No announcement yet.

Answers to Some Questions About Moral AI

  • Filter
  • Time
  • Show
Clear All
new posts

  • Answers to Some Questions About Moral AI

    Recently I was asked to respond to these questions on Facebook. These are not so much answers as jumping off points to further discussion, which are often better than attempts to definitively solve problems humans have been grappling with for millennia.

    ::AI quest::

    How can we, as humanity:
    1. Use AI in a positive way to change the THINKING process/patterns of destructive people with influence (*)

    One of the themes I repeatedly encountered while designing the initial data set for EthicsNet was balancing human autonomy with the (eventually superior) decision making of a machine. This could range from discouraging or stopping an illicit or indisputably self-destructive behaviors to preventing access to information the synthetic intelligence has deemed incorrect, hateful, or perhaps just harmful to the reader’s mental health. An AI programmed to blindly optimize human happiness would block news articles that would upset its owner.

    AIs are already actively altering our perceptions of reality, but for the most part in a manner that reinforces our current interests, beliefs, and biases. A thoroughly utilitarian machine bent on optimizing the happiness of humanity as a whole will almost certainly engage in some form of behavior modification and, with the advent of neurotechnology, directly influencing thought patterns - unless safeguards are put into place.

    2. While leaving them with the freedom of thought, speech, and decision-making?

    By this I assume you are referring to the absence of direct coercion or beaming thoughts directly into people’s heads (as fun as that might be). Still, this is somewhat ambiguous. After seeing a few hundred advertisements over the course of a day are human beings left with the freedom to make educated decisions, particularly after they go shopping after a depleting day at work and are less capable of resisting their impulses? Are shoppers free to choose when supermarkets are laid out to make sure we are lured in by shiny produce and made to traverse the length of the store if we want milk, cheese, or yogurt?

    Or do we suddenly become prisoners to our desires when targeted ads begin flooding their Facebook feeds and mailboxes? Or is it when we’re shown sexually desirable people on a big screen TV? The word “manipulation” has a negative connotation, and rightly so, but enormous chunks of human interaction can be classified as consciously or unconsciously manipulative - from making children equate brussel sprouts with limitless physical beauty to subtly (or not so subtly) hinting at the health consequences of smoking to a friend who probably is already aware of them, we almost constantly attempt to sway other sentient beings to the side we believe is best.

    Aggressive campaigns against drunk driving have been credited with the drastic reduction in alcohol related car accidents. Using powerful images gleaned from the aftermaths of such accidents is certainly a form of “manipulation”, but it is for the common good. We either influence ourselves and others for the better or for the worse. Influence becomes manipulation when the influence is being used for purely self-serving ends. It hearkens back to the second expression of Kant’s categorical imperative: treat all others as ends and never as means.

    What’s important is deriving a consensus on what is best, but also figuring out a range of what is acceptable. Determining what is true and what is not can be done through collective intelligence, harnessing the insights of many people produce better predictions and better.

    3. So that they could benefit to their communities (and themselves) instead of tearing them down?

    The internet has many communities that are not doing anything particularly productive and are, perhaps, incubating the person or group who may destroy civilization as we know it with the increasingly destructive weapons made increasingly accessible by emerging technologies like synthetic biology, nanobots, and AI. Radical religious and ecological groups are obvious examples of individuals with apocalyptic worldviews, but the threat a single hateful person poses should be abundantly obvious based on the radically misanthropic motives of multiple mass killers in the United States.

    The issue my questioner is pointing to is another careful balance, one which every government that allows freedom of speech is intimately familiar. While libel, slander, and yelling “fire!” in a crowded room have generally been seen as violations of the contract between the speaker and the state, the lines of contention are increasing as the stakes are rising ever higher. So much so that it has become necessary to begin formulating a common code that is flexible enough to accommodate differences but stringent enough to keep as many people as possible on the straight and narrow path.

    Last edited by Adam; 12-13-2018, 09:11 PM.