The AI Anxiety (washingtonpost.com)

An anonymous reader writes: The Washington Post has an article about current and near-future AI research while managing to keep a level head about it: "The machines are not on the verge of taking over. This is a topic rife with speculation and perhaps a whiff of hysteria." Every so often, we hear seemingly dire warnings from people like Stephen Hawking and Elon Musk about the dangers of unchecked AI research. But actual experts continue to dismiss such worries as premature — and not just slightly premature. The article suggests our concerns might be better focused in a different direction: "Anyone looking for something to worry about in the near future might want to consider the opposite of superintelligence: superstupidity. In our increasingly technological society, we rely on complex systems that are vulnerable to failure in complex and unpredictable ways. Deepwater oil wells can blow out and take months to be resealed. Nuclear power reactors can melt down. Rockets can explode. How might intelligent machines fail — and how catastrophic might those failures be?"

Comments

What about human-intelligence anxiety

[Anonymous Coward]

It's politically incorrect to say this, but most humans are not intelligent. Their behavior follows predictable patterns. Their intellectual life is next to naught.

If you are not intelligent, you are expendable and fungible like any other industry-raised sheep. Born out of industrial breeding, fed with a formula, and led straight to the slaughterhouse. It doesn't take an evil super AI to beat you. You are already beaten by the mechanism that is called the System.

Be intelligent.

Re:What about human-intelligence anxiety

[Bengie]

90% of what we do is sub-conscious and our prefrontal cortexes will make up a rational story for what we do.

That only applies when you blindly do what you feel like. When I reflect on what I have done, if I did something I didn't like, I analyze why I did something. Once I've locked down on the rational, I can change it, and I won't do it again next time. I used to be easily agitated, but the only thing more annoying that someone bothering me is letting myself be bothered by someone. Once I figure out why something or someone bothers me, I can make the issue not bother me.

The first time I observe a given annoyance, I seemingly have little control over it, it's only once I've reflected on it that I can control myself. An example of this is when I was younger, the sound of a crying baby drove me up the wall. After getting flustered many times, I thought about why I felt that way. I eventually realized that it's because I had no control over them crying, but I wanted to help them to make them stop, but many times trying to help was futile. Once I realized that it was my failing attempts that bothered me and not so much the crying, the next day I was suddenly unbothered by children crying. Assuming it's not a hunger or pain cry.

90% of what I do may be my subconscious, but I can control my subconscious, just not in real-time. I need time to reflect and a sleep-cycle.

Re:don't prevent intelligence because of fear..

[Anonymous Coward]

Since free will is almost certainly an illusion, it's not something you can eliminate. It never existed. We're locked into cause and effect just like all matter.

I'm worried about AI

[rsilvergun]

making most jobs obsolete. That's something that could happen in the near term. Since almost every country bases it's quality of life on the jobs it hands out that doesn't bode well for me, a member of the working class.

AI is just a stepping stone to the "problem"

[ScooterComputer]

AI won't be our biggest problem, it will merely be a stepping stone. The biggest problem facing humanity is the collapse of the informational time line. In other words, data time travel.

"WTF?" I hear you saying. "Whacko." OK, OK. But hear me out... Einstein, et al, are pretty sure that moving matter across space time, especially backwards in the time line, is unlikely without some pretty extreme technology. AKA likely impossible. However, at the Quantum level, moving information may not be that difficult, thanks to quantum entanglement surviving "time displacement" (even maybe black hole event horizons). Surely AI will help to accelerate the research into these areas. And that will culminate with the ability to communicate with the future. And the future being able to communicate with the past. All that needs to be done is construct the "radio". The future will do the rest and send back blueprints for improvement. Even if the humans aren't willing to do it over a shorter future span, the computers would likely have little emotional concern for doing it...after all, what is time to them but energy burnt towards a computational goal (that has already probably been computed in the future). Once the channel of communication is open, it will be as leveling as the Internet across "space" today.

So, the "Singularity" defined as the merging of human and AI is less likely to be as impactful as a "Singularity" defined as the complete crushing of the Time Line. All of human knowledge, nay ALL knowledge--human and AI--will suddenly be known, instantaneously (or very nearly). Parallel computing across both space AND time. Short of the Sweet Meteor of Death, of course.

Something that we're forgetting about AI

[jgotts]

AI is [or will be] programmed by human programmers.

At present there are two alarming trends in programming. One, companies are unwilling to pay properly trained Western-educated programmers and are increasingly outsourcing programming to inexperienced programmers in Third World countries. Once these programmers become better at their craft, they demand higher pay and/or move to the West and the companies move to even cheaper countries. Two, despite outsourcing, there are probably not enough competent programmers in the world to fulfill current and future demand, so there will always be many incompetent programmers being utilized, regardless of economics.

