This is not written by an AI, yet

Large language models (LLM) can remix existing words, sounds and images, but they don’t think, they blend. Of course once my words get uploaded to the internet, an LLM crawler can find it. If it decides my words are worthy it will add them to its model or database. Then when it is asked to opine on this topic it can regurgitate it, in part or in whole, changing the wording and structure enough to ensure no copyright laws are violated. If it’s nice it can quote me, credit me or link to me, but there’s no law to enforce this.?

Of course Google and others already crawl the internet. If you ask Google, using the right words, they will provide you with a snippet of this article and links to this and other similar articles, ordered by their proprietary (quality) ranking systems. On their results page they will insert valuable advertising, but not pay the web page authors a cent, even though they are profiting from our copyrighted materials. Under copyright law, everything written, drawn, recorded or photographed is owned by the creator, unless they have explicitly assigned those rights away, say by signing up for a social media account and then uploading to it. Google is a $280B advertising machine legally profiting from other folk’s copyrights, without licensing those rights, because search is considered fair use, not ‘copying’, even though they make copies (internally) to do search. Or I could mark my web page ‘noindex’ but then you won’t find it, so why write it?

But currently no one can mark their page ‘noAI’, and there are no clear AI laws. Even when there is, there will be ways around them. Maybe Malta or the Cayman Islands will allow AI’s to free roam. Intelligence services and their subcontractors will claim ‘national security’ to override ethics and laws. Startups, humanitarians and researchers, like Google’s founders pre Google, will ‘learn’ from whatever they can grab. If there is a value to scooping up everything, including this web page to feed a LLM, it’s going to happen.?

So shortly after I write this, thanks to Google, someone else will be able to find it and reproduce it, or something like it, simply by entering the right prompts and waiting a few seconds. That’s great when you need to summarize what’s already been written. But even the latest AI systems don’t think, they blend. From millions of examples they can learn the rules of language, spelling, grammar, wording, summarize (reword) or suggest something similar in their database, but they don’t understand. They work by pattern matching. They can either repeat or randomize. They learn and can predict the repetition of letters and words, but not by modeling how people think.

Today you can copy and paste my article into OpenAI and change my style into the style of your favorite author, but if you don’t agree with me, you’ll have to figure out why and start typing yourself. AI’s of 2023 are early, messy, experimental prototypes of what will come in ten or a hundred years. There is no clear dividing line between writing good sentences and paragraphs, and writing good ‘original’ thoughts.?

Since the invention of the electronic computer eighty years ago, AI’s have learned behaviors that are, in some ways, superhuman. But, in terms of human-like ‘thinking’, current AI’s are still children, perhaps comparable to super-literate eight year olds. That’s 8 years of development in the past 80. Our electronic brains are growing up faster now, perhaps one human development year every five calendar years, as the world’s smartest companies invest billions and their brightest minds, questing for workers who learn quickly, aren’t paid, never tire or complain. In the lifetime of someone born today, AI's could grow smarter than adult humans, or ascend beyond our understanding; angels or demons of our own making.?