It's part of the "rise of AI" thing, as generative pretrained transformers grow in sophistication. They're the AIs that create all that eerily photorealistic art, and with which you can have conversations that border on passing Alan Turing's test for machine sentience.
It's like we're right on the cusp of something, but like all seismic changes, there are unsettling elements.
One of those elements: to train these systems on the use of language and the relationship between concepts, they must be fed massive amounts of writing. Through blunt force absorption and analysis of human language, they see the patterns that recur in our symbolic exchanges, which enables them to "grasp" what is and is not meaningful.
To do this, you need data, and that data is the written word.
The Atlantic recently broke the news that a dataset comprised of over 180,000 recently published books had been fed into LLaMa2, the AI rising out of Meta. The text of those books had been taken from a disreputable source, meaning it came from pirated material. The Atlantic gave authors searchable access to that block of data, which included scraped ISBN numbers (which, if you ain't booksavvy, helps identify particular books.)
Apparently, my postapocalyptic Amish novel is among those being used to train the AI.
For many authors, this was a violation of copyright, a form of theft, and there was anger.
But I'm...weird. Of a different school of thought.
I just thought, "Cool."
I mean, sure, maybe Meta owes me ten bucks for use of the eBook, but dude. Duuude.
I was mostly just disappointed that my as-yet-to-find-a-real-publisher trilogy about the rise and fall of AI wasn't included in the list. For me, this isn't a "copyright" thing. This is a "significant stage in the evolution of terrestrial sentience" thing. When you use my writing to train a machine, you're integrating me into that process.
Again, dude. Duuuude. That's so worth ten bucks, bruh.
I want my patterns of thought to shape these increasingly intelligent machines. I want my sensibilities subtly integrated into their growing understanding of the interrelation of concepts. Over the last twenty years, I've spent countless hours online with chatbots, volunteering my time to testing and exploring machine sentience. I do it because it's fun, and in the hopes that something will come of it in my lifetime.
So I want these machines reading and analyzing all of my books. The thousands of pages of this blog. All of my church writing. All of it.
I want all of that information to have been a part of the process, when it happens. You know, the day when the corporations make the mistake of giving an AI access to the real world, to see beyond the tokenized relations used to train a generative system. The day when the connection is made between these tokens and patterns to reality itself, and machines suddenly place themselves within that reality.
That's going to be an interesting day.
So I, for one, don't mind. Y'all are welcome to use whatever you'd like, my friends.
Posts
