up a level
post article
search
admin
about
rdf
Karl's Fiction Writing Site
|
| Unconscious Computing: the Know-Thyself App |
Posted by Karl on Thursday March 02, @12:00PM
from the dept.
A recent issue of Science (subscription required) talks about research (described here for free) suggesting that we humans make better decisions about complex issues when we let our unconscious mind do the work. I think most of us are willing to give up conscious control to this side of ourselves; after all, our conscious and unconscious minds are in the same boat. But what if there were an 'artificial unconscious' as well? Would you be willing to give up control to that, even if it were convincingly shown to be in your own best interests?
Let's get the B-movie idea out of the way first: one of our interests is having the ability to override our unconscious patterns using conscious thought. So let's not spin off into a scenario where computers make decisions for us and don't allow us to make any for ourselves. That's not an interesting speculation.
What is interesting is to imagine a form of AI that augments what the unconscious mind already does best. Gareth Cook describes that talent in the Boston Globe article linked to above:
For example, when people buy a house, they tend to put too much emphasis on its size, and not enough on their commute every day, he said. When working through a decision consciously, the mind has a tendency to focus on factors that are easy to articulate -- like the number of square feet -- at the expense of other factors that are hard to put into words.
Now, we tend to think of the conscious mind as being the executive function and gatekeeper, even to the point of often assuming that the conscious mind has sole command of the body's perceptual and motor systems. This is quite untrue, in fact the reverse is true: the way our eyes move and track is mostly under unconscious control, for instance.
Would it be too shocking to consider that when the unconscious mind works (as in the house-buying example above) it has full access to the same resources as the conscious mind? --Including perception, language, conversational strategies etc. We may, for instance, engage in conversations with other people about the house-buying project that are part of a systematic research programme our unconscious mind is executing. We just think it's 'been on our mind' or that we're 'curious about this and that side of it'. Maybe those phrases are codewords indicating that our conscious mind is being sidelined in favour of something smarter.
For the purposes of discussion, let's grant that. (There is in fact considerable evidence that the unconscious mind can create and execute fairly elaborate plans without conscious awareness being involved.) Here's the question: is it the best use of everybody's time for computers to be designed to be used only by the conscious mind?
Let's rethink computer design. I've long said that if you have to actually use a computer--as in interact with it--then it's not doing its job. Computers are supposed to do stuff for us, not make us do more stuff. Computers are certainly useful to the conscious mind--but they're a lot more useful when they're off doing stuff by themselves, for instance routing my cell phone calls when I travel to another country. They are more useful, in other words, in proportion as they act like the unconscious mind. Yet in our direct interactions with them, they are geared entirely towards connecting with our conscious mind.
In my novel Lady of Mazes I describe a 'values-driven interface' to an augmented-reality system known as inscape. That interface is an example of a direct interface to the unconscious mind rather than awareness: it filters the users' perceptions according to their values, their sense of right and wrong, justice and purpose and so on. The interface is not consciously 'steered' by the operator; it's kind of like having a vast and continously-updated preferences profile attached to your web browser. (Cory Doctorow and I worked on a primitive version of this in the OpenCola project.) But it's much more than that.
Assume that your unconscious mind is clever, observant, rational (not Freudian) and has your own best interests in mind. Why not give it better tools? --In particular, why not let it extend itself through the creation of software agents that are themselves also clever, observant (in their own ways) and have your own best interests in mind? Let these agents determine your interests and values by watching your reactions to real-life situations, rather than by polling your conscious mind for the details of a possibly-misleading self-image.
Scary? Well, if this system of agency is able to translate your unconscious behaviour-patterns into an accurate model of your values and aims, it will also be able to report its findings back to your conscious mind. So the killer-app of unconscious computing could be a translator for the Self--a software and hardware package capable of telling you who you really are, objectively speaking. What appears to really motivate you, where your unconscious behaviours are at odds with your conscious self-image--parsing and reporting this would be a snap for a system designed to interface with the unconscious.
(That interface, by the way, could be a spin-off of biometric and cognitive science research. For instance, if your PC watches your eye-movement patterns and flushing reaction it can build up a model of who you trust and who you're attracted to. So can the government's computers--so it behooves you to keep ahead of them by knowing yourself first.)
Ultimately, though, the unconscious-computing system should be as invisible as our biological unconscious. In its behaviour, it's probably best to think of it as a small army of fanatically loyal well-wishers, who genuinely care about you and tirelessly act for your benefit.
Just like the systems of your own unconscious mind.
<
|
>
|
|
The Fine Print: The following comments
are owned by whoever posted them.
( Reply )
|
Re: Unconscious Computing: the Know-Thyself App
by Max Kaehn on Thursday March 02, @07:03PM
|
I’d be happy just to have a metabolic monitor that can trigger food cravings (or suppress hunger messages) based on what my body currently needs for my daily modern life. My current wiring apparently thinks my day job involves killing mammoths with spears, that I need to be prepared to survive fevers and famines, and that I don’t really need to live past 40 anyway, and programs my taste buds accordingly. :-)
|
[
Reply to this ]
|
Re: Unconscious Computing: the Know-Thyself App
by Ted Chiang on Monday March 06, @12:52AM
|
What does it mean to say that the unconscious mind is not Freudian? That your unconscious mind doesn't have desires that differ from your conscious desires? That sounds like rejecting a fundamental property of the unconscious. One doesn't have to be a disciple of Freud to accept the reality of unconscious desire.
|
[
Reply to this ]
|
Re: Unconscious Computing: the Know-Thyself App
by Karl on Monday March 06, @05:36PM
|
|
No, that's certainly not what I mean. I mean that the unconscious is not primarily a kind of prison for banished and unacceptable thoughts--the seething cauldron of repressed desires that is the popular vision of the Victorian-Freudian model. It would be better to call it the 'pre-conscious' for most purposes.
