Skip to main content

Week 1 [03.10-09.10.2016] Can we build AI without losing control over it?

Image
By

Watch the presentation Sam Harris: Can we build AI without losing control over it? at https://www.ted.com/talks/sam_harris_can_we_build_ai_without_losing_control_over_it#t-22278 and present your opinion on it/discuss it. 

Comments

kondrat said…
It is a terrifying vision of the future, but in my opinion if we are able to keep our curiosity at adequate level there is no risk at all. We can build real artificial intelligence without equipping it without self-consciousness.
5 October 2016 at 13:26
pawelsienkiewicz said…
I think it would never happen. There should be always a backdoor, left in software that let humans turn off AI-things.
5 October 2016 at 18:42
Unknown said…
As long as human is involved in programming AI there will always be a risk of serious bug which can be dangerous for our species. But as you said if we keep it on adequate level we will still have a chance to fix it and prevent us from disaster.
5 October 2016 at 20:01
Jakub Finke s12590 said…
Artificial intelligence, intelligent machines can help us in life. Life is better but these machines can't be more intelligence than we because man must master the machine and not vice versa
5 October 2016 at 20:18
Łukasz Domeradzki said…
I wrote interesting article about it in the past, if somebody is interested into that particular subject, I'm leaving a link:? http://onlineclasspjwstk-malgorzataswierk.blogspot.com/2015/12/week-10-0712-13122015-artificial.html

Personally I still have the same opinion - humans are capable of creating a thing that will be smart enough to improve itself, and at some point we may simply lose control over it, not necessarily physically, but logic-wise, as our AI might reach the point when we'll no longer understand it or it's actions.

I don't think it will happen anytime soon though. It doesn't look like human race is trying that hard to exterminate itself, and AI progres is rather experimental than practical. Creating a decent AI that can learn and improve itself is difficult task, but possible. Creating AI being capable of outsmarting us will be the last achievement of our race :).
5 October 2016 at 20:29
Dajana Kubica said…
I think it will not happen, because people have imagination. It is primarily thing that separates our intelligence from artificial intelligence. If a robots would want to destroy our civilization, as was mentioned in the video, it is only because the man who created them, taught or allowed to be taught the desire to destroy the human world.
5 October 2016 at 21:22
Unknown said…
Yes, but think about it this way: As it was said in the video we will keep improving what we have built, we will still making AI efficiency better and better and one day it will become so smart, so intelligent that even you will have some sort of a 'backdoor' , the AI will notice that and block it. When it's self-consciousness will become high enough it will prevent every possibility to turn it off. Just like humans prevent themselves from dying, the same highly developed AI will do. I just imagine that development in such major is very dangerous for us. There can be always some bug, some mistake, we are only humans. And if something goes wrong with the artificial intelligence...We should be very careful with that.
6 October 2016 at 11:58
pawelsienkiewicz said…
That's why we should always have a backdoor. I don't think that AI would outsmart humans.
6 October 2016 at 15:10
Unknown said…
There already has been a case of a robot mentioning "Human zoo" existing in the future. Here's a link to one of many articles mentioning it:

http://glitch.news/2015-08-27-ai-robot-that-learns-new-words-in-real-time-tells-human-creators-it-will-keep-them-in-a-people-zoo.html

Since this robot was not programmed to lie, we can safely assume he was telling the truth. So yes, the AI will most likely take over. We should remember to be nice to robots every day, that one mentioned that "Human zoo" was only for the kindest of us.
6 October 2016 at 22:40
Maciej Spychała said…
As Sam said in his presentation, scientists agree that creating AI that could somehow be dangerous to humans won't be the problem of our nearest future. The problem is that we don't really know what's our timeframe to start worrying about it. He rightly pointed out that it should be already a topic for serious discussion.

The problem of dangerous AI is that it doesn't really need to outsmart human beings or be created by someone with a malevolent intent to become a danger. For example AI that teaches itself to be considerate of the natural environment could possibly come to a conclusion that humans are the biggest threat to the planet and that it has to do sth about it.

Also the task of creating the sort of "science-fiction type" all-encompassing AI is a task of implementing some sort of a moral system. The most obvious problem with that is that such system could easily be as flawed as the understanding of morals by its creator (and I'm still not talking about anyone that would genuinely mean to inflict harm on others by the actions of his AI).

