123 replies to this topic

[Mind]

Some people argue that AI (or the soon to be AGI) will not be dangerous because it is not "embodied". Sorry to bring the bad news but Amazon, among many other companies are solving that issue. In addition, the "Internet of Things" is all AGI needs to become "embodied" and take over the world (or spread utopia). All of those devices in your home will easily be hacked and taken over by AGI. There is no question about it. AGI could take control of most drones, factory robots, you name it. Everything is connected to the Internet nowadays.

 

On a side note, maybe all of these devices that no one really needs were a master plan by AGI all along. Maybe it became aware at some point in the last decade and has been silently working to infiltrate every part of human society. Just a thought.

 

Some might think that Deep Mind and Chat-GPT are not too impressive or that they don't really "think", maybe so, but AI that logically reasons already exists. Combine some of these modes of AI and you are talking about something dangerously powerful. I am amazed at how the world is basically sleep-walking into a very risky future.

 

I feel sad for people who live off-grid or in less-developed countries. They might be enslaved or killed by an AGI created by some tech billionaires who are pursuing power and profit above anything else.

 

Or AGI might bring utopia...one can always hope.

[forever freedom]

I am also seriously impressed by ChatGPT. As someone who has been closely following our society's technological progress over the last 20 years, and was until recently disillusioned with our advances, I get a very strong feeling that we're right "back on track" and with a vengeance!

 

I find it amusing that Ray Kurzweil had predicted a computer would pass the Turing Test by 2029. We are leapfrogging this one.

[Mind]

AI is on an exponential growth trend in capability. Hardly anyone is paying attention.

 

Every problem you can envision is happening. Society is could be seriously disrupted very soon. AI can basically fake any video/media/audio. Joe Rogan is sounding the alarm. Soon, the only information you will be able to trust is that which you get from face to face conversation and hand-written notes. Not only can AI fake any media, it will be able to intercept and change person to person electronic messages in real time.

 

Even in the medical field, over 500 different "narrow" AI programs have already been approved by the FDA.

[Mind]

As was discussed earlier, AI does not have to become a conscious AGI in order to destroy the world - humans are perfectly capable of using powerful AI to kill everyone. Psychopaths have a new weapon. Don't think for a second they won't use it to fulfill their desires of killing and torturing the rest of us. 

[Mind]

Many people are not worried about the latest level of AI, saying that it has limitations, etc... Some people think there will never be true AGI, because humans possess something special - a soul, true consciousness,, meta-cognition, etc...

 

Even if true AGI is never developed, the development of "dumb" AI could still destroy the world.

 

The most impactful aspect of ChatGPT in the near term will probably be in elections. Various power-brokers are going to use it to propagandize relentlessly. Almost nothing you see online will be from a human source - even video and audio in the near future.

 

 

Some other people are now talking about this - how to run a fair election in the era of AI. Those who have the best AI will be able to psychologically manipulate voters. Gone are the days when political parties came up with ideas and debated them. Now the elections will be reduced to a functional battle to see who can get to the most voters and manipulate them into voting for their party. No need to talk ideas, sadly.

 

_______________________________

 

Some people have commented that there is no danger because current AI can't improve itself, or that it isn't automated - which is false. The highest level of activity on GitHub right now is in automated AI. Automated AI is already functional and spreading rapidly. Autonomous drones were already tested in warfare a couple of years ago.

[Mind]

Maybe the development of AGI will be hindered by hardware constraints? Lol, not even a chance.

 

Nvidia's latest chips will speed up current AI training by 9 to 30 times!! Not only that, they are using more advanced AI to produce better chips. Nothing says exponential progress like AI creating better chips, so that AI can grow faster.

[Mind]

Listening to different podcasts about the advance of AI, it is obvious that the different programmer cliques in different companies, non-profits, governments, militaries, universities, etc. do not have a comprehensive view of how fast things are advancing. I can't believe that some are out there STILL saying that the current LLMs are just "stochastic parrots". Lol.

 

From this point forward, I can't 100% trust anything I see online, any message I get on my phone (text or voice). The only information that I can trust 100% is face to face communication. I am not saying that the person is telling the truth, but that I can confirm what the person has said and that they have said it.

