AI and my Grandfather
Recently, while writing in Microsoft word, I have been frustrated with the AI-based interventions while I write. It is like a mosquito buzzing while I work. To make it fun, I did screen grabs of the errors that the grammar/spelling mistakes it wanted. No, not every sentence is past tense. And yes, AI you can refer to “stair” as a singular. I thought I captured building blocks for a funny series of block posts. It wasn’t funny. In the recent month, the AI-based interventions escalated.
In July 2024, I observed that Microsoft Word is getting slower and slower with its echo (time between keystroke and display). A few reboots, it is still bad. I whine to a colleague. He says go delete X,Y,Z from W folder.
C:\Program Files\Microsoft Office\root\vfs\ProgramFilesCommonX64\Microsoft Shared\OFFICE16\ai.exe
I did and it got better instantly. This was the evidence that what was forecasted to happen had already happened. The rumor was Microsoft would be capturing 100% of keystrokes in their applications and sending them to Microsoft as fodder for their AI engines. As an individual, my writing is protected by copyright. Now Microsoft is scraping all of my work as I type (a nice opportunity for rebellion, y’all). There is a second rumor that Microsoft is (or will be) taking screen captures and sending those back to their AI crap. Therefore, Microsoft is deliberately violating my privacy, your privacy, and violating our copyrights. While privacy is not protected by the United States Constitution, nor protected by many laws, copyright is.
Yesterday, while on the phone with my publisher, we discussed AI. She indicated that many publishers looked to embrace the technology to enhance their daily tasks such as: writing email, writing comments, and the like. Publisher are the artists advocate and defender of our copyrights. In fact, their job is to sell these rights for foreign sales, movie and TV deals, or what ever. It is capitalism and our intellectual property is our product. Their job: sell it and make us all money.
Hey, publishers if you embrace AI to write your documents and email, then you are paying for the destruction of the rights you are contractually obliged to protect. Wait… How’s that?
Let’s step back. How does AI write an email? The first body of email used as fodder in the large language model came from Enron email that landed in the public domain. Free to us. And tech companies did. It wasn’t enough. Limited scope of topics. Limited scope of writers/readers unified by a single corporate culture. Pretty sure that most people would not want their AI-generated email to sound like it came from Enron in the 1990s (Yes, Virginia, there was internet in the 1990s). Therefore, more data was needed for the development of large language models (LLM). LLM are simply repositories of words written by humans (mostly). The next best source for large volumes of data are published works that are out-of-copyright. Once you gobble that up, you recognize that language and style has changed since the 1930s. We need more writing samples. Since we (the tech industry) sits in the middle of the internet every website (copyrighted material) plus every ePub of a book (copyrighted material) plus every email that travels through Microsoft (or Google) servers.
Still not good enough. We need more data says the AI large language model. Our AI needs to be less shitty. Let’s get more data. Now it scrapes all things written in the worlds most populate word processor.
Therefore, when you ask for AI-generated text for a simple work email, you are in fact engaging in AI research and encouraging the violation of copyright. Same thing is happening in music (‘Hey, AI write me a breakup song in the style of Taylor Swift’).
There are times I see the benefits. I work daily with colleagues overseas. The AI-generated emails are often better than the write-natively-translate process that has been common. Cool, we communicate better. I can read-for-content in about six romance languages, but reading messages in my own English is easier.
Today, I spent the first hour of my workday on GoDaddy with the aim of un-breaking an email address. It took 7 phone calls and 2 chat sessions before I figured it out myself. The conversation devolved to ridiculous where I clearly had to battle against the common to find the deeply technical and obscure information that I needed. Every AI chat bot was just wrong. “Are you attempting to setup email on your phone or your computer” it asked at least three times in a row. Sadly, it was neither GoDaddy.
The words of my grandfather echo in my head. He had worked as a syndicated radio news guy during WWII on the NBC Red network. All of his peers and friends were swept up with television and landed jobs that gave them monster careers. My grandfather called TV a fad and opted to stick with radio.
I can not call AI a fad. It isn’t. There are billions invested in it. I embraced network technology in the 1980s as a young IT professional. I knew then that we were not heading towards a fad but something amazing. AI is the next bolt-on to our global internet. It is not a fad.
Some of what it does is funny. And some in scientific fields can happily celebrate huge successes with AI. The fundamental difference between AI for science and AI for creative endeavors is that from the git, science intended to live, breathe, and grow within the public domain. Creating AI models based on public data is legit.
The old rule was that if you paid for a service at Google or Microsoft for hosting your email and your office-like products (word processing, spreadsheets, presentations), then you were protected from your host. When you used their products for free, they used your data as their fee. They harvested your work for their research and you got products and services gratis.
That rule is gone.
I did move my writing to OpenOffice/LibreOffice. That is awkward and likely not much better. My work will be scraped and fed into the AI hopper regardless of my rights, my expectations, my efforts. I don’t even know how to protect myself or advise other on how to protect yourself. Gee whiz, I turned off the AI that captures my keystrokes. But I save the documents on a drive that Microsoft controls on their operating system with backups on their OneDrive. They have 100% access regardless. Yes, of course, I could build a linux based system and start from there. I have no access to Adobe there. Oh, and Adobe, you think that artists using your tools have failed to notice Adobe’s efforts to encourage the use of Adobe’s AI tools. I do NOT wonder where and how it got its data? Oh, right from their own paying customers.
I feared my legacy would be identify this tech as a fad and step away. Instead, I look at this tech for what it is and scream. What is a podcast, but radio?
I bought a fountain pen.