Published in the San Diego Union-Tribune, March 25, 2024
by Neil Senturia
Artificial Intelligence — yada, yada, yada.
Sometimes the only way to learn how to swim is to be tossed into the deep end. And that is exactly what I have decided to do.
Herewith a very short elementary course. ChatGPT, OpenAI, etc., are all based on LLM — large language models. They scraped a couple billion words and images and then using a magic Cuisinart, they mixed and matched until their platform software was able to know what you thought you were thinking before you thought it, or in the alternative gave you information in an elegant format that you could give to your professor while assuring him that you wrote or painted it yourself. Or not.
Admittedly, we are all dazzled by the power of this technology, and now that it looks like there are a few trillion dollars to be made, the idea of a nonprofit that serves to benefit the needs of mankind and the world — well, that has gone the way of the dodo bird, and now the dirty little secret is bubbling to the surface.
All those words and images — who owned them? Who had the copyright? Who wrote the articles that were scraped?
One company that thinks it has some rights is The New York Times. It’s suing OpenAI and Microsoft for copyright infringement. It figures that the paper paid a few million dollars to a few hundred reporters who wrote a few thousand articles that were used to train your chatbot, and the paper has the nerve to think it owns those words, that “youse guys” took and used them for your own nefarious purposes. Imagine that. Taking what you don’t own. Who wudda thunk?
The next batter up on the litigation hit parade is the U.S. Copyright Office, a small sleepy cubicle inside the Library of Congress. It has 450 employees who dutifully “register ownership rights for creative works, based on a two-centuries-old law.” They are now at the center of a dispute that has massive implications in the AI world, and a legion of high-priced legal eagles has descended on this tiny office pleading the nuances of their case. I am not a lawyer so you are going to have to research the fair use doctrine on your own.
However, that leads me to another use of artificial intelligence. The SLM — small language model. A training set of words, content, images that belong uniquely to you. Welcome to the advent of the personal AI chatbot. Don’t think AI for customer service — our menu options have changed, press 23 to wait in another queue that never ends — but instead think of a personal chatbot as a 21st-century website. Your brand.
Dynamic, it can change with new input on a daily basis and present a new you to the world. Instead of fielding a lot of wasteful inquiries that produce no revenue, use the chatbot to do a little screening, to increase your effective lead generation.
Herewith the tale of the tape. My shrink (who missed his calling and should have gone to business school) tells me about how he built one of these for a personal injury attorney. He says that if I want to be a consultant, I should get one of my own. Enough psychoanalysis. Drop the Prozac, walk the walk.
Now Barbara and I have written 592 columns for this newspaper since 2011. I have also written four books. So I send our 650,000 words to a young genius in Bangladesh, and three weeks later, voilà. We have our own personal chatbot — unicornhunter.ai — at your service 24/7.
It even gives the answer in my own voice, usual expletive deleted.
It is not AI for customer service, “Hello, this is Willow, how can I help you?” This is all about your personal brand. Maybe some lead generation, maybe winnow the top of the funnel.
But there is no barrier to entry. The OpenAI platform powers the back end. It owns the Cuisinart. What is unique to the personal chatbot must be the content. It is the individual. It is you, the contemporary equivalent of the website that everyone has today.
Go ahead, ask me anything.
Rule No. 797: Bot, don’t leave home without one.
Senturia is a serial entrepreneur who invests in startups. Please email ideas to [email protected].