Over the holidays, I had to call an online retailer’s customer service line to find out why a gift was taking so long to arrive. After going through a few keypad prompts, the system connected me to a very pleasant, automated female voice that offered to help me and asked me to describe my problem. The fact the voice belonged to a robot was barely discernible.
I prepared for circular phonetree hell as my thumb hovered above the 0 button, but I went ahead and described my problem. The automated voice replied by asking for my order number, which I had printed out in preparation.
Having dealt with automated voice assistants before, I expected it would transfer me to the correct human customer service rep, who would resolve my issue. Instead, without missing a beat, the chatbot explained that the order was being held by the shipper because the delivery driver didn’t have the subdivision’s gate code. That’s when I realized that I was talking to a chatbot more advanced than I had ever encountered.
The chatbot then explained to me that it would send me an email where I could respond with more detailed delivery instructions. The email was there by the time I opened my Gmail and the package was delivered the next day.
I spared a thought for all the poor customer service reps, who had organized their post-pandemic lives around remote work, allowing them to care for family, work from the road, or just avoid a daily commute. I recalled that many of these low-paid workers had been refugees of previous waves of economic displacement after millions of manufacturing jobs were offshored.
Not Your Average ChatBot
A few days later, I learned about OpenAI’s newest release of Chatbot GPT. After watching a few YouTubes, and test driving the app online, I realized that the risk to customer service workers was just the tip of the iceberg.
GPT (Generative Pre-training Transformer) is an AI neural network trained on a massive body of online text to generate natural, human-sounding responses. It can answer complex, compound questions, write essays and stories of fiction within defined parameters, generate elegant computer code, and even give medical advice — always with disclaimers that the user must perform real world verification.
That’s because, as of yet, the AI behind GPT hasn’t compared its responses to real-world results, but have no doubt that will happen quickly — probably within a year. When it does, it will be faster, cheaper and more intelligent than any human workforce across a whole range of occupations, including journalism, teaching, healthcare, and computer programming.
Ironically, OpenAI was founded in 2015 as a non-profit by Elon Musk, Peter Thiel, and other tech investors to prevent artificial intelligence from being centralized by making its patents and research public domain. The thinking was if advanced AI was broadly distributed, abuse would be less likely.
But Musk left the board in 2018 over a conflict of interest with Tesla AI, and in 2019, the board changed the company to a “capped” for-profit, offering potential profits of 100X to investors. Microsoft stepped in that same year, buying a $1 billion stake, and the company promptly announced it would commercially license the product.
On the OpenAI website, I asked ChatGPT what the worst possible outcome of its implementation would be for human employment.
“The worst possible outcome of human employment of ChatGPT would be if it caused widespread unemployment and social upheaval, leading to societal collapse,” it replied.
Later that night, I began rereading my old high school copy of Aldus Huxley’s Brave New World.