(and why can’t I claim ownership of items I “created” with it??)
Today’s AI is not what you – or the rest of the technology world – actually thinks it is. While it is very much artificial, it is by no means intelligent, conscious, self-aware, or even self-thinking. Yes. There are some very life-like robots out there, but nothing advanced enough to be considered AI. This is just very advanced machine learning (ML) and large language model (LLM) computing.
That’s not to say that any of this isn’t sophisticated. Just look at what those life-like robots can do, including Hanson Robotics’ Sophia. Even that robot isn’t true AI, though it can process visual, emotional, and conversational data and has spoken on the Tonight Show with Jimmy Fallon, The United Nations General Assembly, and (very controversially), a commencement speech at New York College’s 2024 graduation ceremonies.
So, is this artificial intelligence?
Very simply put – no.
It’s not, despite how freaky some of the interactions Sophia or other life-like robots may have with users, it’s still just an interactive response to human input. It may seem like the robots are alive, but they don’t yet new, independently, cannot outgrow the parameters of their programming, and can only respond to queries (or human interaction) with the sum of their programming, be it the full sum of human knowledge or just part of it. These robots cannot create or ideate new, independent thoughts. They can only mimic or predict the odds of something being chosen, used or selected based on the data on their “hard drive” and a very powerful probability engine.
Is it stupefying? Is it freaky that they can “see,” measure the tone, timbre, and tremolo of your voice to detect stress, agitation, etc.? That they can read or “sense” your facial expressions and predict what you might be feeling based on vocal recognition, and all the other inputs you are subconsciously providing it with??
Oh, HELL yeah! Its freaky, like it was the very first-time human beings were able to start a fire or turn on an electric light or take a still or moving picture. Epiphonal advances in technology are always seen as incredible, or shocking, or frightening to some people, even those of us who are technology experts. Regardless of our level of experience or expertise, we’re still only human…
Ok, so if this is not REALLY “artificial intelligence”, just very sophisticated behavioral mimicking, and analysis, and what not, how does it do what it does? As I mentioned, this is really machine learning or a type of computing that allows computers to learn and improve from data it has access to without explicit programming. It uses algorithms to analyze immense amounts of data, identify patterns, to create a model that can make predictions. The more data a machine learning system is exposed to, the better it becomes at making predictions.
How much data is required?
LOTS.
No, more than that. I mean, LOTS!!
We’re talking petabytes or exabytes of data (a Terabyte is a trillion bytes. Peta- means a thousand trillion. So a Petabyte is 1,000 Terabytes. An exabyte is 1,000 petabytes); and the more data, the better. When I say that ML and AI systems need the “sum of human knowledge” (estimated to be 463 Exabytes by 2025) to work properly, I really do mean it. The more examples it can work into its predictive model – and again, that’s all ML is really doing – predicting with a margin of error (no matter how large – less data – or more accurate – more data) – what a user is looking for, wanting, or desiring – the better chance it has of providing the “answer” you’re “looking” for.
For example, “AI-based” image creators need examples upon examples, upon examples of previously created images – be they man-made or ML generated – in order to be able to generate an image that satisfies what you want to “create.” The amount of processing power needed to predict where each pixel is placed and with what color and hue that satisfies the needs of the engineered prompts used to start the ML action is HUGE. (Remember – machine learning cannot ideate on its own. It can only predict what you might want from it, based on all the examples it can access, and then again, with some margin of error.)
This is why as of this writing, new “AI-computers” require a Neural Processing Unit (NPU) dedicated to processing the request and the amount of data needed to satisfy it. (Its also why ARM-based – vs. Intel-based – computers are better suited to AI. The amount of heat generated, and electricity/battery power required is better suited to the ARM platform, which runs cooler and manages power consumption much, much better. But… I digress…)
Claiming ML/AI Generated Content as your Own
Right now, the US Copyright Office will deny any copyright requests of AI generated, or even AI-assisted material by persons submitting it for registered ownership. You can’t claim the “raw” generated ML/AI results as your own because in their opinion, an engineered prompt is not long enough, not specific enough, and/or doesn’t represent enough human ideation and original work to warrant registered ownership. In fact, they have ML/AI applications of their own to help them identify works submitted to them that contain what they consider “an inappropriate amount” of ML/AI generated results.
Here’s a couple of great examples:
AI Artist Loosing Millions from Stolen Work
A well-known AI artist who won a painting contest in Colorado using a Midjourny-generated creation claims he is losing millions from people stealing his work. The artist claims the US Copyright Office is reluctant to register his AI-generated art collection with a copyright. Their refusal has allowed repeated instances of his work being stolen without attribution or compensation. His lawyers are arguing that his prompt engineering and scene-selections/ selection skills should be considered creative input and human authorship, which would allow him to claim ownership and register his copyright.
AI Generated Music
I’ve been writing and playing original music since the age of 14. I’m nearing 60 now, and that my friends, is a lot of experience in recording studios, band rehearsals and live performances. As a serious amateur, I can tell you that I’ve played in live performances in Nashville, TN – Music City, USA – as well, and I feel that anyone who can hold their own there, knows what they’re doing. Nashville isn’t (necessarily) kind to those wannabes who can’t keep up.
So when I say I’ve been playing with “AI” music generation apps and that I’ve had them produce at least 3-5 hit songs from the 70 or so I’ve generated, I mean that these are HIT songs. Yes, the work may be derivative – meaning that it might sound like other music on the radio right now – but when you wake up in the morning and you start humming one of them out of nowhere, you just might have something.
Important Links!
- Subscribe to our Monthly Newsletter!
- Subscribe to our Weekly Podcast on
The post How Does Today’s AI Actually Work? appeared first on iTechGear.org.