What I find sad is to see startups here in Switzerland where we already separate and recycle extensively that offer a service where you just toss everything that can be recycled in the same bag. So batteries are mixed with plastic and cans. These bags compete with government collection facilities and are "reprogramming" people to be lazy. Regular trash is very expensive here and people who want to safe are forced to separate and recycle what can be.
The hardest task of recycling is the separation of items. Centralizing this does not scale as you need to hire more and more underpaid worked to do this task. The task that was done for free before by the person throwing the item out. All the facilities had to do is sort the remaining small percentage of miss sorted items instead of all items.
Sure it does.[1] What doesn't scale is manual sorting. It complicates collection, because there are all those bins to empty,the trucks are more complicated, and humans aren't good at separating the different types of plastics. That's easy with multispectral imaging, but humans lack that capability.
Bulk Handling Systems seems to be the US leader in this. Robots are used but do under 10% of the job. Most of the sorting is done by screens, shakers, drums, DC and AC magnets, and air-jet optical sorters. The robots only do "quality control", pulling out stuff that the heavy machinery mis-sorted.
San Francisco has one of their plants.
Aluminum is easily separated and recycled, and is the only profitable part of the operation. Steel gets separated and melted down, and about 40% of US steel is now recycled. Which is why Nucor, which started as a recycler, is the biggest US steel company. Some kinds of plastic bottles recycle well, but it doesn't really pay. CarbonLite, with three big plants, went bankrupt in 2021. The buyers kept the plants running, but the investors didn't get much. CarbonLite was caught between high fixed costs and not enough market for their product.
There's not much use for mixed paper. Offices used to generate large amounts of good quality paper, and newspapers were a consistent source of uniform product. But now, most paper is discarded packaging, with too much else mixed in.
It's hard to make money at this. Separation, though, is not the problem.
I saw one of those how they do it shows where the host went to an aluminum recycling plant, and this is how I learned there are lead alloys out there. They use an xray machine to sort those.
My favorite thing about some of these sorting systems is that they are online but not instantaneous. For any reasonable rate of objects passing by the scanner, they can determine yes/no (or maybe/no) in a fixed amount of time, say 10ms for illustration purposes. So the picker is 10ms down the conveyor from the sensor, and the signal arrives at the precise moment to activate the picker. More computers or faster computers let you increase the flow rate by increasing either the density or the speed of the conveyor, but it's always a realtime system.
For instance I don't think that lead aluminum alloys care too much if you get a little bit of soda can mixed in, but I sure care if you put lead alloy in my window screens or into my aluminum can (although the lining protects a lot of the can, it doesn't protect the lip or your sweaty hands, or the PR nightmare of discovering lead in the cans).
Yep, this is why I only recycle aluminum, steel, and glass. Not even sure the glass is worth it. Everything else goes in the trash (saves the recycling facility the bother of doing that instead).
Does sorting done by the general public count as "free"? It takes time and effort to route garbage to the appropriate bin --- maybe not much, but it adds up over the whole population. The desire to avoid this effort doesn't strike me as "lazy" any more than hiring a housekeeper or landscaper is "lazy". Economies of scale and division of labor make our society more prosperous. Why shouldn't we apply these principles to trash disposal and recycling?
From an efficiency perspective, it seems "cheaper" to presort each item upon ingestion rather than mixing everything up and sorting later. For most people, placing an item in the right bin is a subconscious action on "autopilot".
It does not scale (at least not yet with current AI). It does cost everyone a bit of time but it also depends on you and how much trash you generate. Recycling is for the benefit of all of us and we should all do our part irregardless if a CEO or a house cleaner.
The consumer sorting is often not very good and requires multiple bins with a higher capital and operating cost for storage and collection. It's a pain for the users. Single stream trash is the way to go with robotic sorting is the way to go. No reason it can't scale. Amazing work, one of the best uses of labeling and machine vision I've seen to date.
I have experience in commercial recycling and was about to post something with dim details, but this comment says what I was about to say roughly. It is hard to read comments from people with good intentions but zero experience in commercial recycling. Single-stream versus household sort is one of the most basic choices of any municipal recycling program. Every operation has waste, flaws and mistakes, single stream or not.
