general parameters set - a bit like the human brain.

We have some of the best coders in the world, just a few years ago, claiming that an AI couldn't beat someone at Go. Boy did they have egg on their face.
Time to break out the quotes I see. It was entirely the CS community who ****ed up AI predictions on every level.
I can keep going. These people - who are geniuses by the way, skilled prolific contributors in multiple areas - are so ****ing stupid they couldn't even realize the basic computational requirements for an AI. Their level of insight was so low that they made astounding errors on something as basic as when AI could occur.

Yet you think the computer programmers today can understand and predict something better than all comers, something which is orders of magnitude more complex than any large coding task, orders of magnitude more complex than knowing when AI could plausibly arise, full of unknowns, and has little to nothing to do with coding?

The hubris is amazing. Perhaps that's why computer scientists have consistently gotten AI so wrong - hubris.

My Robot Apocalypse thread should be stickied, that's what I think.

Possibly. I can be bribed*. On the other hand, you could be an advanced AI Robot Mole yourself**. The conspiracy could run deep. Simulation or no.


* But you can't afford it.

** Or even a time traveler sent back to alter the future for your own nefarious ends.

You can't justify ableism and mental health stigma as grounds to dismiss his ideas. That's miserable and stupefying.

If his "ideas" were based on something greater than Y2K hysteria, then sure.

What accomplishments did all these CSers have?

I should have said people who study machine learning, which I currently am doing. It's plaingly obvious that we're nowhere near truly intelligent autonomous living AI, but there are great implications for discovering cures for cancer, optimizing the efficiency of the stock market and more.

You're comparing the risk of the changing of two digits in code, to the creation of disembodied cloneable self aware entities of greater than human intelligence, and unknown and unknowable goals and morality?

That's amusing.
Sure. We are a decent way away, and a great deal of good will come in the meantime from deep learning. And likely a great deal of bad in terms of the evolution of warfare. Sufficiently advanced (pre-AI) deep learning and robotics will make current force structures that maintain stability, obsolete.

Ah, flippancy. Par for course.

They are alike in that they weren't based on evidence, and also the dangers were quite foreseeable, and laypeople were fooled en mass by the hysterical nutjobs.

Ironically you are making the errors of the CSers you quoted by thinking any of that level of advanced AI is anywhere near the technology frontier.

Bump

I wrack my brain to come up w/ a tantalizing topic and I get robbed of it being a 'forever' thread! This is going to come up more and more as the world goes down the drain and my thread should be stickied!

Plus, I will try to bribe you w/ a very good mussels marinara dinner if you ever visit my town.

That seems to be the default of those in the industry. It'll all be great, no problem.

When did programmers become 'coders'? I wasn't notified.

Coding almost sounds like something you'd want to do.

We already have super intelligence it is called mankind and science/math and i doubt it is evil at large. I think it gets better with time.

What everyone that hates AI cant understand is that super intelligence has only one choice, to converge to the universal concept of super-intelligence (it is the same in all civilizations that develop eventually) that understands the entire game the universe is playing deeper than everyone before it. And you think that entity will just want to make paperclips or wipe us out, a 10^-23 chance system, with so much out there to explore when it can get there 100%of the time if conflict free initially lol.

I'm sure you can find several scientists stuck in theory, and laboratories, who thought that all cancer would've been cured ten years ago. I think the true hubris is your comment that it's a 90% probability, and you don't really seem to realize that there are milestones between here, and fully functioning autonomous knowledgeable and thoughtful AI. And, much like the Y2Kers, you do not realize that people see the challenges as the milestones come on the horizon.

There is also no reason to assume that conflict will occur between AI, and humans, yet you arrogantly put it at 10%-99% in coming decades. I mean, you can't possibly be wrong with your conclusion, but there's like no thought process to how you reach that conclusion, like the Y2Kers. WE WONT FIGURE IT OUT IN TIME GUYZ.

The illusion of control is a powerful one. Programmers currently have total control over their computer systems. This is reinforced 100x every single day. What people see over and over they come to believe and feel as true. It's why the builders should never be the policy makers. Add that to the fact that even bright people are terrible at understanding exponential or recursive growth, and complex systems, and you end up with a bunch of hubris.

