Will Artificial Intellegence Make The Human Race Extinct?

by Brokeback Watchtower 46 Replies latest social current

  • Island Man
    Island Man
    "I believe AI will just do what it is programmed for"

    That's just the problem. The kind of AI currently being developed and experimented on, isn't programmed to do a specific task. It learns how to do its task by being fed lots and lots of data. It finds patterns in the data and creates and tweaks its own rules for processing data based on those patterns. It is like a child that learns to talk and learns the meanings of words by listening to others talk and the context in which specific words are used. It's learning the way we learn, albeit in a far more rudimentary way - for now.

    A lot of people get hung up on the issue of consciousness and self awareness. But these aren't needed for AI to be a danger. All that's needed is superintelligence. AI has already proven itself superior to us in basic strategic reasoning used in games.

  • ttdtt
    ttdtt

    Well something better than us has to come around eventually.

  • ttdtt
    ttdtt

    What Machine Learning Robots will really do (not take over the world) but take about 70% of all jobs.

    That will be when the Sh#t will really hit the fan, and massive unemployment will be the order of the day.

  • redpilltwice
    redpilltwice

    Thanks for your input Island Man. I appreciate different approaches of the matter, that's what makes it intriguing. We are uncertain about the amount of control we can have in the future and we're talking about huge things that certainly will influence our lives. Therefore I also agree with ttdtt because we develop AI to help us and perform certain tasks more efficient, not because scientists want to become digital alchemists and deliberately open pandora's box.

    It learns how to do its task by being fed lots and lots of data. It finds patterns in the data and creates and tweaks its own rules for processing data based on those patterns. It is like a child that learns to talk and learns the meanings of words by listening to others talk and the context in which specific words are used. It's learning the way we learn, albeit in a far more rudimentary way - for now.

    How did it "learn" to learn in the first place, by itself or by intelligent human input?

    How does it "find" without acting within the laws of programming? Does it "want" otherwise? No, it has no will or desire.

    I believe that we can only reason and try to predict from experience and so far our experience is that codes do not write themselves, they are written by us. A defect or virus will not improve its working, a program can even be shut down because of it. So why expect such a huge positive "mutation" in non-living matter/digital machinery?

    A lot of people get hung up on the issue of consciousness and self awareness. But these aren't needed for AI to be a danger.

    True, and that's why I mentioned the millenium bug. There was potential danger (so we were told), and no one will claim it had to do with some AI being self aware.

    All that's needed is superintelligence. AI has already proven itself superior to us in basic strategic reasoning used in games

    I disagree, it is not superiour to the sum of our collective intelligence in any way. Algorhythms are not capable of learning, they can solve problems because they are programmed to perform certain tasks, such as calculation and data processing and darn, they do it superfast! But let's take chess. It has no freedom to create new rules, it can only search all options, no matter how many. It can even detect patterns of the human player because it can have access to all past and current data that has been stored during its gameplay history. None of those options is unknown to our collective knowledge of chess. It is not intelligent in the sense that it can create something "new" in the world of chess.

    I know that people like Stephen Hawking fear a 'rise of the robots' in the future, but so far every smart machinery that has been invented by humans has been improved or replaced because it became outdated. So why would a machine or computer suddenly be "miraculously blessed" with the will and intention to jump ahead its own purpose, become smarter and do its own thing?

    Furthermore, wouldn't any attempt to create, test and develop AI only take place under strict and controlled circumstances/protocol? How could any certain artificial attempt to think for itself go unnoticed by the scientists involved? Wouldn't they rigorously study everything that is happening in detail before going any further?

    Like was said in the OP, I also don't worry too much but for my own reasons. I do NOT fear a human extinction because of AI.

  • scratchme1010
    scratchme1010
    Will Artificial Intellegence Make The Human Race Extinct?

    No. Not at all. What it can make is either our lives better or more difficult. Like everything else, AI can be use for the good or for the bad.

    What I see is is AI being used in ways where companies suck up all the information they can about us, to use it for their purposes.

    In a separate post, I posted about a conference I went in Las Vegas, recently. Two of the main speakers were professional Futurists (I didn't even know that such profession exists), and their talks were about using AI as well a advances in the approaches on how it is developed (as opposed to how it was originally thought that it would work).

    They presented things like the Google toilet, capable of detecting pregnancy, stress hormones, alcohol, infections, etc:

    They presented a computer generated pop star who gave concerts and interviews:

    And a technology named VGG-Face deep learning algorithm that can detect people's sexual orientation with face recognition technology:


    I like the way one of the speakers finished her speech, with this quote:

    "Technology generates as many new problems as new solutions. Most of the problems we will have n the future are going to be caused by the technologies that look like solutions to us today".

  • WeatherLover
    WeatherLover

    @redpilltwice

    I also don't think that ai will make us extinct, but there has been remarkable progress in ai over the past few years.

    "Over the course of millions of AlphaGo vs AlphaGo games, the system progressively learned the game of Go from scratch, accumulating thousands of years of human knowledge during a period of just a few days. AlphaGo Zero also discovered new knowledge, developing unconventional strategies and creative new moves that echoed and surpassed the novel techniques it played in the games against Lee Sedol and Ke Jie."

    (Bold is mine.)

    https://deepmind.com/blog/alphago-zero-learning-scratch/

  • Brokeback Watchtower
    Brokeback Watchtower

    Another possibility I see is that AI and humans evolve together and become one and the same perhaps a Homo evolutis.

  • redpilltwice
    redpilltwice

    Thank you too WeatherLover, I've read the article.

    AlphaGo Zero also discovered new knowledge, developing unconventional strategies and creative new moves that echoed and surpassed the novel techniques it played in the games against Lee Sedol and Ke Jie."

    What exactly is this new knowledge other than extremely fast calculation and trial and error of every possible move that could've be done? If the data is build up by playing against itself, it has to know the rules before it can decide what will be a better move and what isn't. By being able to play a zillion games in no time and ongoing trial and error it seems to have reached perfection. That's cool, having the potential to scan and evaluate every move that could've been made by anyone anytime anywhere. Of course, the outcome will surprise us. We are individuals, not knowing what all other players know. Humans also have emotions and sometimes do things by intuition, not knowing where and when the same situation occured in the entire history of this game. This "cold" approach and the relentless way the algorhythm is programmed gives it a unique helicopter view that is unparalleled in exhausting all the meta-data.

    Impressive? Yes, very! But I remain sceptical because it still didn't go beyond what is has been programmed for... play against yourself, remember and store the data, improve by trial and error and become better and better and better up to the point of perfection. No human being will be able to have access to so much data at once by using memory, but we can build machines that do it for us because we program them to do that specific task, they have no free will.

  • ozbrad
  • Brokeback Watchtower
    Brokeback Watchtower

    Hey I was just thinking about John Nash's game theory when reading your post.

    https://en.wikipedia.org/wiki/John_Forbes_Nash_Jr.

    Maybe some one that studied this subject could tell us the out come?

    https://link.springer.com/chapter/10.1007/3-540-45634-1_4

    https://www.quora.com/How-are-game-theory-and-machine-learning-related

Share this

Google+
Pinterest
Reddit