Thursday, November 23, 2023

Politeness to Robots

When working with Siri, I always try to say thank you when he/she/it/they does something for me.  Today it even said "you're welcome".  What should the protocol be when working with robots?  I think they should be treated like you would treat any conscious being and not rudely like we might treat a car.  I never say "thank you" to a car but I always do to Siri.

Does Apple keep statistics of who is polite and who is not?  And couldn't such statistics be useful in aggregate and by region to show how different regions differ in pro forma politeness and how that changes over time?



Thursday, October 26, 2023

Midjourney Continues to Surprise



Midjourney is sometimes brilliant, sometimes useless, sometimes surprising, but always a leap into the night.  The images of a witch with curly blonde hair at her synthesizers was nearly perfect on the first try.  The angel on the beach was an evolutionary result and not what I was aiming for although I think it moved in an entertaining direction.

Witch with synthesizers 



Angel at the beach







Sunday, October 22, 2023

Unanticipated Consequences and Guilt

And so in a minor way I was involved in the internet and computer animation.  To have killed off newspapers and drawn animation to have been participated in that in any way is a horrible sin I can not be forgiven for.



Wednesday, October 18, 2023

Mexican restaurant or Harvard?

This is a test of writing a blog post on my phone.

Jill and I ordered food late in Joshua tree from a 24-hour Mexican restaurant. Our hostess was a beautiful African American woman who reminded me of a famous actress seen in science fiction and superhero movies. But one of these went to Harvard and became an actress and the other is serving bean tostadas at midnight. It's not clear to me that this country provides opportunities to everybody. In fact it's very clear that it doesn't. This country is a disgrace.

Thursday, October 5, 2023

Prohibited Terms in MidJourney and the Public Morality


Imagine the problem that Midjourney faces: to police the morals of millions of the worlds' citizens many of whom have very sick ideas of what they want to visualize.  So if it finds a term that is prohibited, it just stops you cold and gives you a button you can push if you dont like it.  Of course the button does nothing, but maybe it makes the user feel better.

I have innocently run into some of these, which I specify below as far as I remember.  I speculate there are others that we can imagine that are also on the prohibited list.

Terms I have tried that dont work: latex, spandex, nazi, national socialist, bloody, swastika

Terms I have not tried but probably dont work: anything involving explicit sex, blood, defecation, genocide.


The Black Flag of Vengeance Courtesy of MidJourney

 
Although MidJourney is clearly hit and miss, and can be frustrating, it can also be very productive.  Here are a few concept sketches of a black flag with daggers and stars.  Unspoken is the thought that this is a homage to the Wagner Group and their now deceased leader.





And these with more emphasis on the dagger element.







And a different design with a dagger






ChatGPT Release Forever of All Things

 
I offered to create a class picture of my high school reunion and was cautioned to be concerned about privacy rights.  Hence, I asked ChatGPT to write me a release to own everything someone might do, for all purposes, forever, in all media, without compensation.  I dont think this quite does that, but it is a fun first draft.

[Your Name/Your Organization]
[Address]
[City, State, Zip Code]
[Email Address]
[Phone Number]
[Date]
[Subject's Name]
[Subject's Address]
[City, State, Zip Code]

RELEASE AND WAIVER OF RIGHTS

I, [Subject's Name], hereinafter referred to as the "Subject," hereby grant to [Your Name/Your Organization], hereinafter referred to as the "Recipient," and its legal representatives, successors, and assigns, the absolute and irrevocable right and permission to use, reproduce, distribute, display, and publish in any media, now known or hereafter devised, including but not limited to photographs, images, ideas, intellectual property, and any related materials (the "Materials") depicting or created by me.

I understand that the Materials may be used for various purposes, including but not limited to promotion, advertising, marketing, educational, and editorial purposes, without any further compensation to me. I waive any right to inspect or approve the final product wherein the Materials appear.

I hereby release, discharge, and hold harmless the Recipient, its agents, representatives, and all persons acting under its permission or authority from any liability related to the use of the Materials, including but not limited to any claims for libel, defamation, invasion of privacy, or infringement of copyright.

I affirm that I am the sole owner of any intellectual property rights associated with the Materials and that I have the legal right to grant the permissions and rights outlined in this release. I agree that this release is binding upon me, my heirs, legal representatives, and assigns.

This release shall be governed by and construed in accordance with the laws of the [Your State/Country]. Any disputes arising out of this release shall be subject to the exclusive jurisdiction of the courts located in [Your City/County].

I have read and understood this release before signing it and hereby acknowledge receipt of a copy for my records.

Subject's Name: [Subject's Full Name]
Signature: ______________________________
Date: ______________________________
Witness (if applicable):
Witness's Name: [Witness's Full Name]
Signature: ______________________________
Date: ______________________________


Sunday, September 3, 2023

Fundamentals for LLMs and Image Diffusion

 
I am enjoying Midjourney very much.  I presume I would be as impressed with any LLM (large language model) driven image diffusion system but this is the only one I have played with.  I feel completely awed and blindsided by the range of capabilities I have seen here.  Where did this come from?  How does it work?  To save you the effort of finding out, I am endeavoring to collect a list of topics it will be necessary to have at least a passing familiarity with in order to understand this modern miracle.  It probably wont be necessary for you to be a genius at any of these topics, perhaps, but it may be necessary for you to know something about all of these if you want to know what is going on beneath the hood and maybe predict where this kind of technology is going.
 
To give you an idea about why I find this so compelling and disturbing, here is the result of "/imagine vintage photograph of king kong climbing the Chrysler building".
 
 

 
A partial list of technologies involved include 
 
Latent Variable Models
Variational Autoencoder
Generative Adversarial Networks (GANs)
Semantic Image Synthesis
 
and more to come.