AI and ChatGPT: There’s Less There Than Meets the Eye (And More Too).

The media are filled with stories about artificial intelligence (AI) and the newest version called generative pre-trained transformers (GPT).

The ChatGPT app you’ve heard so much about is just a user-friendly version of GPT-AI that you can interact with. You download ChatGPT and then ask it to answer questions or perform various tasks. You can also give the app raw materials such as spreadsheets or other data to help it perform those tasks.

I have a friend who uploaded a business model spreadsheet to ChatGPT and asked it to write a typical Wall Street-style research report on the company. The report came back in minutes, and my friend rated it better than most actual Wall Street research.

Needless to say, high school and college students were early adapters asking ChatGPT to write term papers and answers to take-home exams. Reports are that the term papers are also excellent, and plagiarism is not a problem because each product is unique. So, yes, the GPT version of AI does have amazing capabilities, and it is taking the world by storm.

AI itself is an old story. Aristotle speculated on what the world would be like if a spindle could weave cloth without human hands. Alan Turing (father of theoretical computer science and the man who cracked the Nazi Enigma code) wrote essays on the ethical dilemmas of AI in the 1950s.

My first discussions about AI occurred with a consultant from Arthur D. Little in 1979. The concept has been around for decades. What changed was the massive increase in available texts (starting with the internet, world wide web, and search engines in the 1990s), and the exponential increase in processing power (with NVIDIA and other processors in the 2010s).

The missing pieces were large language models (LLMs) and natural language processing (NLP). When these elements were combined with GPT algorithms (that allow the computer to be “trained” on certain texts), the result was apps that could read billions of pages of text (LLM) in a common-sense kind of way (NLP) and produce answers in plain English.

Are these machines poised to take over the world? Not really. To downplay them a bit, they’re just speed-reading machines that can spit back results with some nuance.

They also make things up. That’s what I call the “puppy factor,” where the machine wants a pat on the head for answering your question the way a puppy wants a pat on the head for fetching a ball. The problem is the machine will invent facts in order to please its master.

One writer asked ChatGPT to write a biography of himself. The machine promptly did so, including his date of death and an obituary.

A real and immediate problem does exist, as described in this article. ChatGPT is offered by a consortium controlled by Microsoft. It is a threat to Google because GPT tech means the death of search engines.

Google has responded with its own GPT app called Bard. Google engineers have warned management that Bard is a “pathological liar” and offers solutions to technical problems that can cause injury and death. We’ll leave it to readers to decide if and when they want to use GPT technology. Just take it with more than a grain of salt.

Corporate leaders and institutional fiduciaries looking to incorporate state of the art predictive analytics to their risk mitigation and strategic analysis should click the link to learn more about Raven Predictive Analytics®.