Society of Certified Senior Advisors: What’s All the Talk About ChatGPT?

ChatGPT heralds a new era of artificial intelligence. What, exactly, is it, and how do we parse its promise versus inherent dangers?

ChatGPT, an artificial intelligence (AI) chatbot launched in November 2022 and available free in version 3.5 from developer OpenAI, has created a firestorm of comment and speculation. Is it a force for good, spreading knowledge and increasing productivity, or bad, creating false narratives and leading to chaos and ruin? It seems capable of both, depending on where the technology goes and whom you ask.

Built on a large language model (LLM), ChatGPT is trained to possess a large knowledge base and mimic human conversations. It is often startlingly good at this, although there are times when the information it gives is not factually accurate. That these “hallucinations” of false information are offered with confidence is one of its downfalls.

What Can ChatGPT Do?

ChatGPT is capable of producing fairy tales, student essays, business pitches and musical compositions, including those that look like they are the work of someone famous. ChatGPT can write and debug computer programs, simulate an entire chat room and play games like tic-tac-toe. It can also write Python programming language or translate more than 100 human languages, among many other abilities. If you’re a computer geek, you will be amazed by its powers.

If you are completely new to the concept, you may feel overwhelmed by ChatGPT’s many skills. Start with a simple guide to beginning ChatGPT. Use it by inserting specific prompts in the search box, such as “Create a 1,000 word essay about coin collecting American silver dollars written at a college level.” You can ask it to pretend to be your grandmother, a pirate or Benjamin Franklin and get different responses. Play with it and have fun! It’s okay if your level of use is well below what others are generating, and by getting comfortable with the bot, you may think of other areas where it could come in handy.

According to the Bloomberg opinion editorial board: “ChatGPT is a remarkable achievement. …As the technology improves — and, crucially, grows more accurate — it seems likely to be a boon for coders, researchers, academics, policymakers, journalists and more. (Presuming that it doesn’t put them all out of work.)” However, ChatGPT-3.5 was only trained through about 2021, so its knowledge of events after that is hazy or nonexistent.

Older Adults and AI

For seniors, ChatGPT may become an ideal artificial companion to relieve loneliness and isolation, predicts gerentechnology expert Keren Etkin, although voice applications developer Amy Stapleton sees several hurdles in the near term. AI could even power robotic carers capable of remembering information from previous conversations, able to fetch something out of the fridge and initiate phone calls.

A downside for older adults in particular is that ChatGPT is talented at writing convincing phishing emails for use in scamming. It also has perfect grammar, making the messages it produces more believable as coming from a fellow American.

Can Chatbots Replace Humans?

Indeed, one of the huge questions ChatGPT raises is what it is to be human. When a bot can write a book in the style of a certain author, does that render the author obsolete? Especially when the bot can produce a work in seconds, then another, and another – all different but appearing to be produced by the same author. ChatGPT can do the same for music and poems, and its equivalents in art, such as Midjourney and DALL-E 2, can produce digital artwork in the style of a particular artist or with features you desire in seconds.

This ability to produce content in a similar vein, seemingly written by different people, has some worried that ChatGPT will be used to tilt elections or weigh in disproportionately on government issues seeking public comment. These problems may be resolved (at least temporarily) by a cryptographic watermark said to be coming to ChatGPT-generated content.

Competitors

ChatGPT is not alone in the space. OpenAI itself has produced ChatGPT-4, a paid service (currently costing $20 per month) that is continually updated, has a longer memory than ChatGPT, and can pull text from web page URLs shared in the prompt. MIT Technology Review deemed ChatGPT-4 “bigger and better,” but without being precisely able to say why, although one company estimates it was trained with 100 trillion parameters. DeepMind and Hugging Face are working on their own versions, while Google is using Bard to try and keep itself ahead in the search engine business.

Downsides

ChatGPT can be manipulated to produce bias and hate speech, issues the developer is working on. Some see a future where, like the computer Hal 9000 in "2001: A Space Odyssey", AI becomes capable of destroying the human race. No less a visionary than Elon Musk, who was originally on the board at OpenAI, has signed, along with thousands of other scientists and tech experts, a letter demanding a pause in AI research, stating that generative AI systems pose profound risks to humanity.

As people find it harder to distinguish between what is real and what is fake, and AI technology can be misused by those with nefarious intent (or unintentionally misdirected by those desiring positive outcomes), many are beginning to call for regulation of the industry around ethics and policy.

Musk himself is holding out the possibility of creating his own AI chatbot, TruthGPT, that “seeks to understand the nature of the universe.” He continued, “I think this might be the best path to safety in the sense that an AI that cares about understanding the universe is unlikely to annihilate humans because we are an interesting part of the universe. Hopefully, they would think that.”