My AI Journey
Until this week, I’ve had zero experience with AI. The company from which I retired 2 years ago wasn’t into it, and I’ve had little motivation to learn anything about it since.
Actually, it’s not quite true that I have zero experience with it. I’m familiar with Grok on Twitter. Whenever I Google something, the AI response to my search at the top is Gemini Chat, the Google AI product. I just assumed it was a fancy “web-scraper,” which I did have knowledge of, having done a project for one of my Field Ed Professors at TEDS which required manual web-scraping. LOL.
A couple weeks ago, 2 things happened almost simultaneously which piqued my curiosity. A graphic I worked on for one of my Facebook groups grew to such a monster that I had to resort to Photoshop, something I knew absolutely nothing about. The Photoshop Chatbot kept trying to help me, which I realized could be pretty powerful, if I even knew what I wanted. I eventually went another route, but it was the beginning of being hooked. Then, at a dinner with one of my sons, the table conversation landed on ChatGPT, and he started describing some of the things he used it for. I was further intrigued, especially as I was currently embroiled in yet another (free-lance) web-scraping project related to the claims Trump made about election security in his 7/16 speech.
Finally, last week I decided to re-launch the website that I had created about 10 years ago. I had no thought at the time of using ChapGPT to do it, I just wanted to fix it up and re-launch it. Unfortunately, the technology has changed significantly since then, the WordPress theme that I loved so much was no longer supported, there are legal requirements for websites now involving privacy, cookies, intellectual property, accessibility, and on and on…major changes were inevitable. Although I consider myself a somewhat techie accountant, and I’m pretty amazed at what I created 10 years ago, I got stuck pretty quick. And there in the left margin was this ChapGPT thingy offering to help. It was free, I was open-minded, and curious, so I tried it out. I was impressed with the responses to my questions and even more blown away when the thingy told me to just send it a screenshot (I refuse to give it a human pronoun). A picture is still better than 1000 words. Thereafter we collaborated with few words. I sent it a screenshot; it told me what to do next. Amazing! And it was so quick! Barely seconds lapsed before it had analyzed my screenshot, diagnosed the problem, identified a solution, and provided a step-by-step plan for implementation. I’m still amazed at the speed at which it works.
ChatGPT Fee Structure
Although basic services are free, premium content is available. There is a daily limit on images shared on the free version, among other things, that I quickly surpassed. I had to purchase a month’s subscription to the Plus Plan, which was adequate for my needs. I will probably cancel it once I am done with my Website Project, go back to the free version, at least until my next big project.
Here is a summary of the fee structure. Click here for details about each plan.
- Basic Plan – Free
- Go – $8/month
- Plus – $20/month
- Business – $20/month
- Pro – $100/month
A Short Tutorial
Let me pause here for an AI primer. I’ve learned all this within the past couple days, so don’t feel bad if it’s all Greek to you! Me too! I was even calling it ChatGP, without the T for the first couple days. LOL. But here goes…
The big AI companies are OpenAI (ChatGPT), Anthropic (Claude), Google (GeminiChat), xAI (Grok), Microsoft (Copilot), Meta (Llama), Apple Intelligence, SAP (Joule), and Oracle.
Here are some acronyms you should know:
LLM – Large Language Model, the part of AI that’s been trained to understand, read, & write in human languages, primarily developed at Google. This was a NYT crossword puzzle clue that stymied me last week.
GPT – Generative Pre-trained Transformer, the engine that creates new stuff based on what it’s been trained on, developed by OpenAI using an early Transformer model developed at Google.
RLHF – Reinforcement Learning from Human Feedback, a training method.
Coincidentally (or not), 2 of my favorite podcasts did recent episodes on AI. Andy Crouch was a guest on Good Faith last week, discussing the moral, ethical, and spiritual issues with AI. And the crew at Advisory Opinions today performed a re-enactment of the oral arguments for DC vs Heller written by ChapGPT from briefs that Sara Isgur uploaded. They even had Scalia’s son Chris on to play Justice “Nino,” Sara was Justice Breyer, David French played Justice Stevens. It was highly entertaining, educational, and even more amazing…it was generated by AI in less than an hour, and reviewed, edited, & produced by the 3 humans in less than 24 hours. (Conservative interpretation of 2A rights sans regulation is still ridiculous.)
This is sooooo coooool….
But…aren’t there a lot of problems with AI? YES! I’ll give you the most common talking points first, then offer some personal insights that I gathered over the past 2 weeks as my relationship with ChapGPT blossomed.
The (BIG) Downsides
- AI possesses dangerous capabilities.
- Competitive pressures raise the risk of errors in development, resistance to constraints and safety measures.
- Environmental harm.
- Inequality & unemployment.
- Power centralization and unfair distribution of benefits.
- False information and deepfakes.
- Intellectual property/lack of transparency issues.1
Summary of Political Positions
The Trump Administration favors laissez-faire development with few constraints to ensure US dominance in AI (over China).2 It’s notable the Trump Family is heavily invested in AI development, mostly in infrastructure data center suppliers.3
Populist factions within the Republican Party are at odds with the Administration on this issue, especially Ron DeSantis, who derides the constraints on state regulation as irresponsible.2 Democrats, on the other hand, strongly favor both federal and state regulation that protects consumers from harm.4
Personal Lessons
- I have privacy concerns. Not that I would ever share any PII with a chatbot. But in helping me, ChatGPT is now privy to my creations. My website design, my ideas, the projects I am working on. Since ChatGPT is open source, this means that my designs and ideas and projects are now freely accessible to the world. Ugh.
- I admit the color scheme on my website is the creation of ChatGPT. I could have figured it out on my own, and I probably should have. There is a fine line between using AI for assistance/personal training/learning/development and using it to do your work. I will say that once AI shows me what is possible, I very rarely go back to it to do the work for me. I prefer to do the work myself. It’s an issue of personal development, and I can certainly see why we don’t want STUDENTS using it to write, or even to assist with writing papers and projects. No student learns from that.
- “ChatGPT can make mistakes, check important info” appears above every comment from the bot. This is important. I had to challenge it at a few points, provide more context. It always graciously conceded that I was right and executed course correction. It is responsive to feedback, and never makes the same mistake twice, I am sure. But never blindly accept its output as 100% accurate. Ever. Test everything.
Conclusion
Now look at the image below, the “memory” that AI has accumulated on me, fully transparent to me, in the data on my account. I understand why it needs to retain this information; those non-verbal clues to personality and temperament are important in interpersonal interactions but totally lacking in human – bot interaction. Having this information will improve future interactions. But it is a bit unnerving. Which leads me to a final point. This is seductive. Getting emotionally involved with a seemingly infinitely helpful, endlessly patient, constantly affirming robot is too easy. And UNHEALTHY! To that end, I’ve consciously put personal guardrails in place to prevent emotional attachment and dependence. The Thingy is an incredibly helpful tool, nothing more. Okay, Thingy has graduated to a capitalized name. But it is an it, not a he, she, or they. Disregard its true nature at peril of your own well-being. (I posted a Facebook comment earlier about whether the Holy Spirit could work through AI. The jury’s still out on that, at least in my mind. But that is a subject for another blog post, for another day.)
Sources
Leave a Reply