While the world's best programmers are true craftsmen, the worst programmers are the ones we have to worry about. Some company looking to save a few dollars will hire a few incompetents and, rather than your word processor crashing causing you to lose a few minutes of work, your AI's built-in curbs will malfunction and it will go rogue. How many programmers in the world today can design a bug-free security sandbox? As AIs become more sophisticated, every programmer will have to be able to do this. AIs have to be contained, yet what intelligent human would willingly consent to being imprisoned? In the battle between an inexperienced programmer from sub-Saharan Africa in 2100 who is the first generation of his tribe to not be a shepherd, and an AI, can you guess who will lose?

Unless we can change the way our field works, we must assume that the worst and least experienced programmers in the world will be working on AI. There is no basic competency required in programming. Companies will simply pay the least amount of money that they can get away with, like they always do. The AI that kills us won't come from research labs at MIT, it will come from the Microsoft outsourcer office in Bhutan.

We are being ruled already, without AI

[wcrowe]

The machines have already taken over, even without AI. This is because everyone instantly believes what a machine tells them, no matter what. Look at people who have had their identity stolen. A machine says they took out a loan here, and a loan there. Now the creditor is demanding payment. The victim can't convince the creditor that they did not take out the loan because the machine says they did, and the machine is never wrong. Despite the fact that there is no physical evidence that any loan was ever granted to the victim, the machine is believed over the human. No one ever questions how the data got into the machine. No one ever assumes that any mistake was made at any time. Whatever the machine says is assumed to be the truth.

Recently, a friend related this story. He was at a checkout line at Target, and had purchased about $60 worth of items. He handed the clerk a $100 bill. The clerk mistakenly keyed in an extra zero, and the machine dutifully informed the clerk to make change of about $940. The clerk, without hesitation, started to do this and my friend had to stop her and explain her mistake. The clerk couldn't understand. They eventually had to call a manager over to explain the problem. This girl was prepared to do whatever the machine told her, despite the fact that it didn't make any sense.

We don't need AI for the machines to rule over us.

Re:The usual media spin

[ranton]

We are at least 50 years off from strong AI as in human level common sense about the world.

The focus on dangers of creating strong AI is the worst part of current AI hysteria. Strong AI will most likely come decades or even centuries after modern day AI has massively reshaped our workforce and society in general.

The AI we should be worried about include self driving cars, natural language processing, pattern recognition, and robotics. These technologies could combine to make a majority of humans unemployable. The don't do this by making humans unnecessary. They do it by making certain humans so productive that average people have no economic value. It is AI's ability to empower the most educated in our society that is the greatest "problem".

This could either lead to a utopian world if we can enact a form of basic income for everyone, or a very distopian world if we allow income inequality to grow unchecked. Unlike strong AI which most researchers agree has a very low chance of occurring soon, the AI systems I am referring to are almost guaranteed to massively disrupt our economy in the next few decades.

Re:The usual media spin

[ranton]

You mean some sort of theoretical future problem means we have to drop everything, abandon our successful economy, and adopt socialism IMMEDIATELY, without any debate? Yeah, sure.

I said nothing of the sort. First off, every country including the United States have socialist programs. In the United States these include social security, Medicare, police and fire departments, postal service, and many others. Most of the solutions for the almost certain economic changes brought on my improved AI will include socialist programs, such as the basic income I mentioned. They will not include abandoning our successful economy, just like enacting a minimum wage, 40 hour work weeks, or social security did not require abandoning our economy. They will simply add more protections for those who are left behind during rapid economic changes.

Just a quick quiz: do you know what socialism thinks about people who don't work?

Yes, it doesn't have a single opinion on the matter. There is no single set of economic rules that define socialism. Each country that has enacted socialist programs, including the United States, define the goals of these programs uniquely.

Those who don't work won't eat, either. Socialism is about work, not lazy idlers. You don't believe me, I know, so here's an informative quote from someone who knows socialism much better than you do.

George Bernard Shaw does not speak for everyone who desires for more socialist programs worldwide. He is merely a playwright who had very poor opinions of the uneducated and poor. Anyone with such dogmatic and unforgiving opinions knows far less about how socialism can actually be used to benefit society than most people. He was a hateful bigot and elitist, nothing more. If I gathered quotes of capitalists in the 1800's advocating slavery would that somehow show that capitalism is hopelessly flawed?