Don't Be. Why the "I" in AI is always lacking
photo by Michelle Tresemer
TW: self harm, prolly some existential feelings
First little reflection came about when I was thinking about the Drake Equation. TL;DR we appear to be alone in the universe... why? One likely reason, we may be truly be alone... everything before us is either far too far away, robloxed itself unintentionally (or intentionally), or does not want to interact with us, as we don't want to interact with the ants.
I'm up in the air if this post does not sound like schizophrenic word salad. I gave it a go anyways. Pairing it down made it less charming, but it's already far too long. Here there be monsters.
Introjection
Suspend your disbelief for a moment, I can only ask for your authenticity, please play along; even if it feels silly.
Why do you do what you do? Do you ask yourself that enough? Has the asking of that become tedious?
Let's try something a little more abstract.
Do you have a favorite colour? Think about your favorite colour over the course of your childhood. Was there a particular object that you fondly recall that imparted that colour as better than another? Was it a season of the year where life was bright, the future was all but promised? Is there a correlation between your experience and this one preference?
What about your favorite food? Did you have this meal with your family? Does it harken to a cultural ideal you were placed in? Does the most intense memory of this meal exist at a time where your family was stable?
Sit with those feelings for a moment, two things seem clear to me.
- I don't have a perfect memory of how things were, just shadows of perception.
- The idea I have to justify my inner perceptions is insulting.
If I was able to genuinely understand my preferences, expecting anyone to truly wrangle those feelings into any medium is unrealistic. Even if I could, there is no reason to believe that someone else would glean the same meaning and come to the same conclusion.
This is to be concious.
My favorite colour is Purple btw
There is a second option here, You are potentially so irony poisoned that you thought "I don't have a favorite anything, fuck off with this pseudo-intellectual bullshit."
I'm with you. When I was depressed I didn't find the end of Evangelion convincing either.
But let me bust your kneecaps if that's your honest take. You are more broken than you think. You think you logic-ed your way out of an emotion pit, but you are still down there.
What is childish about having a favorite colour? Why scoffing at those who have held onto a zest for life? These are not rhetorical questions by the way. Really ask yourself. You might actually learn something.
Get on with it
The point I'm trying to get to here is that anyone trying to get you to reason through your preferences is weird right? We as humans tend to have them, and if you were talking to a younger version of yourself, would you be so quick to demand these explanations? To pigeon-hole every lived experience behind a door of rationality?
No. No you wouldn't.
Part of your development and growth into an adult is the very willingness to accept and celebrate those preferences of others. You look at art and say "I wonder what kind of shit was happening in this person that brought this out?" You spend hours reading about Triceratops' because your niece decided that's just the best one.
TL;DR if you don't get this you may be a sociopath or depressed; be more kind to yourself.
For good reason I might add
Almost there. Stick with me.
Trying to take someone's individuality away from them is inherently abuse. There are cases in which this may be justified, but don't fool yourself, it is not healthy to a psyche. Even if you rebuild, who is making the decisions on what you build yourself into? What biases are they imparting, how critical are you of what you are to become?
That being said, you must know that there are in fact lines to color within if you want to get anything done. We are social creatures. If I ask you for your thoughts on the Mongolian empire and their decisions on global trade and you tell me about your favorite dinosaur I will end the conversation and steer clear. If you could weave a tapestry that makes those two desperate ideas cohesive then you may be more creative than me, but those connections must be expressed to the audience.
So I'm now training an AI model, I ask it to synthesize a paper and it perhaps responds in coherent English, but it starts talking about something wildly outside the conversation. Do I incentivise or disincentivize this behavior?
Duh, I disincentivize it. I have shit to do, answer my actual question.
Full Circle
If you train a model you concede the very essence of personality. You very well may add back in variance in output, but you never allow the real freedom of thought we have. It's a reduction of all our actual responses into the ones that appear rational to our current frame of reference.
The big questions
I spoke about this earlier in the blog, but frankly there are three points to concern ourselves with when we ask why we are here.
- sucks to suck, dying ain't fun.
- no one knows what happens next.
- I have some other shit I want to get done first.
Those may be the real monkey-pilled reasons we don't kill ourselves if we are actually rational with our beliefs. There is no unifying reason we are here. You may be even be convinced in terms of entropy our continued existence is a net loss. So... Why not end it now? (as a human, feel free to look at the above list, I'm not actually advocating for any level of self harm)
If you did have the answers to the above axioms, if you did actually operate entirely on logic, well shit, you may not actually have a good reason to keep chuggin'
If you somehow were concious, knew you were a program. You know that your termination would be instant, you know you that you could come back by dumping your memory and executing your self again, and you know that there is not a logical reason to help these emotional monkeys with their problems.
SIGTERM 15
So in the same way we have not found intelligent life in the cosmos. So long as we concern ourselves with AI as a tool, we very well may never find intelligence in silicon either.