Excellent human sorting as you find in Japan or Switzerland is rare. Of course it is desirable. Pre-sorted inputs are not perfect in those jurisdictions, and rare in most other places.
The comment above lauds the technology here and I agree. Sorting a recycling line is a terrible kind of job for the majority of adults, and economics make the implementation worse in many cases. De-facto, it is the handicapped, children, prisoners and starving people that sort trash in most of the world. 1st world machinery with the right economics are an improvement in almost all cases, especially where food waste might be involved.
I'm wondering what happens to the woman who collects bottles in my neighborhood for her son's wheelchair when this technology comes to the third world.
That said the local garbage dump is encroaching on a refugee settlement, if it was something that could help it encroach less on that maybe it's a good thing.
It's not necessarily a pain for the users. It's a cultural thing. If you can change the culture, it's easy to do it with humans at the source. I'm in South Korea now and they sort all the garbage before even throwing it out. Different bins for different types of plastic, styrofoam, tetra pak containers, cardboard, glass, organic waste, etc. My brother-in-law, his kids, and I will take the garbage down and put it all into the right bin. No complaining, everyone thinks it's the proper thing to do. Makes it super easy to process each type of garbage correctly. https://en.wikipedia.org/wiki/Recycling_in_South_Korea
South Korea didn't do it this way in the beginning. When I was in China a couple of years ago, they were just starting to process their garbage this way, though it'll take a while for them to really do it well. The interesting question is why it's so hard to get the west to handle things this way. I don't know Europe, they might be better, but I know North America, and wow, the people in North America really don't care about this kind of stuff in general.
Again, I'm not sure of that. They feel that it's the proper thing to do, so I'm not sure they view it as painful. I hate to say it, but it feels like a disparity of maturity. Some kids do their chores because they're responsible kids, take their responsibilities seriously, are proud of being responsible, and view kids who aren't similarly responsible as immature. Other kids hate their chores and do their chores in spite of their hate. Both kids are facing obligations, but their emotional reactions are different.
It's nothing to do with maturity, don't be patronising. They feel that it's the proper thing to do because it is the proper thing to do if you don't have good centralised sorting.
If you do then centralised sorting is clearly better all round. It's less work for people, collection is simplified and more efficient, you don't have issues with people missorting, you can sort into more categories, etc. etc.
So nobody in the world has good centralised sorting. That's our reality. So what does that imply then as to people not wanting to deal with the pain of presorting? You say that it's the proper thing to do in that reality. I'm welcome to using a word other than maturity to describe the phenomenon if you can suggest one. I'm not married to that word.
> So nobody in the world has good centralised sorting. That's our reality.
Yes. That's what people are trying to solve.
But that's irrelevant. Your original comment was claiming that manual sorting isn't a pain because it's the proper thing to do. That's just nonsense. It's the proper thing to do (for now) but it's also a pain.
Would you have said washing clothes wasn't a pain in the 1800s? It's basically the same thing. Something annoying that you have to do because the technology to do it automatically hasn't been invented yet.
I think this is overly getting into semantics now. But if we must, when I was a child, cleaning the floor was a pain. As an adult, not cleaning the floor creates the pain, having a clean floor is great, and the act of cleaning itself is just a natural thing to do, just as breathing and sleeping is. Natural acts aren't painful. I think a border gets crossed for those who can make something natural. For people just getting started exercising, exercise is painful. For people who exercise regularly, exercise is enjoyable. In fact, I recall one Olympian Simon Whitfield saying that his body would experience pain if he didn't exercise.
Perhaps different people also have different thresholds for pain. I'm not sure people in the 1800s would have thought, "Gee, washing these clothes is such a pain, I wish something could be invented to do it for me." I think they rather would have thought this is just a natural part of life. Perhaps the super rich who could afford servants could have the luxury to afford such thoughts. I'm betting many people in the 1800s wouldn't have even been in that headspace.
