The Perils of Using Artificial Intelligence (AI)
A common theme of mine is the danger of relying on secondary source information (such as claims made in books or television programmes) compared to primary source information (such as accounts from people who witnessed an event first hand or documentation produced by people directly involved in a ship’s construction). We know from the study of primary source material that so many claims or popular beliefs about Titanic are not true. Many of these are addressed in Frequently Asked Questions (FAQ) posts on this website.
The dangers of relying on secondary sources are also apparent from using Artificial Intelligence (AI) tools. I recently came across a post on Facebook about White Star’s Oceanic (1899). Underneath, it had a prompt to use Meta AI to answer various questions. In response to clicking on one of these, this tool claimed that Oceanic held the Blue Riband at the time she was launched. This is simply not true. She was not built to be the fastest liner afloat or to hold the Blue Riband, which was an unofficial speed record held by a succession of German passenger liners from 1897 to 1907. Moreover, it would have been impossible for her to demonstrate any kind of speed at all until she had entered service. At the time of launch, her propellers were fitted, but her propelling machinery and the key mechanical components to operate the ship were not yet in place!
Another example is a post I made recently on my Mark Chirnside’s Reception Room Facebook page. In response to inaccurate (apparently AI-generated) information on another Facebook page, I shared a link to my article discussing Britannic‘s length and breadth. My article provided detailed evidence demonstrating that, contrary to popular belief, Britannic‘s breadth had been increased prior to her keel being laid. It had nothing to do with any of the design changes that followed the Titanic disaster (specifically, the inner skin running the length of the boiler and engine rooms).
Beneath my post, Meta AI once again had a series of prompts. One of them was a question asking ‘How did Britannic‘s width change?’ I clicked on it and saw that it made exactly the same untrue claim that my article had already disproven, as well as adding further confusion by saying that she was ‘a few feet wider than its [recte: her] sister ships…’ In fact, she was only eighteen inches wider in total (1.5 feet). This shows the challenge in providing accurate information to people when tools such as Meta AI generate so much inaccurate information.
Although tools such as AI potentially have their uses, it is important to verify any information obtained from them independently!