How Do AI Right So Good?
Recently I published a piece called, Who Is This Guy Al? Well, I looked him up using Artificial Intelligence, and lo and behold, he’s behind it all! Now that I’ve cleared that up, I figure it’s time to do a deeper dive into the realm. Having used AI several times consciously, and many times unconsciously, it’s important to understand how it works, its limitations, and how to best use it.
Obviously, AI is all the rage. Its technological advances have far outpaced our ability to control and build protections around it. That’s not to say that we have to worry about a global takeover akin to Colossus: The Forbin Project, but we must recognize the risks of allowing it to create misinformation or make critical decisions that we then build or act upon.
For my own part, I’ve consciously used AI to search for information, answer questions, and create memes. I’ve used it unconsciously as it’s anticipated my next words (there, it just did it!), corrected misspellings, or argued with me if I tried to make a clever turn of phrase, create a nonsense word, mix metaphors, or otherwise mangle our language. In all cases, I try to be vigilant, but also seriously consider its advice.
While it may be tempting to use AI to write essays like this, it would defeat my purpose: expressing my own ideas in my own words, and addressing feedback directed at me. I’ve found that I enjoy putting my thoughts into words on paper (or laptop), and organizing them into something that causes others to think, react, or be entertained. I prefer to be an originator, not an appropriator.
So, back to the question raised in the title: How Do AI Right So Good? ChatGPT, widely known as the quintessential AI tool, was released by OpenAI in 2022. ChatGPT, a Generative Pre-trained Transformer, was developed by training it on massive amounts of data to identify, learn and anticipate language patterns in order to generate new material, and allow it to transform a query into a response. GPT is actually not synonymous with AI, as it’s one form of it, but it is perhaps among the most useful and familiar.
GPT uses its extensive training to correctly spell and write responses to questions, create content, or to analyze and summarize documents, meetings, and communications. Other AI tools use algorithms to identify patterns as well, and can correct spelling, anticipate words, and suggest grammatical improvements. As such, GPT may be the most comprehensive and useful, but other forms of AI can help to bring clarity to information and communication.
A limitation of GPT is its inability to infer context or background if questions or situations are not well defined. Some forms of GPT may ask questions for clarification, but as one can imagine (which is unique to living things), our own thought processes are subject to biases, lived experiences, intuition, and other factors that artificial intelligence may be too logical to comprehend.
One danger with all forms of AI is the risk of a user accepting things at face value. Like the internet, which has its own issues and limitations, AI is dependent upon the data and information it is trained on and uses, and not all available information is correct or accurate. Even more fundamentally, AI is dependent upon its programming, which at some level traces back to human roots. Therefore, it is vitally important to recognize information it derives from other sources, and verify that those sources are well-founded and contextual. The user must be discerning and think critically about the validity of the information.
I’m reminded of a Morganism I developed a number of years ago: “As humans, we’re endowed with the incomparable ability to think for ourselves. It’s incomprehensible how so many of us refuse to do so, allowing others to do it for us.” AI is still developing, and we’re probably still a long way from seeing how far it will go. But AI is a tool and not a finished product. As long as we maintain that perspective, we can wield it deftly and improve our lives and the lives of those who depend on us.
p.s. Just for yucks (and perhaps out of curiosity) I had Microsoft CoPilot (my system’s built-in AI platform) review my essay. Aside from a few minor edits, reordering of a couple of paragraphs, and the introduction of subheadings, it left most of my writing intact. While my CoPilot didn’t concede, I took that as a win and didn’t change a thing!
p.p.s. Early in the essay, I mentioned using AI to create memes, but didn’t explain it further. I’ve generated a number of political cartoons and posters using MS CoPilot, and am amazed at its ability to take my directions and create something that looks professional. It usually interprets my instructions well, and allows me to refine them and make changes. The finished products are often better than I expected. It does, however, sometimes censor me if I take an idea too far, and make the political connections too overt. I can usually work around that though. Here are several examples of recent creations:
Well done Bob, and the closing graphics are outstanding. Your point that AI is a tool, programmed by humans at its roots, is one that we must remember even when AI becomes less Artificial and more "Intelligent".