Should the AI take the proverbial "bird's eye view" to see the whole of the Amazon forest, would it flag damaged goods? From my point of view I'd say yes.
Amazon is reportedly deploying artificial intelligence to inspect goods and look for signs of damage before they are packed and shipped to people. Jeremy Wyatt, director of applied science at Amazon Robotics, said workers today check the quality of items as they package and sort them – but they often fail to spot damaged …
Easy to poke fun at this but it looks like boilerplate for what ai and ai driven robots are going to be used for.
By the time the politicians have finally actually worked out what the problem really is, it will be too late.
At some point we have to decide if we want there to be low grade repetitive jobs with people doing them.
Also, if I was Amazon middle management I'd be more concerned about my future in the face of ai and wonder if downing tools to protest about climate might be wise...
"we have to decide if we want there to be low grade repetitive jobs with people doing them"
This. There's no reason to expect that net jobs available will decrease, since they have been on a consistently upward path ever since automation began. I expect there will be more jobs in the service industry and more higher-skilled manual jobs (for example robot mechanic). The dangers as I see them are (1) change happening too rapidly to allow gradual workforce retraining, leading to a combination of massive unemployment at the same time as a huge amount of vacancies for high-skilled positions that can't be filled due to lack of resources and (2) too much wealth continuing to accrue in fewer and fewer hands, which has accelerated highly in the last 20-40 years
Actually there is plenty of reason to expect that job availability will decrease short term as you suggest.
Also the expectations of the level of skill & education for future replacement jobs will be higher.
The Luddites were mostly illiterate, 1970s factory jobs were similar, there are very few jobs that don't expect you to read & write nowadays.
Training and legislation will be most of the answer, time for governments to govern - oh dear we are screwed!
As you say wealth accrual is going to be significant, I wonder if I will be a Bezite or a Muskie slave?
Probably not, since it's just looking at the package it comes in, but it's not designed to do every task they have. It's just intended to reduce their costs from dealing with stuff that's sent back by people who are unhappy with its quality, and they want to catch the things that are likely to lead to such complaints. That isn't going to eliminate the existence of fraud.
That said, you could always train a new one on all the products that are typically sold and give it an x-ray generator so it can investigate the contents of packages. Why does this not sound like a good idea?
Three times not much is not much better.
This is supposed to be AI, right ? Why is it not 99% efficient ? Is it so hard to detect that a package is not what it should look like ?
If anything, I would have expected complaints about the fact that a mere scratch sent the package for evaluation, but apparently it takes a lot more than that to declass a broken package.
Either that or the system is shit at its job.
So, your accepted outcomes are:
1. Near complete perfection.
2. Ridiculous false positive rate which causes complaints
3. The system must be entirely bad.
These seem to leave out a lot of options. Perhaps the three times better than a human deal is because, if they ask for 99% accuracy, they get too many false positives? Maybe my incredulity at your choices is because I'm having trouble seeing "three times better than a human" as a bad result. Sure, it would be nice if it never made a mistake, but it's a visual recognition task which is not deterministic.
For that matter, once it's built, the acceptable rate might have been if it was equal in accuracy to a human, because a computer with a camera running software you already have is probably cheaper than a human who also has a camera. Things that are three times faster or better at performing a task as I am tend to be rather useful things.
"deploying artificial intelligence to inspect goods and look for signs of damage before they are packed and shipped to people"
It would be great if someone could use AI to do something similar for goods in transit. I had a custom made door delivered with two corners broken off, and the same company delivered the replacement with a huge dent in one side. Not sure how AI could help, but HI (human intelligence) doesn't currently seem to pass muster.
