I was there, there was a yellow disc in the sky (aka the sun), I was sitting and merely contemplating stuff when I got hit with a video.
It was 60 minute with something on underwater smuggling and how people were unprepared. It took me 15 seconds to set that premise to solved. OK, Google or Amazon need to get involved. It is not ‘that’ easy, but that is what Deeper Machine Learning is for. Funny enough, my ships engineering skills (outdated since 1981) got into field and my thought patterns resembled one I had in UTS when I came up for a system to weed out false positives in bomb detection. Whilst everyone was focussing on where the bomb was, I decided to look at a way to remove false positives which took mere seconds and when you have 4 million passengers a year, having certain points where you can scan a passenger in less than 5 seconds matters. The fact that you weed out 80% of the false positives also matters as it suddenly leaves you with a manageable number of people and with Deeper Machine Learning that system merely gets to be more accurate, as such within months that number would have increased to 90%+ which makes is an possibility. It was merely a concept but I was happy (as was my professor). Now we get back to the story. You see, it took seconds to find this puppy.
Here we have a commercial Japanese solution of a underwater drone. It is not enough, because we have to tinker with it and to make a drone an autonomous underwater vehicles (AUVs) takes work and the battery would require an update, the function and the added hardware will be murder on the regularly installed battery. The nice part is that these puppies do not need sleep and they could scan the hull of any vessel in minutes. Two might get it done in a minute and now we get the setting, a set of two one to scan and one to validate the scanning by weeding out the false negatives. Hulls are simple, they are one setting, they are smooth and waterproofed. The idea that a hull is tampered with is not laughable, but it tends to be slightly ridiculous, as such an ‘adjusted’ hull is noticed by any AUV and teaching it a few additional things is not hard, not for the right Deeper Machine Learning expert. As such we need to consider like an autonomous underwater vehicles (AUVs). You see a place like New York might have millions of containers a year, but it does remain a relatively small about of vessels, as such a dozen drones would be able to scan all the vessels BEFORE they dock and that is the busiest port in the world. The drones could also be scanning for other things, like divers going on a tourist tour past any vessel which would be a big no-no.
These settings alone were solved (by little old me) in less than a minute, so why were these methods not considered? Perhaps they were and they found a snag I never considered. I am not prefect, but I try to see the solution in a challenge, not the hiccup.
Still the exercise was fun for the minute I had it, it gave me something else to consider for a moment. And when you think on how I got there, wonder what else I can come up with tomorrow, but that is a setting I will consider in 18 hours. The drone will need adjustments too, scanners on the top (two sets) facing 30 degrees up and 90 degrees up, it also needs to be altered into an autonomous underwater vehicle (AUV), which will a little work. So when we added the initial and verification scan, we get a vessel with the ability to do it at the same time and it is done in seconds per 10 metres. The learning curve needs to be adjusted and it can be set against type of vessel. You see a coaster, a tanker and a cargo ship have slightly difference hulls, but the same principle applies, waterproof or sink. It is really that simple at times. The smugglers ‘adding’ a box at the hull will fall through the hoops in the initial minute and as such the boys in blue (with flippers) can capture the haul. The ones who were clever and added a ‘valve’ to allow the merchandise to sit between the outer and inner hull is a little harder, but when Machine Learning considers that these valves should not be there, the pattern adjusts as well. This will create some initial false positives, but there is also the gain that we eliminated 90% of all vessels making this a relatively easy exercise.
Wow, 3 minutes of my brain productively used. I am getting good at my old age. So consider this a concept, consider this a joke, it is all up to you and the boys in blue. I did my bit on Saturday and I am not going to get paid for it, so use it as you see fit.
I am now 230 minutes from Sunday, have fun and enjoy the sunshine (if there is any).
Lawrence van Rijn - Law Lord to be 