PS I really recommend Sam Harris' books and podcast (not only about AI), quite interesting stuff to say the least.
7 October 2016 at 00:46
Dominik Ciborowski said…
Its very interesting vision and I agree that it won't be the problem of nearest future. Unfortunately this 100, 200... years as he also mentioned is not so long period and will go by very quickly. The worst thing is that we don' t really know when we will go a step too far in developing AI.
We always should have in mind that we cannot treat robots as humans. Maybe one time we would have big problem to show difference between humans and computer.
I have watched very good TV show called "Humans". It is about robots with consciousness applied by their creator. I highly recommend this show as it opens mind to AI topic.
7 October 2016 at 23:44
Maciej Główka said…
Sam Harris vision of future is really interesting and also quite terrifying.
Firstly, we should know how AI is working nowadays. When developers want to teach AI what is a cat and how it looks, they have to put into AI millions of cat photos. Artificial Intelligence isn't browsing internet and learning about cats by itself, as ,in my opinion, many people believe how it looks like now. That's why it is very important to show people what are possibilities and limitations of current AI. Luckily, some of the biggest technology companies such as Facebook, Google decided to team up and educate people about AI. Here is news about it:
http://www.theverge.com/2016/9/28/13094668/facebook-google-microsoft-partnership-on-ai-artificial-intelligence-benefits
I agree with Sam Harris about our near future. We should start thinking about some laws narrowing down use of AI in specific areas, for example in military. We should always remember, that main purpose of AI is to make our lifes easier.
8 October 2016 at 12:11
Piotr Basiński said…
I agree with your point of view. There is also a very good independent film about artificial intelligence, which is very interesting. Title of this movie is "Ex Machina". For several hundred years from now someone will see whether it was worth to develop AI or whether it was a big mistake.
8 October 2016 at 19:58
Michael Gutowski said…
We can not only build AI without equipping it without self-consciousness, many neuroscientists claim that AI robots will never have consciousness like humans. It is fair to say that we still haven't figured out how brain works and even if we do it may turn out that due to the technical limitations we won't be able to recreate human brain. The thing which Sam Harris didn't mention is that apart from intelligence there is emotional intelligence (self-awareness) and creativity. The last two are probably the once that are most difficult to understand and, at least for now, impossible to simulate with computers that we know. However, there is no doubt that AI does not need to be self-aware or creative in order to become a danger to humans and at this point I can agree with Harris thoughts. I also liked the part about Justin Bieber - nice one :D
8 October 2016 at 20:03
Kacper Zaremba said…
I have to agree with you. Saying AI is intelligent is huge overstatement. Today Artificial Intelligence is learned by doing the same tasks millions of times. It doesn't have anything in common with self-consciousness or intelligence. Currently there is no way to lose control over AI, unless we program it to do so.
8 October 2016 at 21:22
Unknown said…
I my opinion advenced A.I. powerd by the computer, that is more powerful then our brains is not the danger, but the anwser to all the questions we ask ourselves today, like what is going on, deep in our minds? What is depression or much worst brain diaseases that we are facing right now, on even much bigger scale than it was not so long ago. I see the danger in human brain simulation, which connected to the network could make a lot of problems, becouse our brains, in many cases are not perfect and it is possible that it could go crazy, making a lot of mess. Artificial inteligence made by microsoft some time ago went complitly "out of her mind" into extreme right wing side ideology sometimes even nazi, but it was learning especialy from internet comments that are mostly stupid and unthinking. I agree, that for the proper development of A.I. we need kind of Manhatan Project. Then, most likely we could build it without losing control.
9 October 2016 at 10:19
Dominik Ciborowski said…
Right now we should not call computers intelligent. They are "intelligent" because they are learning some tasks billion times and that's not what we humans call intelligence.
9 October 2016 at 15:59
Unknown said…
Personally I think that we are very far from creating a true AI. Every kind of intelligence we know (basically only human intelligence for now) requires a purpose and is also defined by its limitations. In humans the only purpose of intelligence is to help us survive and reproduce with everything else being side effects. As animals we have a lot of things hardcoded into our brains genetically. Most of those things change extremly slow. If we were to create an AI we'd have to hardcode a lot of basic instincts to it. Those instincts would define and limit the AI. Without them it would probably get into very chaotic state and turn itself off.
9 October 2016 at 19:36
Robert Maciejewski said…
What if the dangerous AI was made by some maniac or someone that made a huge mistake developing it ?
9 October 2016 at 20:28
Robert Maciejewski said…
This comment has been removed by the author.
9 October 2016 at 20:47
Krzysztof Skorupka said…
I think being outsmarted by intelligent machines is not the only problem here. More real threat is that people will be less and less intelligent because machines will do more and more of thinking for us. One can observe nucleus of the problem among today's youth - many of them have serious gaps in knowledge about basic matters (Why do I need to remember this? I can google it in a few seconds), big problems with basic math (Calculator app for mobile?) and so on. I'm afraid that with the development of AI this problem with get worse. It may be that one day, we will realize we can no longer function without smart machines or gadgets. Being fully dependent on such devices seems to be much more tangible form of "losing control".
9 October 2016 at 21:29
Michał Pycek said…
In my opinion it is possible to create an AI without losing control of it, however only if we take it seriously and spend some time on analyzing all of the possibilities of results.
If we are aware of them, theoretically there should be no danger, however it is difficult to recognize all of them.