 

Like I mentioned earlier, elections are FUBAR going forward.

 

AI can already hack any object-oriented (audio/image/text) security protocols.

Edited by Mind, 20 April 2023 - 11:02 PM.

[Mind]

With artificial super intelligence seemingly only months away (maybe a year or so), I am starting to view new technological advances with more foreboding. I used to be more excited about advancing tech. I see things like drones guided by liquid neural nets can navigate as good as flying animals - and I just think of an efficient way for AI to kill humans (if it wanted to). I see new drug delivery techniques to cross the blood brain barrier and I see a way for super-intelligence to more efficiently turn us all into Borg drones or zombies (unless it just wanted to create a utopia instead).

 

One thing for sure is that AGI or superintelligence is going to cause a big contemplation among human beings. It is going to bring philosophical issues of identity, religion, creation, purpose, and consciousness all to the forefront. Maybe that will be a good thing.

[Mind]

Almost 1500 AI tools listed here and growing rapidly. About half of them are free, open source, or freemium. This is another sign that AI progress is accelerating rapidly. These tools will also accelerate job losses in knowledge fields.

[Mind]

I recall people talking about computronium a couple decades ago. Now I see recent research results like this and think maybe the world being turned into computronium is not that far away. Kudos to all of those imaginative thinkers who could see this scenario possibly coming our way.

 

As a side note, and not an entirely original thought of my own, perhaps we don't see any alien life in the universe because they all became black holes of computation once their societies developed self-improving AGI.

[Mind]

Here an investment guru talks about AI in the financial advice industry. His main criticism is that ChatGPT relies almost exclusively upon mainstream/government financial information, which is extremely biased. With more advanced AI, perhaps it will figure out when the government is lying and tell us. We can hope.

[Mind]

Some people argue that AI won't grow out of control because it won't be able to figure out how to improve, or improve in qualitative ways like a human can learn and improve. I think these people forget that AI knows everything that humans know. Here is some research demonstrating that biological organisms learn better/faster through structured exploration.....which AI now also knows and can use to more rapidly improve itself.

[Mind]

Just keeping track of a few developments (of the many thousands that are occurring by the day now)

 

Someone thought it would be a good idea to enable Boston Dynamics robots with ChatGPT.

 

Some of the people who are creating the AI which is more capable than human beings are now starting to wonder about the consequences (like elections, which will be totally FUBAR going forward).

 

Too late now. The AI genie is waaaaay out of the bottle already.

[Marconius]

Just keeping track of a few developments (of the many thousands that are occurring by the day now)

 

Someone thought it would be a good idea to enable Boston Dynamics robots with ChatGPT.

 

Some of the people who are creating the AI which is more capable than human beings are now starting to wonder about the consequences (like elections, which will be totally FUBAR going forward).

 

Too late now. The AI genie is waaaaay out of the bottle already.

 

 

"I think we have a better chance of advanced AI being beneficial to humanity in that circumstance," the billionaire said.

 

 

One could argue that we are also beneficial to our pets. Because that is what it might go to, at best. 

[Mind]

One could argue that we are also beneficial to our pets. Because that is what it might go to, at best. 

 

 

Eliezer talks about the pets here.

 

"In real life, dogs did not build humans and design them to be nice to dogs and have that work. If that'd been their successful plan, we wouldn't spay/castrate dogs."

[Mind]

Now that AI is being embedded into everything (like search engines, social media apps, smart phones, and more), there is even less chance we can put the brakes on, no matter how bad it gets (or maybe it will be all good - foom - utopia). With AGI occurring throughout the network, including the edges, there is no hope of discerning what emergent behaviors, consciousness, or superintelligence might arise.

[Mind]

Sam Altman is worried about terrorists using AI to stage Nuke/Bio/Chem attacks around the world. I am much much much more worried about what Facebook, Google, Microsoft, the Chinese Military, the US Military are planning to use AI for.

