AI — The Good, the Bad, and the Ugly
The era of AI is taking off in earnest. Are we going to just “take it as it comes?” OR, “work to understand what Machiavellian Plans the Hardware/Software world is planning for us?
Like the Protostar Within (L1527 only about 100,000 years old), still embedded in the cloud of gas and dust that feeds its growth, AIs growth is set similarly to become a Star of its own. What the new AI star, in the universe of “WE” the people will be, has NOT BEEN DETERMINED YET — Copyright definition for AI is still being manufactured from the cloud of confusion dust today. Some (scientists) hope AI will give us indications of the future course of humanity, others are looking to determine the possibilities of where “life” came from. What it will really do and how WE will use it (or will “it” use us?) is TBD.
In the last six months of 2022 we, here in Medium, started seeing people talking about programs we had never heard about. The boom of AI models started. Today, we have a deluge of posts discussing AI in all aspects, it is the prominent subject of either discussion and use given the proliferation of AI images in our midst.
Several detail though and like the heading of a November 15, 2022 article in “The Verge” spouted: “The scary truth about AI Copyright is, nobody knows what will happen next.”
Money is pouring into AI and its growth will be exponential and only limited by the speed and quantity of fast GPUs we can muster. After studying conference discussions and reading articles on the subject the most salient question is — “Who represents the Users in all this?” Not once anyone has mention this.
The other question that comes to the surface is if billions are being spent on this Endeavour, somebody must think there is money to be made at the other end. I can see end users like:
- Corporations — products that use computer programs that require decision making like, flying an airplane (with pilots still), drive a truck (ditto).
- The Medical field for diagnostics, Ai assisted surgery (with expert doctors still)
- Medicine Research, to hone in on virus vaccines and other difficult not yet solved problems like Alzheimer and others that impair people’s quality of life.
The problem is “how do we get there when it comes to copyright?”
Given everything I have read and heard, I see two areas of interest here.
1. Large entities requiring AI can afford to pay their way there and have all the lawyers they need to lock-it-up
2. People whose livelihood depends on the human brain for authorship creativity, art spark, New music inspiration, or the next billion dollar idea, the niche that drives progress through ingenuity, and makes a living through hard work.
I would suggest all entities that can afford AI can afford to copyright their software. The rest of us little people (item 2) cannot afford to fight Google for writing books based on my inspiration. That is an unfair fight therefore Google has to be prohibited from entering my market. Since one of the functions of the government is to make the playing field flat and the Copyright office is there to serve the people, then the copyright office needs to serve me a lawyer that defends my right to a livelihood without unfair competition. “fair use” is no longer a good defense for those with an unfair advantage. it is there to serve WE, the little people.
The answers that I see are binary (a one or a zero). Millionaire entities cannot be allowed to abuse the little people.
Ditto for those entities that allow me to create art with their AI but write horrific Term and Conditions I have to agree to on the fly. I am not a lawyer and cannot afford one for every “app” that wants me to “click here to agree.”
The era for the Giant BOTs has begun and we are still, “the little people.”
The rest is not all of it But more detail onto the main thoughts rambling out there, in the “Professional” arena (some are below). Except for my final thoughts at the end.
What is the status of copyright in 2023 and how do we edit Medium stories that may contain AI redacted paragraphs and images?
As reported by “IPWatchdog,” whose focus is “on the business, policy and substance of patents and other forms of intellectual property,” in their article “Thaler Loses AI-Authorship Fight at U.S. Copyright Office,”
“In an opinion letter dated February 14, 2022, the Review Board of the United States Copyright Office (Review Board) affirmed a decision of the U.S. Copyright Office (USCO) denying registration of a two-dimensional artwork generated by Creativity Machine, an artificial intelligence (AI) algorithm created by Dr. Stephen Thaler.”
How About the User?
Previous to Thaler, in a real life example for user authority, in 2020 James Grimmelmann, Professor of Law, Cornell Law School, Wrote an essay for the Ohio State technology Law Journal 16, p. 25, “Spyware vs. Spyware: Software Conflicts and User Autonomy.” Where he informs how Apple:
“pushed a software update to macOS to delete the Zoom server and prevent it from being reinstalled…. Apple deliberately broke an application feature on millions of users’ computers without notice or specific consent. And then, six days later, Apple did it again.”
Which can be taken a warfare being waged by the “owners of software” on the public right to choose how they want to run their machines.
However, the problem is really bigger than that, as he explains, “this is not an isolated and unrepresentative incident,” there are further examples of not only invasion of software to unsuspecting users, but warfare happening beyond the screen the user sees, inside every machine. He calls it “further examples of programs doing drive-by’s on each other like warring street gangs:”
· “Malware: Antivirus software attempts to prevent malware from being installed on users’ computers, and to remove that software if found. Malware tries to install itself and evade detection and removal, so of course its first order of business is often to turn off any antivirus protection.
