LawrencetheMDer
Active member
I imagine somewhere someone is building an AI metal detector. Combining AI with metal detectors (AIMD) might/could/will change metal detecting as we know it. The advent of AIMD will be a shock to the field, somewhat similar to when metal detectors first became relatively easily available and swept beaches clean in the 60s and I believe as early as the 1930s. Everyone would be an expert in the use of such simplified machines. Yikes!
AI algorithms are already readily available (e.g., ChatGPT) and are large language models and are trained on, basically, all the words on the internet or on a particular company’s sales force text database. Theoretically, such algorithms could be trained, not on words, but on parameters specific to the electrical field properties, at different frequencies, of the output of metal detectors. Training would involve not words but say 2000 gold rings (specifying weight, size, ct) to train the AIMD on such rings. Next training would involve, say, various coins, again 1000s of them. Next training would involve pull-tab, then push-tabs, then their various parts. Lets’ not forget those pesky bottle caps, again we’d need 1000s of them. But once done, what a heck of a detector we would have.
AIMDs will differentiate pull tabs and push tabs from gold and silver rings. Imagine no VIDs, no 2-D images, perhaps no screen at all! The AIMD would simply state “gold ring”, “quarter”, “pull-tab”, etc, when the coil went over the target. And unlike the early metal detectors of the 1970s or 1980 (?) that would specify “coin”, for example, AIMDs would be deadly accurate, say 99% or, ok, 98%.
Well, off to hunt. Gotta find those gold rings before AIMD finds them all. The availability of AIMDs would cause me to become an “Early Adopter”, first time in my life.
Happy Hunting
AI algorithms are already readily available (e.g., ChatGPT) and are large language models and are trained on, basically, all the words on the internet or on a particular company’s sales force text database. Theoretically, such algorithms could be trained, not on words, but on parameters specific to the electrical field properties, at different frequencies, of the output of metal detectors. Training would involve not words but say 2000 gold rings (specifying weight, size, ct) to train the AIMD on such rings. Next training would involve, say, various coins, again 1000s of them. Next training would involve pull-tab, then push-tabs, then their various parts. Lets’ not forget those pesky bottle caps, again we’d need 1000s of them. But once done, what a heck of a detector we would have.
AIMDs will differentiate pull tabs and push tabs from gold and silver rings. Imagine no VIDs, no 2-D images, perhaps no screen at all! The AIMD would simply state “gold ring”, “quarter”, “pull-tab”, etc, when the coil went over the target. And unlike the early metal detectors of the 1970s or 1980 (?) that would specify “coin”, for example, AIMDs would be deadly accurate, say 99% or, ok, 98%.
Well, off to hunt. Gotta find those gold rings before AIMD finds them all. The availability of AIMDs would cause me to become an “Early Adopter”, first time in my life.
Happy Hunting