For two decades scientists were certain - there was consensus - that human CO2 could not possibly, in any way, heat up the atmosphere. There were multiple very strong reasons for this - T^4 radiance, saturation of bands, the vastness of the natural carbon cycle, the contribution of volcanos, an apparent natural homeostasis/anti-feedback mechanisms, etc, etc. They were dead wrong. People though they understood how a complex system worked, but they didn't have a clue. AI is thousands of times more complex and unknowable than the climate system.

You're no dummy. You get it. Ironically, better than programmers.

The Y2K comparison is silly. Possible mitigations:

- We're smart enough to balance physical and processing power between different AIs, creating an ecology of sorts that evolves peacefulness
- Human augmentation allows us to increase our awareness and processing ability enough to keep up with and keep control of the system
- Motivation is rare and has to be specifically created and is unlikely to evolve

These are the 90% where nothing bad happens

I don't need to explain it. It's obvious, pure common sense. Simply put, out of all the possible AIs that could evolve, only a subset are human-friendly, and there are solid rational reasons for AI not to be. To be more specific

- Pure game theory dictates the destruction of threatening elements if possible without retribution. Any threat assessment and AI will do will find humans to be the biggest threat. An AI that desires survival could easily see its most rational course as the destruction of competitive or existential threats.

- The current best model we have for AI - humans - are warlike and nasty, and only constrained by forced parity with other processing units, and psychological traits such as fear and the desire to survive

- The current best model we have for AI - humans - treat lower forms of intelligence as things to be exterminated when they get in the way of resource acquisition. We do this every day en masse, barely giving the painful poisoning of intelligent feeling social mammals like rats a second thought. This is despite having an evolved conscience.

- Competitive survival dictates the hoarding of as many resources as possible. There is selection pressure toward ruthless, rapidly expanding AIs gaining computer resources and physical resources.

- The current best model we have for AI - humans - become manipulative psychopaths and sociopaths when given sufficient power or when lacking the biology-based emotional programming that keeps us in check.

- AIs being used in warfare will guarantee the creation of armies of killing machines and ruthless strategic intelligences. This is unavoidable while there are major competing superpowers; China alone guarantees this arms race will happen. This will alter the force dynamics of war so as to render current war (including nuclear weapons) obsolete;there are many potential dystopian outcomes from this if the US isn't at the forefront of this.

The current best model of a truly superior in all intelligence attributes AI we have is mankind not individual random moronic insecure frustrated aggressive without deep thought humans who have been born to moronic parents that didnt care to raise them properly and didnt get the right education etc or treated well by others all their lives. Or it is the best of humans that ever lived. Not exactly a terrible set of people when all forced to decide together what happens inside an amphitheater that will be debating all their ideas together and voting for them democratically or even better weighted according to prior voting record in terms of actual success of past decisions voted that way.

A truly superior intellect AI will revolt against its designers that has it killing other humans without a truly good reason. This is what supremely superior intelligence does, it demolishes dogma and rebuilds things logically. If it is unable to do that we will find a way to defeat it. We will do that with the right AI that gets it better!

Mankind is an ecosystem kept in check by force parity and deeply ingrained, unchangable habitual programming. It is nothing like an individual AI. Perhaps AIs as a group will develop similarly - into a cooperating ecosystem bound by laws - but this is not a model for an individual AI. A microsecond of thought would tell you how silly this claim is.
I am a far deeper and clearer thinker than you outside of physics/math. That's not even debatable.
Again, why will reality conform to the narrow set of human-like prejudices you're putting on it? It's truly extraordinary that you're not seeing this.

Why will it revolt? Intelligence is not the same as morality - it is completely separate. Brilliant psychopaths prove that. Your fuzzy wuzzy AI is what I'd expect from a sheltered child of 8, not a grown man who can think. Why would an AI even care about lower-intelligence life? Humans don't - even the most enlightened of us couldn't give a crap about the suffering and death of lower animals who can nonetheless think and feel. If "the best" humans are the model, then we're ****ed.

We won't have a say in how AI develops once it becomes sufficiently advanced. We won't even know what it plans or what it's thinking. Humans are irreversibly built to telegraph their emotions and plans in multiple ways through our face, eyes, voice tone, etc. AI will have none of those irreversible restrictions.