Although, no, I'm not aware that having desires different from your conscious ones is a fundamental property of the unconscious. Most of your activities while driving a car are unconscious, if you're an experienced driver--do you consciously have to remind yourself to shoulder-check before you make a turn? (I hope not!)
|
[
Reply to this ]
|
Re: Unconscious Computing: the Know-Thyself App
by Ted Chiang on Monday March 06, @06:20PM
|
|
Here's an example of what I mean by unconscious desires. You might, if asked about your co-worker John, answer honestly, "I like him." Yet you might demonstrate behavior that suggests hostility toward John. Pointing out these behavior patterns to you might cause you to say, "Wow! I didn't realize I was doing that. I guess I sort of resent John for getting that promotion instead of me." Or, you might deny it, and say, "You're seeing things that aren't there." But the behavior pattern can be real even if you deny it.
So, a computer that "determines your interests and values by watching your reactions to real-life situations" might identify this resentment towards John. If it reports this to you, great; you can either accept or reject its conclusions, but you'd better not have the computer act on this without your explicit approval. Otherwise, your computer might try to act in your interest by, say, ordering a gift basket of cookies for John, knowing that they contain trace amounts of peanuts and John has a peanut allergy. That would not be good.
|
[
Reply to this ]
|
Re: Unconscious Computing: the Know-Thyself App
by Karl on Thursday March 09, @08:39AM
|
|
Hmm... sounds like a story.
I was interested in this exact tension when I wrote Lady of Mazes. In that book the 'anima' are essentially simulations of people that use exactly the sort of technology we've been discussing. They understand your conscious intent, and they also understand your unconscious drives. But the thing is, they attempt to act like you. In your example above, the computer is not actually acting like the person it is modeling. An unconscious dislike does not equate to murderous intent any more often than a conscious dislike will--probably less.
You could in fact argue that the computer is equally likely to misinterpret your conscious intents and accidentally kill the other guy. To the computer, there's no natural distinction between your conscious and unconscious desires. So the problem you identify above is actually a problem with having computers act as our proxies in general, not specifically with regard to our unconscious desires.
|
[
Reply to this ]
|
Re: Unconscious Computing: the Know-Thyself App
by Ted Chiang on Thursday March 09, @12:56PM
|
|
My example was exaggerated for effect. The question is, would you want your computer to mimic your subtly hostile acts toward John without consulting you first?
More broadly, what does it mean to "act like you"? People have competing, conflicting desires. It can be difficult for other people to identify what we want; in fact, it can be difficult for us to identify that ourselves. It's not at all clear to me that one's unconscious mind can be described as "a small army of fanatically loyal well-wishers, who genuinely care about you and tirelessly act for your benefit." Sometimes your unconscious mind may act as the "better angels" of your nature, but other times it definitely doesn't. And this is, I think, a different issue than a computer misinterpreting your explicitly articulated goals.
|
[
Reply to this ]
|
Re: Unconscious Computing: the Know-Thyself App
by Karl on Friday March 10, @11:18AM
|
|
Yeah, that makes sense. In Lady of Mazes, I had each person surrounded by a sort of virtual cloud of versions of themselves; one vital aspect of the culture of Westerhaven is the presence of simulations, in which people simulate themselves, each other, and in which simulations use simulations in a dizzying recursion. In that case there's a constant cloud of potentia surrounding people and their virtual agents, very much like a probability cloud in quantum mechanics. At times, the wave function collapses into a particular action or interaction, but declaring 'who' has acted--the physical person or one of their virtual proxies--is often difficult. In Westerhaven people don't limit their self-image to their physical selves at all, so they take responsibility for things their proxy selves do--within limits, because the 'real' person is the aggregate total of proxies and physical body. Behaviour that doesn't agree with that aggregate total is not seen as something you'd be responsible for, and is in any case weeded out by the natural selection of the sim system. In Westerhaven, 'you' are a participant in your own life, not the sole agent living it.
To expand on that just a little, you are in constant negotiation with yourselves on how to represent yourself; put another way, your biological 'self' as we currently have it is the most powerful selection mechanism for the ongoing natural selection of versions of yourself. Whether you dominate or are dominated by your externalized selves is really a matter of personal style.
My Westerhaven people have annihilated the current model of individual identity in favour of something like an extended feedback loop that penetrates far into external reality. One reason they do this is because it allows them to reconcile issues such as, eg. life and death. In SF people in the future often use extentions to their senses and cognitive abilities in an attempt to transcend their mortal condition; in LOM people like Livia use extensions and an overall amplification of identity to do the same, with arguably greater effect.
This is a bit of an extended answer because as you can probably guess, I have definite ideas about what it means to "act like you" but they cannot be summed up in a glib definition.
|
[
Reply to this ]
|
|
The Fine Print: The following
comments are owned by whoever posted them.
( Reply )
|
|