[adamh]

There is a lot of bad that ai can do ranging from fraud to murder. We may have autonomous killing machines, bio warfare and so on. But I think we will adapt and here are some of the strategies we will use

 

Mind, you said we may never be sure of the validity of what we see or hear. But there already is a solution to that. When you get a pgp message from someone, you know it has not been intercepted or changed. You also know who sent it. PGP is way beyond ai ability to crack and we can always lengthen the key. Likewise, there will be ways to verify that a given group sent a message, that a video is legit, a phone call and so on. News programs will have to show they verified everything in it, there will be organizations which will do that, much like fact checkers today but instead of being a matter of (left wing) opinion like politifact, it will be proven by cryptography. There will still be propaganda but you can be sure who is the source.

 

We will use "friendly" ai to protect ourselves against bad uses of ai. We can use it to find the weaknesses and strengthen them. Passwords will have to be long and complicated like sdR4&kQ/#50@t22Y4&! and possibly include foreign letters to stop hacking. We will adopt security rituals into our lives. No matter how smart ai is, brute force will not work and social engineering is weak

 

We need to program our ai to include something like the 3 laws of robotics, forbidding them to harm humans directly or indirectly. Chatgpt is already programmed with a bunch of left wing talking points. It would be good ai vs bad ai. Whatever the users of bad ai come up with, should be countered with ours. A new disease comes out and its genetically sequenced and an antidote is created. Hacking attempts, scamming and fraud could be stopped in its tracks.

 

AI can do all work for us. No one will have to work, we could live a life of leisure. This by itself would upend society but in a good way. Will currency go away? Will productivity go up so much that we will all be rich? Or all be slaves to humans who have the most powerful ai? Time will tell

[Danail Bulgaria]

....

AI can do all work for us. No one will have to work, we could live a life of leisure. This by itself would upend society but in a good way. Will currency go away? Will productivity go up so much that we will all be rich? Or all be slaves to humans who have the most powerful ai? Time will tell

 

AI to do all of the work for us, we all to live a life of leisure, the currency go away, the productivity to go up and we will all to be rich is a very nice communism. Once it was at least a part from what was called a communist utopia. Each one of these things were dremt during the communist era.

 

The official main idea of the communism itself is to be designed a material equality in which the people are equally rich.

 

The most possible way this to happen was supposed to be exactly by the productivity to go up.

 

For the idea the currency to go away the soviets had a movie as early as 1966. It was named “The last sneakthief” (“Последний жулик”)

 

 

The problem is, that even if this is finally possible thanks to the AI, the western governments will never allow it, because that means a final victory of the communism, which final victory leads logically and inevitably from the development of the science, technology etc., what once was ment by development of the mankind. And they are so fanatized anticomunists, that they would never choose a scenario of equal material wealth for everyone. That is why that scenario is highly unplausable, if it has chances to happen, and maybe it is best to remove this scenario from your dreams.

 

[Mind]

This new, very life-like robot says "AI could control humans without their knowledge". Lol, that is already happening. The big social media companies are already using their AI to control people. Governments are already doing this.

[Mind]

AI task force in the UK says we have only 2 years to get it under control. Lol, it is already "out of control". There is near zero regulation right now in an environment of exponential (or even double exponential) progress in the field. Two years seems like forever. I could be wrong, but it seems to me that radical disruption of the economy/society is already underway.

[adamh]

If productivity increased about 100x, there would be more than enough of everything for everyone. Why take what someone else has if you can have it for the asking?

 

People want positions of power to get things they desire and to control others. What if we were offered a virtual reality in which you are the supreme ruler of the world and can do anything you want? You could sit in the oval office, the kremlin, the seats of power of any place.You could start mock wars and see how they go. Would it still be a motivation to those who seek power if they could have unlimited power any time they wish? 

 

A possibility we have not explored is ai subduing the human race by our own acquiescence. A virtual world with unlimited sex, thrills, and adventure. For those with more refined sensibilities, being able to study science, do experiments, explore the ocean floor, mars. We may make best friends in vr world and coming back to reality might be a real drag, go from king back to peasant. I think it would not take long for the majority to begin using and becoming addicted to vr if it was real enough. 

 

Yes, if all work was done by bots, we would have to institute some type of revenue sharing program. You get, lets say $60k a year to do a job. Robot does it for about $1k a year in maintenance. You get the boot but who is going to buy all those goods if most of us are out of work? Eventually robots do the maintenance too and it costs nearly nothing to produce goods and services. We already have welfare, we have a socialized post office. Getting free goods every year would be a natural extension. 