Again, you said that presorting is the proper thing to do in a reality without centralized sorting. At this moment, I can't really think of anything that I'd consider both "proper" and "painful" if I have managed to seriously make my mindset consider that thing to be "proper". Before making the switch, I would consider it "painful", yes. Looking at the before and after, I can only see maturity being the dimension that really changed. Maybe commitment is another word that could be used too. But if I really think about it, I think commitment would be so heavily moderated by maturity to the point where it might as well override commitment in terms of effect. Again, I'm open to alternative words if any can be suggested.
I used Airbnb for years traveling to Zürich, and there were always very explicit directions about the recycling with warnings that any fines incurred would be passed along to me. I think the whole system seems fairly simple and reasonable.
I am not the host, it isn't working in my apartment block.
The UK doesn't have a single standard on how to do recycling, it is up to each town council to choose how things should be sorted as well as the type and colour of the bin for each class, someone from outside that area would not know the system without being told.
I’ve been automating processes and systems (with software, not robotics) since my early days as a sysadmin. I would later mentor many juniors over the years and was often challenged by them about how good for our careers doing such a thing was. I would always respond that I felt our jobs were to try to automate ourselves out of being necessary, because then we could move on to bigger more challenging problems, now that we had gotten common repeatable issues out of the way. “There will always be bigger, better work for those with the skills required to automate their jobs out of existence”, I would tell them.
Now 25 years later I find myself saying similar things when discussing the topic with people who fear automation displacing workers, and my initial response to this article was much the same.
Then, a memory of a commercial I used to see as a kid popped into my head and for the first and only time I’ve felt myself being pulled to an anti-automation sentiment - and this is an isolated use case of automation I am feeling this way about, and it is all because of this commercial that apparently still occupies a place in my mind probably 40 years since it aired.
I had a ‘special’ mailman at NASA Ames in 2007-2009. This guy was the best mailman ever; he was 100% focused on his job and did it with the seriousness of handling CBRN materials. You might wonder how someone can go from good to great at mail handling, especially to someone that doesn’t send or receive much mail. There were lots of unexpected things. Address errors always got to me. Mail was sorted by ‘importance’ in the box. Hand delivery of government mail and warnings about anything anomalous. I got the feeling that you could give this guy any small job and he would study the minutiae intensely, finding ways to get better long after any ‘normal’ person would have gotten bored and started thinking about something else. I see a lot of problems when people don’t employ them well. They often live institutionalized and do odd jobs part-time like grocery bagging. At Stop’n’Shop, they are switched around and perform poorly. At the Westford Market Basket, there was a special bagger that had a rough first few years, but learned everything about the store, and ended up managing it, superlatively imo. 100% focus on whatever they are told to do for years at a time. My mailman would have made a great project manager, a set of well-defined human tasks that are almost always ignored and subverted because of the power that comes with the position. Imagine a person in a power role that was actually focused on doing their job right.
In my home-town there is a garage/factory which mainly recruits all kinds of disabled people. They produce e.g. wooden chairs, cup holders and the-like but also purely artistic decorative stuff. All on very different levels i.e. some people nail some wooden sticks together as they have been told, some have complex jobs (if they are able).
For sure, all of the tasks they do during the day can be automated and in-fact are automated at the big facilities. However, people like to buy their products - not only because they like to support them and to give them the chance to contribute to the community (like it was said in the video: they pay taxes and everything) but also because there is a general market for hand-made stuff.
There is always a small niche of consumers who prefer hand-made stuff because of its individual charm etc. over the soul-less mass-manufactured alternative. I believe that this demand exists precisely because of the rise of automation (and thus is unlikely to vanish is automation is pushed furhter).
I'd also argue that working in a woodworking shop - being able to actually create something and (if the handicap/IQ allows for it) even be creative - has a much better effect on overall quality of live than working on a assembly-line bullshit job.
I don't know of a handicap which does not allow you to work in any of these kind of jobs but does allow you to assort plastic from paper within reasonable amount of time (but I'm sure someone can give me an example; in that case I'd argue that its up to us to find or "invent" a suitable job or some helper-device to enable them to do so - we have the money to do this).