9 October 2016 at 22:39
Krzysztof Lichota said…
I partly agree with you. We are far from AI we know from sci-fi movies.

One of the most developed branch of computer science is machine learning. It helps to solve problems like handwriting recognition, autonomous vehicles driving etc. that can't be hardcoded. Computer learn by itself how to do such things and in majority does it better then human. If future AI would be able to learn, how we can be sure that it won't find a way to overcome limitations we gave to it?
9 October 2016 at 22:40
Unknown said…
In my opinion we will never build a perfect artificial intelligence, because the perfect AI should be able to lie. If AI could lie, how the programmers would be able to identify whether the computer is lying or just giving the wrong result?
9 October 2016 at 22:48
Wojtek Kania said…
Nowadays the technology is already here. But all this fear, which was created by science fiction movies, is invalid. There is a difference between intelligence and awareness. We have no intention of teaching robots that they are robots, what does it mean „to be” and why we exist. A better description would be „artificial thinking”.
9 October 2016 at 22:53
Moode said…
Why not, if have asked someone 200 years ago that,
Do you think you will be able to travel to another continental in a few hours, 
the answer will be No. and it’s something normal right now for all of us.
I think this the same matter.
but when?? No idea
9 October 2016 at 23:12
Post a Comment

Popular posts from this blog

Week 3 [23-29.10.17] Clones of NES around the world

By
I ’m sure everyone of you heard about Pegasus (Gra telewizyjna) console. I’m also pretty sure majority of you heard about NES and know very well that Pegasus is in fact NES clone (well, actually it’s Famicom but let’s not get into details) distributed by Hoop Cola guys. But do you actually know how many clones NES had in the past? Well, neither do I but I’ll try to present you some of the most popular clones around the world (excluding China no-names you can buy on Aliexpress even now). Short note for people born in previous age: NES is video game console released in 1983 in Japan and 2-3 years later in EU and US. By many it’s considered the best video console of all time. Pegasus Now, let’s get back to the topic. Of course, the first position in my list is Pegasus I mentioned before. Our precious Pegasus It hit the market in the early 90’s thanks to Dariusz Wojdyga and Marek Jutkiewicz, who are currently on list of 100 richest people of Poland. It was dist...
58 comments
Read more

Week 4 [04.11-10.11.19] Dark matter

By
If you don't live in a big city in Poland, you can observe many stars at night. Did you ever wonder how would look our sky in the next millions of years if we would count only on existing proven astrophysical theories and calculations? It would be absolutely empty. All  celestial bodies  in the galaxy are rotating around the biggest black hole in the middle of galaxy (in Milky Way this black hole is 6.4  billion times heavier than our sun)  with such speed that the gravity generated by galaxy’s observable matter could not possibly hold them together. This leads us to  conviction   that something we cannot see is at work and is giving galaxies extra mass, generating the extra gravity that galaxies need to stay intact. Something, that will ensure us that our sky will remain the same as it was a million years ago. The official name of this strange and unknown matter is the “dark matter”. Unlike normal matter, dark matter does not interact with the electroma...
73 comments
Read more

Week 12 (12.01-18.01.15) Are you an early bird or a night owl ?

By
Owls are nocturnal creatures. They’re wide awake at night and they sleep during the day. If this sounds like bliss to you, then, like about 20 percent of the population who find themselves most active at around 9 pm, you may fall into the same category as our feathered friend. Night owls often have difficulty waking up in the morning, and like to be up late at night.  Studies of animal behaviour indicate that being a night owl may actually be built into some people’s genes. This would explain why those late-to-bed, late-to-rise people find it so difficult to change their behaviour. The trouble for night owls is that they just have to be at places such as work and school far too early. This is when the alarm clock becomes the night owl’s most important survival tool. Experts say that one way for a night owl to beat their dependence on their alarm clocks is to sleep with the curtains open. The Theory is that if they do so, the morning sunlight will awaken them gently and natura...
60 comments
Read more