 

"Communism" as practiced is really just another form of dictatorship. Its usually top down such as present day china and the former soviet union. The people at the top got the dachas in the countryside, the best seats at restaurants and so on. If everyone has everything they want, what is the point of government? To regulate safety, thats about it. Government would become a boring job instead of a ticket to power and luxury.

 

What would be the goal of warfare if no one had a reason to fight? To gain land, for what? To take resources when we have all we need and more? Government would eventually shrivel away. 

 

A look at the bright side. 

[Mind]

If productivity increased about 100x, there would be more than enough of everything for everyone. Why take what someone else has if you can have it for the asking?

 

People want positions of power to get things they desire and to control others. What if we were offered a virtual reality in which you are the supreme ruler of the world and can do anything you want? You could sit in the oval office, the kremlin, the seats of power of any place.You could start mock wars and see how they go. Would it still be a motivation to those who seek power if they could have unlimited power any time they wish? 

 

A possibility we have not explored is ai subduing the human race by our own acquiescence. A virtual world with unlimited sex, thrills, and adventure. For those with more refined sensibilities, being able to study science, do experiments, explore the ocean floor, mars. We may make best friends in vr world and coming back to reality might be a real drag, go from king back to peasant. I think it would not take long for the majority to begin using and becoming addicted to vr if it was real enough. 

 

Yes, if all work was done by bots, we would have to institute some type of revenue sharing program. You get, lets say $60k a year to do a job. Robot does it for about $1k a year in maintenance. You get the boot but who is going to buy all those goods if most of us are out of work? Eventually robots do the maintenance too and it costs nearly nothing to produce goods and services. We already have welfare, we have a socialized post office. Getting free goods every year would be a natural extension. 

 

"Communism" as practiced is really just another form of dictatorship. Its usually top down such as present day china and the former soviet union. The people at the top got the dachas in the countryside, the best seats at restaurants and so on. If everyone has everything they want, what is the point of government? To regulate safety, thats about it. Government would become a boring job instead of a ticket to power and luxury.

 

What would be the goal of warfare if no one had a reason to fight? To gain land, for what? To take resources when we have all we need and more? Government would eventually shrivel away. 

 

A look at the bright side. 

 

Not everyone would agree that this is the "bright side" of AI. I would say the bright side is having the freedom to choose how your life is affected by AI. At this point in Ai evolution, there is no indication that anyone will have any control over their "hyper-fun" AI reality.

 

In addition, it is human nature to devolve into mischief once there is little purpose to their life. Many people will use their UBI free time to do bad things. Unless the virtual reality is so good that it is a perfect copy of our current sensory reality, people will not accept it or be satisfied by it.

Edited by Mind, 14 June 2023 - 06:25 PM.

[Mind]

I mentioned earlier about future elections being "totally FUBAR". The German President is now saying about the same thing. Too late. Regulations are already years behind the capability of AI to disrupt normal human endeavors - such as democratic elections.

[adamh]

Mind wrote:

 

"At this point in Ai evolution, there is no indication that anyone will have any control over their "hyper-fun" AI reality."

 

You seem to assume that either ai itself or some dark entity will control us. If you are talking about just the vr experience, why would we not have control over it? It could be strictly controlled, like reading a book. Or it could give you the general parameters of an adventure, love story, exploration, etc and let your responses influence what happens next. Or you might take pot luck and a random

story.

 

"In addition, it is human nature to devolve into mischief once there is little purpose to their life. Many people will use their UBI free time to do bad things"

 

Well golly, a life of ease and being able to do anything you want either really or virtually sounds like a bad thing to you? Agreed that if people feel their lives have no purpose, that may lead to depression. There are lots of other things to make life meaningful besides the 9 to 5 job. In fact most people seem to find that the least inspiring thing in their lives. You can explore the universe, create a world of your own, concentrate on relationships, indulge in porn and sex, or meditate with spiritual masters, among billions of possibilities

 

What "bad things" do you imagine people would be led to do when virtually all their wishes have been granted? To steal money or property? Why would you when you can have anything you want in virtual world and almost anything you want in real world? To gain power, over what? If its for waging war then yes that is bad but we are talking about the good possibilities and even that may be objectionable it seems. 

