The March Philosophy Book of the Month is Final Notice by Van Fleisher. Discuss Final Notice now.

The April Philosophy Book of the Month is The Unbound Soul by Richard L. Haight. Discuss The Unbound Soul Now

The May Philosophy Book of the Month is Misreading Judas by Robert Wahler.

Relationship to error as distinguishing trait between automated process and creative minds?

Discuss the January 2019 Philosophy Book of the Month, The Runaway Species: How Human Creativity Remakes the World by David Eagleman and Anthony Brandt.
Post Reply
User avatar
Scott
Site Admin
Posts: 4320
Joined: January 20th, 2007, 6:24 pm
Favorite Philosopher: Diogenes the Cynic
Contact:

Relationship to error as distinguishing trait between automated process and creative minds?

Post by Scott » January 13th, 2019, 10:01 am

When comparing automated processes to the creative aspects of the human mind, the authors of The Runaway Species write the following (page 160):
We count on automated behavior to be free from mistakes. In situations where outcomes need to be reliable, such as getting the fork to our moth, neural pruning removes superfluous options. We want to type correctly, run without falling, play a perfect scale on the violin. But proliferating options requires a different attitude towards error. Error is to be embraced, not avoided. In automated behavior, error is failure; in creative thinking, it is a necessity.
That is an interesting way of distinguishing the automated and/or subconscious aspects of the human brain with the more creative and conscious aspects. Even more, I think it starts to speak to one of the differences between current machines and human minds, perhaps illustrating one of the obstacles in creating strong general AI.

What do you think?
Online Philosophy Club - Please tell me how to improve this website!

Check it out: Abortion - Not as diametrically divisive as often thought?

User avatar
Dorian
New Trial Member
Posts: 3
Joined: January 13th, 2019, 12:51 pm

Re: Relationship to error as distinguishing trait between automated process and creative minds?

Post by Dorian » January 13th, 2019, 1:26 pm

I have been a reader of this forum for some time, but the serendipity of this post inspired me to create an account. Just yesterday I read an paper titled "Erro, Ergo Sum" in the Philosophy Pathways journal which hit on this exact subject. The author takes a more extreme stance than The Runaway Species to hypothesize that error making isn't just a necessity for creative endeavors, but the defining characteristic of a conscious mind.
Although errors vary greatly in nature and degree, this discussion will focus on the
most fundamental manners in which we might err, which will be referred to as perceptual
errors. These occur when sensory information is either ambiguous or misinterpreted, and
suffers an encoding error from which we derive an inaccurate picture of reality. The
theory proposed is a framework in which consciousness, cognition and free will may
have emerged from a single evolutionary adaptation to safeguard against these perceptual
errors.
I have been thinking on it all weekend, and it is really weird when you hear a new idea and it seems to start popping up all over the place. I watched David Eagleman's PBS series on The Brain, and now it looks like this book will move to the front of the queue when it arrives.

To your point regarding error making as an obstacle for a strong AI, I agree it is an obstacle and one of the most profound differences between current machines and human minds. It seems to go part and parcel with the enormous undercurrent of hidden motivations (evolutionary pressures, social dynamics, biological drives, ect.) that there is no obvious way to program into a computer in order to replicate or approximate human thinking.

User avatar
cavacava
Posts: 55
Joined: October 12th, 2018, 11:10 am

Re: Relationship to error as distinguishing trait between automated process and creative minds?

Post by cavacava » January 13th, 2019, 5:47 pm

The process of creating automate behavior, involves error, and I imagine the engineer working on such a process does embrace error as something to be overcome, a bug. The machine process once complete, may be error free and as such it is available for use by others who may want to incorporate part or all of it for some other project. Common structures may arise depending on purpose and these may get incorporated in many schematics.

If all creativity is confined to syntactical changes then perhaps machines can be as creative or more creative than its creators (although it may be that 'the cause cannot exceed its effects').

Tehsorso
Posts: 10
Joined: December 23rd, 2014, 3:49 pm

Re: Relationship to error as distinguishing trait between automated process and creative minds?

Post by Tehsorso » January 17th, 2019, 2:52 pm

Dorian and Cavacava - very good observations. I strongly recommend The Fourth Age by Byron Reese, incidentally the February club book, for an interesting discussion on the error-related limitations of AI. The book is really philosophical in its approach, by the way, with a multitude of perspective being explored equally convincingly.

Cavacava - error may be a bug for the engineer if we think of error as a malfunctioning operation, but building an AI that would process error in a constructive way would be something of a holy grail. Wouldn't that be what you mean by "incorporated in many schematics"?

I agree, though - if AI were to use error the way human do - that is, for creative (generally speaking) purposes - they would probably be able to generate more creative output than humans. But, as Reese suggests, it's highly unlikely that they'll be able to replicate human error processing.

User avatar
LuckyR
Moderator
Posts: 3375
Joined: January 18th, 2015, 1:16 am

Re: Relationship to error as distinguishing trait between automated process and creative minds?

Post by LuckyR » January 19th, 2019, 3:16 am

This topic's controversy reminds me of the splash and then backlash for and against fuzzy logic awhile back.
"As usual... it depends."

Judaka
Posts: 251
Joined: May 2nd, 2017, 10:10 am

Re: Relationship to error as distinguishing trait between automated process and creative minds?

Post by Judaka » January 21st, 2019, 2:50 am

Creative thinking doesn't even require error and dealing with error is necessary even when you're not trying to be creative. If you want to learn tennis the standard way and just copy the understood methodology, techniques and strategies... you're still going to make errors all the time. To become good, you need to acknowledge the error and take it as an opportunity to improve rather than getting frustrated. That's without any creativity.

Creative thinking is more about not making assumptions which restrict the possibilities, you can be extremely creative and still be terribly inefficient and ineffective. Let's say you're cooking in the kitchen and you want to be "creative", you can achieve this without your food tasting any good. You don't need to make an error, you can be creative on the first try.

This philosophy forum is a great example of that, I've seen so many creative and wacky ideas, that they are based on incorrect premises, are of no use to anyone and aren't coherent has no relevance to their creativity.

Seahunt
Posts: 66
Joined: October 12th, 2010, 10:21 pm

Re: Relationship to error as distinguishing trait between automated process and creative minds?

Post by Seahunt » January 27th, 2019, 11:24 pm

I was just glancing at this and while I haven't read the book, my ears pricked up when I read the word "errors". I mean what could be more up my alley than errors! My middle name is "error prone". Looking at the comments though, I was baffled by a few things. My company was recently discussing automation and I pointed out that I didn't see where the design allowed for errors. Well, that seems silly. Also, there was no provision for human intervention. Well, that just seems like a weak design. It did curtail the conversation a bit, but they know about me and my software. Seriously, my system mediates between servers all over the country. If it can't handle errors, it's not going to work. It's that simple. Servers go offline all the time... especially those pesky Pheonix ones. It's almost impossible to break my system because it is meant to expect errors.
While on the topic of errors, I'm not sure if this is talking about AI, but one thing they learned was that you can typically write a system that can solve a problem maybe 65% of the time. What they found though was that if you teach it to use another method to solve the problem or force it to find another way, it will then be accurate with the two methods perhaps 75% of the time. Add another method, which the Ai folks now call "filters" and it can reach maybe 92% and so on... Well, this is true of humans as well. before I ever read that, I wrote in one of my books about how the best way to develop intellectual ability was to make mistakes. Errors are not necessarily something to be avoided. They are to be managed, sometimes to be taken advantage of. ... Just a thought.

Post Reply