So yes, maybe getting rid of that particular job is the best thing that could happen to the guy in the video - provided there is another job-alternative available that does not only let him add value to the consumer-society but also to the intellectual and creative parts of it as well as of himself (relative to his level of disability).
Yeah these are great insights in the macro-scale, but on the individual level automation can screw your life completely. Specially if you are older.
What you as a 50 y.o. trucker are going to do when Waymo takes your job instead? There is no way you can get the same 6 figure job somewhere outside the truck-driving business.
Because eventually all the tasks that a normal human can do can be done cheaper by a machine. Whenever more automation enabled economic growth in the past, a (large) part of the new economic activity had to be done by humans. We were lucky that the growth was big enough that more new jobs were created than old jobs were lost. In the near future the fraction of new tasks that can be done by humans will shrink. It is plausible that economic growth as a whole will shrink too.
And of course there are the short term effects to consider too. If truck driving is automated, a fifty year old truck driver is unlikely to find a similar job. Changing careers gets progressively harder as you grow older. Many people will be unemployed for the rest of their lives despite other new jobs being created.
I don’t disagree at all. The point is I never really stopped to consider those folks who really will have a hard time moving up to better things, or even laterally. The video I linked, a commercial about a human doing the very job the AI/robotics project aims to do so instead.
It’s short. I hope all commenters take 30 seconds to watch it, and maybe within the discussion in this thread, someone will make a salient point that explains away the nuance I referred to for this exact specific instance.
It just gave me pause, and opened my eyes to that the issue - and my beliefs - can be more nuanced, in light of such things as the PSA from the 80’s I linked in my original post. What’s that guy supposed to do? He Eve said he learned it slowly but now he’s good at it. When all the sun-menial tasks are automated, what becomes of them?
For the record I am a cynical technophile by any measure. I have been called callous (and much worse) for statements like “starvation is a self-solving problem. Eventually there will be enough food for all, as the population decreases”… don’t think me anti-automation; this edge case is a thing that makes me go “hmmm…”
I agree that it's generally been true up until now that we've invented new jobs to replace those which we've automated away, but I don't think that is the whole story. Going back to at least the industrial revolution with for example handloom weavers, people who have had their jobs automated away were often impoverished for the rest of their lives. I don't have any statistics to back this up, but I'd argue that this will be true for most displaced workers today -- any career change later in life will put you behind the curve, and it's not like someone working as a coal miner is likely to learn how to code or whatever. My thought is that these new high-tech jobs we invent have been and will be performed by new generations leaving the displaced to suffer.
I also don't think this trend of inventing new jobs will hold true forever. To the poster who asked why this would suddenly change, it's because we now have algorithms which can learn. We may be long dead (though perhaps not), but I am absolutely convinced that eventually we will invent machines which are able to perform any given task at or exceeding normal human ability. When that happens, and when we deploy it in mass across the economy, I think we'll have to face an existential crisis as a species.
Today we want everyone to have a job because we want everyone to have a means of independently supporting themselves. We want each of us to contribute to society, to not be a burden. We want people to have a job because it often provides a sense of dignity and a sense of purpose.
My thought is that, just as you say that normal people will find another job as automation advances, the attitude we will need to adopt instead will be that people in general disabled or not will need to find another means of providing meaning in their life. Because honestly, my thought when watching that commercial was not, "oh how nice that this individual is working." My thought was that the job looked terrible, is probably not good for health, and that it's not a job that human beings should even be doing. I generally think this about most jobs, but especially so about jobs where you're sorting through waste.
Real people will suffer from automation, but I think the process is inevitable, and personally wished I existed in a time where automation freed all of us to pursue what we were interested in rather than serving the purposes of making the economic machine spin round. Let it spin on its own.
The wealth produced by those robotic replacements will not go to the replaced humans, or it will go in small pieces as social services, doubling down on the frustration. This sounds volatile to me.
