Hi - just look for the work of Patrick Fuchs (www.ndt.net/search/docs.php3?showForm=OFF&MainSource=-1&AuthorID=38978). He has done some great work using U-Nets on Computed Tomography results.
Very insightful and a great overview of neural networks. The regulations and legislation around the use of AI in this regard are tricky though especially in highly regulated fields like aerospace, somewhat similar to problems with self driving cars. If the AI model fails to correctly identify a defect that ultimately leads to a failure who bears the responsibility? Or if it identifies a false positive that results in days of downtime and lost revenue. Is it the operator of the test set with AI implementation, the company that designed the AI model or was it the company that provided dubious training data? AI technology is currently advancing far faster than regulatory bodies can legislate for. We're certainly in for an interesting few years.
You are absolutely correct. AI does not eliminate human factors. With AI the human factors are shifted from the inspector to the programmer/trainer. The benefit: if done correctly this can bring a great boost in reliability but if it is done incorrectly the issue will be applied consistently and all results might become questionable.
You are a great persoanje
Muchas Gracias
full of knowledge, thank you so much for the video 🎉
You are very welcome
great video
I'd like to have more details about that U-Net process, but still the video is great
Hi - just look for the work of Patrick Fuchs (www.ndt.net/search/docs.php3?showForm=OFF&MainSource=-1&AuthorID=38978). He has done some great work using U-Nets on Computed Tomography results.
Very insightful and a great overview of neural networks. The regulations and legislation around the use of AI in this regard are tricky though especially in highly regulated fields like aerospace, somewhat similar to problems with self driving cars. If the AI model fails to correctly identify a defect that ultimately leads to a failure who bears the responsibility? Or if it identifies a false positive that results in days of downtime and lost revenue. Is it the operator of the test set with AI implementation, the company that designed the AI model or was it the company that provided dubious training data? AI technology is currently advancing far faster than regulatory bodies can legislate for. We're certainly in for an interesting few years.
You are absolutely correct. AI does not eliminate human factors. With AI the human factors are shifted from the inspector to the programmer/trainer. The benefit: if done correctly this can bring a great boost in reliability but if it is done incorrectly the issue will be applied consistently and all results might become questionable.