· video game bots: Some online game players use bots to play the game for them, leveling up their characters and obtaining resources. Blizzard, which operates the popular game World of Warcraft (WoW), added a program called Warden to WoW, which detects bots and reports them to Blizzard so it can ban their users from connecting to Blizzard’s servers. One bot maker, MDY, modified its code to evade detection by Warden. Others developed techniques to modify Warden itself and disable its surveillance without alerting Blizzard.”
The list goes on with examples of:
· “Ad Blocking
· Ad Injection
· Browser Tracking
· Email Tracking
· Jailbreaking” — Operating system that make it difficult/impossible to install software not approved by them. And the market of programs that will allow such installations.
The bottom line being that “WE” don’t know what is running in our machines anymore. To where he says:
“One way to make sense of these program-versus-program conflicts would be to proceed methodically through the bodies of law that could be (and have been) brought to bear on them. But their sheer number is stunning.”
Regarding copyright there are plenty, where “the length of this should give pause.”
Three Theory on the view of User Concepts that are typically assumed because they are so straightforward and widespread:
1. “Bad Software is Bad” — certain program behavior is intrinsically harmful and should be prohibited. Programs should not spy on users and delete their files.
2. “Software Freedom” — users should be allowed to run whatever software they want.
3. “Click to Agree” — users and software vendors should be held to the terms of whatever contracts they enter into.”
For purposes of AI we are interested in “Click to Agree” since their terms may/may not, be harmful to users.
“The standard approach to ascertaining user consent in the United States circa 2020 is contractual. Users agree to the terms of a contract, privacy policy, or other instrument when they provide a specific manifestation of assent to it after having been given notice and a reasonable opportunity to review its terms.”
So, the question is: “Are users lawyers?” does it make sense for me or you to “play lawyer?” Does this hold up in court?
In the October 26, 2021 conference “Machine Learning and AI,” Dr. Aude Oliva, Director of the MIT-IBM Watson AI Lab, gave a brief introduction to Machine Learning — How the software is configured such that, through given rules the machine, can learn by association, digestion, and repetition of millions of “input data.” Back in the early days (software 1.0 she called it), “The classical code stack were explicit instructions to the computer written by a programmer.”
“Today, call it “Software 2.0,” in “Machine Learning, the code exists as the “weights” of a (like a neural) network. No human is involved in writing that particular code because there are a lot of weights (millions to billions of parameters).”
Human programmers pick up the “data” (real information) -> use computer resources (Parallel computing, GPUs, Cloud, Hardware) -> write algorithms that form a neural network that allows learning -> it forms a “Deep Neural Network.”
Deep Neural Networks can create very powerful models.
There are two types of data (the gold mine) that “fills” the model, “Real world data and Synthetic data.”
“Real world data” there the collection of billions of messages sent to apps, billions of prompts on search engines, data that is captured from the real world. Real data is captured by companies that have the most customers (the gold mine). These can build the best model.
That is the crux of the problem, IF on the input side the program was fed millions of pieces of copyrighted data created, authored, painted, by people who made their leaving from the fruit of their labor and the output now competes with those human authors is that fair?
Cullen O’Keefe, Council for Policy and Governance at OpenAI (ChatGPT), has created a “General Purpose Language Model with Transformers.” Their purpose is to:
“create language models that learn the general functioning of the English language which can then be applied to a large number of natural language tasks.
For example, produce a summary of a long piece of text, translate from French to English, answer a question.”
The “Transformer Model,” or General Purpose Transformer (GPT), as it is called, uses function specific models that yields a General Purpose Model.
A question of copyright was asked of Dr. Oliva, “Are the AI created videos or photos you showed eligible for copyright?
“First, I am a machine learning person, I am not a lawyer. I can tell you what is going on into those neural networks. If you use a large dataset of images and you train a GAN to produce a new image, some of those images or videos, will look like some of the samples you used. Some others, will be a transformation between two or more images. The criteria then is subjective because the question arises, are these images different enough, to merit being “original?” Are they different enough and creative enough to merit originality? The information is hidden in those weights of the neural network. There is research going on about this.”
Final Thoughts: The entire Conference, “Copyright Law and Machine Learning for AI,” co-Sponsored by the United States Copyright Office and the United States Patent and Trademark Office, lasted about half a day and can be found here, for those interested. It took 24 pages of notes and about four days of patience to collect. I watched it twice and at the end, I felt it was one sided because I was not represented by anyone there. The closest was Mary Rasenberger, CEO of the Author’s Guild. Each participant had an axe to grind and they did that in a very civilized manner. Those giants will eventually have their day in COURT. However, AI is a whole new Game. While the Copyright as an entity represents the Constitution, therefore “The People,” this is not true any longer, they are just playing the referee between giants. When that does not work, the stop the game and go for what may be equivalent to an instant replay, the Courts to settle the score. The royal “WE’ do not figure into the equation.