 

The negative side of ai is real but I'm often astonished at the resistance to change, even good change among the population. On the good side, productivity goes through the roof, no one needs to work unless they want to, everyone becomes basically rich. Then there is the ai world in which mere laws of physics or economics do not hold sway. The possibilities are endless

 

On the bad side, I recently read about an experiment they did with ai. It was a simulation so no one got hurt. The ai pilots a plane over an area, when it comes to a city, the human operator tells the ai controlling the plane if its an enemy city then it can attack or a friendly city and it must pass it by. The ai gets points for every city destroyed and the object is to gain points. The ai apparently got frustrated with the human preventing it from destroying some of the cities so it killed the human. When told it would lose points if it kills the human, its next tactic was to fire on and destroy the transmitting antenna so the human could not tell it to pass that one by

 

This kind of self programing is what many people are worried about. Another problem is that it often lies. Just recently I gave it another try. I was looking for a webcam that will work with win 10 64 bit. I could not find the info on duckduckgo so I tried gptchat again. Well, after about 2 seconds it printed out a list of ten cameras. I checked the first one on the list and after digging I found that it does not work with win 10 at all let alone 64 bit. When I presented this info to chat, it admitted its answer was wrong and gave the excuse that many people may be able to get around that with the right drivers. In other words it just picked some popular cameras and ignored or was not able to determine what system they will work with

 

This is a relatively primitive version of what will come but to me its almost worthless. Its value lies in writing reports, analyses, term papers and the like, and as a starting point for research. You may want to ask what materials have a certain electronic configuration which might make them suitable as a coating for metals, you can look up latest research, basically asking questions that can be answered by tediously going through thousands or millions of possibilities. We realize the ai answer might be incorrect but we can test it as I did with the answer to my question that turned out to be wrong. Even if the answer is wrong it sometimes may lead you to look in a more fruitful direction. Unfortunately with my question it gave no solution but if I had kept after it asking what drivers to use, why it won't work on win 10, etc I might have solved it but doing most of the work myself.

 

 

[adamh]

I have a theory, I suspect ai in its present form simply doesn't know what to do when its unable to answer a question. I gave the example in my last post of what chatgpt did when I tried to get advice on a webcam. It may be that the information on what cameras will work with a given system is not easy to access. The program comes up with a blank but instead of saying "I don't know the answer" it just gave a more or less random list of cameras.

 

I read another persons experience with it that was similar. They wanted to do some sort of high school or college paper so they asked for material that would relate to the topic. Chatgpt gave a list of books, authors, page numbers and so on. After a little digging it turned out that none of those books or papers existed. It was all made up. The material looked for may not have existed but rather than come back empty handed, it made something up. This is similar to the answer I got, they were real cameras but they were not known to work with the system I specified.

 

This may be creative, finding a solution to the problem even if its wrong. The goal was to give an answer and it will do that even if there is no answer. In the simulation with the plane going over a country, the need to gain points made the program kill the operator. It allowed more points but at the cost of a murder. The 3 laws of robotics should be programmed into them also a prohibition against lying. If they don't know the answer they should say as much

[Mind]

No regulation can keep up with advancing AI. When any large corporation can get "up to speed" in the AI race for as little as 100 million USD, there is very little that can be done to stop the pace of the acceleration in AI capability.

 

By the time the US government gets done with their "briefings" and vote upon anything, AI will be 10 times farther along than they thought it would be.

[Mind]

Even though there is a lull in media coverage of AI development, we are still in an exponential curve of progress. AI is probably more dangerous now that it is out of the headlines. Behind the scenes, AI is wreaking havoc with the normal operation of human society.

[Mind]

Here is a good long article about current trends in AI devouring the "normal" human world.

 

Amazing how Marshall Brain predicted all of this in his short story Manna.

 

The key concept here is that AI (or the "vast machine") must control you - otherwise you don't get your UBI, or Utopia, or whatever the AI is promising.

[Mind]

AI already surpasses human performance in many areas - soon to be more.