My SO worked for a certain relatively popular baby goods store up here that was sacrificed on the altar of corporate greed at the beginning of this year. (The things that DONT make the news). They received and sold standard packaged goods of all baby sorts, and also carried several lines of rather well made baby furniture. Said furniture was indeed made overseas, and then container shipped to the Good Old US of Hay. Whereupon goods destined for the units north of the 57th were transshipped through 3 (three) different shipping companies and up to 7 warehouse transfer stops. Suffice it to say that on average, 3 in every 5 units suffered from outrageous amounts of shipping damage. And you know of course, the shipping firms insisted that it had nothing to do with them, the transfers, the repacking on new skids, the onloading or offloading (no that slot in the bottom of the box that goes all the way through the case and is 2" by 6.5" has nothing to do with the forklift operator stoned out of his gourd in Philly moving it off the truck) or the overnight parking in uncontrolled and unmonitored roadside way-stations. Absolutely was the fault of the people receiving the goods at the end point.
Perhaps putting that tech on cameras at the loading/unloading doors of trucks, warehouses, aircraft handling equipment etc might be better application of the tech?
From an Amazon Associate at a robotic FC, I can say that these issues aren't because they get missed, it's because we're expected to do too much, too fast. Stowers have to do so much stowing, as well as having to do 6 sided checks of everything, as well as using safe working techniques (proper bending and lifting etc) all within a certain time, as well as hitting rate. They have automated cameras on stow now that completely miss the correct bin where the stower puts things. When it comes to pick, we get sent to a bin opposite! The manual scan-item-then-scan-bin-after stow method was much more accurate.
Pickers are the same, in reverse. We have to pick the item, do six sided checks, and check for stuff that the 6 sided check would miss (melted deformed chocolate, or a smashed Easter egg in an otherwise perfect box, it does happen, for example), as well as hit rate of 330-500 depending how many packers we have on us, and their individual rate. We have to do so much, that people just don't bother. Damaged stuff gets stowed, then picked, then packed, because the motivation isn't there, we're worked to death.
Personally I hit the middle. Accurate but not the fastest, and I catch a lot of stuff that would otherwise slip through. When management come to me threatening ADAPT, I ask them to choose between quality, or having £1.6 million of damages a year at our FC. They never come back to me for months.
Those £1.6 million of damages doesn't cover stuff that falls out of robots on the AR floor either AFAIK. Amazon's idea of robots replacing us is a fantasy, a massive one. There is no way robots are as quick or as dextrous as humans, they won't be for years yet. The tech is still hit and miss, full of delays.
Just a reminder that the America that can import almost all of it's basic manufactured goods via Amazon from over-there is not sustainable. Here's an answer I found online to the question "are there any motherboards made in the US" that could be re-phrased "outside of Taiwan/China":
There were Chaintech things made in the USA (now their California facility is just a branch office) in the past, but while they started good, they quickly became cost-cutting bottom feeders, and wholly unable to compete with high end Chinese quality. Trenton Technology might still be active, manufacture in New York, but last I heard their quality control was pretty awful, and their single board computers more expensive and more poorly designed than proper Taiwanese goods. Might be why they're called Trenton Systems now - not sure if they're under new ownership.
PNY manufactures in the US, and has managed to keep quality at least on the good side of average (and usually up there with the best of the Taiwanese), but PNY doesn't make motherboards.
EVGA is based in California, USA, but manufactures in China and designs in many countries. Intel desktop boards *used* to be made in the US, but I think even they're outsourced now. Why does the US suck for quality and end up expensive? Because all the components are made in China and have to be imported. We in the UK are the same: The makers of the Raspberry Pi wanted to manufacture in the UK, but it was much more expensive to import the components than just get it fully manufactured in China. As this is expensive, it means quality has to be cut in order to stay competitive, or price has to be raised. Often both.
This ties into the question - is the US (or UK for that matter) budget deficit sustainable? It's not the deficit directly that is the problem, it is that the luxury of the deficit has enabled the withering away of the capacity and ability to manufacture the basic building blocks required for a functioning economy.
The present day Amazon-marketplace will not remain as it is when the shit hits the fan.
when restaurants are fully automated, for then I shall break into one and live out my life eating porcuswine burgers and drinking whatever passes for liquid refreshment for the rest of my life. I shall get away with it my adding my meal to the bill on larger orders, so the money tallies correctly for the corporation. I will only need to hide once a month, when the delivery truck delivers the raw materials for the printers that make the food.