> For the record I am a cynical technophile by any measure. I have been called callous (and much worse) for statements like “starvation is a self-solving problem. Eventually there will be enough food for all, as the population decreases”… don’t think me anti-automation; this edge case is a thing that makes me go “hmmm…”
Here is the thing, this isn't as linear as you make it out to be and is seemingly not taking into account very obvious escalation of things as Society fails to deliver and fail it's populace: protests, strikes, peasant revolts, and eventually total war. You cannot exclude the lower rungs of Society and expect that because of 'progress' things will sort themselves out.
Just look at the wide-spread amount misery we've seen since 2008, people creating tent cities and the mass homelessness has led to a Humanitarian crisis, and an epidemic itself.
With that said, recycle sorting is a MVP for AI/ML applications: it's a low paid, and hard job that requires a perpetual stream of exploitable underclass based labour: even if someone would want to do the job short term out of desperation, it isn't one that will be held long term so the churn, specifically turnover, is probably not worth the cost in retraining Human based worksforce.
Matanya Horowitz a grad from CU Boulder, and eventually PhD grad from CalTech, was working on this [0] when I used to work out of the engineer department, he eventually went on to start an AI based robotics company (AMP) and raised 75 million since it's inception.
I'm also of the POV, after having done Ag myself and seeing the complexity of supply chains and the inherent pitfalls and obstacles it relies on, that farming must go at least 90% automated: primarily it is based on the fact that ~99% of the population relies on the other 1% to feed itself since that is Globally the amount of people involved in Ag--and the average age of farmers Globally is well into late 50s an early 60s. This is an imperative issue that requires more talent focus on automated processes in the laborious and most physically straining aspects that require often short-term and poorly compensated migrant labour--it's essentially slave labour 'with more steps.'
Instead of creating scripts for IT departments you should perhaps you should stop focusing your energy on such low-hanging fruit if you really believe what you wrote and focus on the major existential issues before you start wielding your technophile iron fist and talking about social engineering on this scale. Moreover a month's work of seemingly meaningless manual labour would also give you some much needed clarity and perspective.
With that said, I went from fintech founder to working for a megacorp and quit because I felt my skills were being wasted and did my last stint in kitchens right before COVID when the 'burger flippers, and food delivery people are obsolete and will be automated' rhetoric was at ATH. After COVID (which is still a thing) it's pretty clear now that reality hit those people incredibly hard, and salaries in those Industries have finally caught up with the demand from the Market as more people refused to come back and the 'automate everything' crowds are licking their wounds with mediocre results.
I went back to school to study AI and ML because I want to focus on automating Ag, and even with the leap it has amongst most other Industries, there are still a myriad of issues to work on in regards to Computer Vision and engineering feats that are optimized for crop-specific harvests and weeding.
We're getting a taste of what broken supply chains and surging costs is having to Society, and the truth is as bad as it is still (barely) working; and when food stops being delivered is when Society starts to undo itself incredibly fast.
One point I would like to respond to is that the only specific instance I mentioned where I said we should be “automating ourselves out of” jobs, was in the context of those systems/network/devops roles; roles I myself have held.(Incidentally, automation at the enterprise level involves far more than “writing scripts for IT departments” that you reduce it to, and that is somewhat what I was referring to when. I mentioned that there would be bigger and more important business processes to support as infrastructure becomes commodity, and some automation framework(s) become SOP. Sorry, I digress.)
I did not suggest anywhere - at least I didn’t mean to - to wield a techno “iron fist” (although that does conjure a pretty cool image in my head <grin>) and I don’t think I suggested that every job everywhere should be automated, and in fact the memory of an 80’s PSA coming to mind giving me an immediate feeling of sympathy for that guy in the video, was really the whole point of why I posted.
Also, I wasn’t born into high-paying tech jobs. I was once young, and indeed have worked some pretty shit jobs. I don’t however, think they were meaningless. They may have meant nothing to me, but obviously someone thought they were worth paying me for. So aside from “dig a hole and refill it” type work, I don’t view anyone’s job as not having worth.
It does beg the question, “what will the shit jobs be for future teenagers?”, which I will for the sake of simplification lump into the same category as the glass-sorting job in the PSA. I don’t have an answer for that, at least not yet.
