I used ChatGPT to re-write my post about ChatGPT.
The Internet is abuzz with OpenAI's ChatGPT, evoking thoughts in me as I watch others and read about its output.
Machine learning algorithms need data to build their library. OpenAI reportedly crawled the open Internet for content. We can ponder what this may entail as we move on.
The bot's responses must be verified independently, as they may not be accurate despite its confidence. Humans must check and confirm the output.
AI/ML are as smart as a chicken, yet the results of ChatGPT output have been impressive. Astonishment often arises when people encounter new knowledge, like a PhD student's deep understanding of forestry in British Columbia or being unaware of the latest events in the Royal Family.
A concern is that ChatGPT is now being trained on its own output, which will likely lower the mean. The average response will become less impressive, as it feeds upon its own output. Without strong input from OpenAI, ChatGPT could devolve into Idiocracy.
ChatGPT is a tool, and the output depends on its wielder. The same tools that can create great art can also scrape old crust off a table. The handler has the largest influence on the output. For now, it's just an average response statistical model, not the world-changing AI we've been promised.
ChatGPT, much like the famous Clever Hans, is highly convincing in its superficial display of intelligence, but is nothing more than a clever illusion when examined more closely.