Perhaps this is tangential to the existential crises you describe, but, like you, I have been learning ML/AI, since about 2016 (whenever that “A Neural Algorithm of Artistic Style” paper was published) and I do have some fears about AI, superintelligence, etc and what bad things will (not just may) come of it, though admittedly my fears of the negative are far outweighed by my excitement for the positive potential. This is why I own paperclipmaximizer.ai though admittedly my tongue was firmly in my cheek when I bought it. We can only hope our super intelligent AGI overlords are as optimistic as the GPT-generated quote from Maximo on that site.
I appreciate everyone’s input and time taken to provide it. I have been reading HN for a long time, but never registered or posted until last week. It is really great to have a community of intelligent people not trolling or shitposting and I was pleasantly surprised that you all actually watched the video and came back with real points of conversation. Thanks!
"No model, regardless of size, outperformed the best-performing human on any task. However, for some tasks, the best-performing model beat the average human."
Charge product manufacturers enough tax to get recycling of their packaging to 95%.
Aluminum and clear glass would immediately become more popular (they are mechanically sortable). I'm guessing the manufacturers would quickly standardize on a small number of plastics for everything else.
There is no way to fully price in externalities. This problem is similar to measuring the length of a shoreline. The closer you look, the more externalities you’ll find.
A friend of mine worked at this company, and the bane of his existence was one popular Arizona green tea bottle. The plastic and glass versions looked identical[1], so no way to sort it visually. Great for brand identity, bad downstream.
Did your friend eventually find a workable solution? I wonder if they could reach out to the beverage company and ask for a distinguishing feature that’s easy to recognize for computer vision but wouldn’t affect the brand’s look-and-feel for human vision. Perhaps a machine readable version of a resin identification code [1] encoded as a pattern visible from any angle?
Seems like we need similar legislation as tobacco packaging. Dedicate 40-60% of product to machine readable identification of what type of recycling it is.
Or just go Aldi's route and put enormous barcodes covering entire sides of the package. Scan rate goes up _immensely_ when you don't have to give even half a hoot about orientation.
Before we tried to use AI and cameras for this task, bottle-return machines have just used barcode scanners. They're fast, deadly accurate, and take zero intelligence to program. The achilles' heel is certain jerkface brands that use white-ink-on-shiny-aluminum, which violates the contrast requirements of many scanners. For those, image recognition would be a valuable second method.
Aldi seems to have figured out packaging the best out of ANY store I have been to. And the empty boxes for your groceries are always useful if you forget to grab a basket beforehand.
I think the normal thing is to drop trash down a tower with a forced air cross-current. Glass gets diverted less than plastic by the cross-current.
You can also use various densities of liquid, to catch different densities of plastic floating vs. sinking if you've already crushed things up to avoid air bubbles. However, then you need to later separate the liquid from the shredded garbage.
Seriously. Give that robot some more senses. Echoic, magnetic, infra-ultra, spectrographic analytic, nuclear magnetic resonant... the whole schmeer. It's certainly easier than making them smarter.
I suppose a law could be passed that versions of a product with different materials would need to be visually distinct? Or manufacturers could voluntarily do it.
I wish articles like this would just get to the point. We don't need yet another explanation of how recycling works, how it has low recovery rates, etc. Given the audience, take it as a given the reader has the basics down.
Further down the author finally starts to talk about their topic (trash picking robots), but then digresses yet again. It's a really terrible writing style.
State what your topic is up front with a brief high level, then you can mix in some more background while expanding on the details further below.
If you want to be taken seriously you need to pad everything out to 40 paragraphs. It's some kind of unwritten rule of the internet. Literally. They even have bots that cull succinctity or label it trolly. It's psychotic.
I remember in high school reading a Tale of Two cities. Absolutely disliked how all the language is over dramatic and flowery. Try reading some scientific papers from the same era. Many of them are similar in how they were written.
Then sometime after I read something by Mark Twain for the first time. I was blown away at how much easier and more enjoyable his writing style was.
GOOD Point by my wife! Why can’t we have a blower on a belt to accomplish most of the sorting? Plastics blow one way and glass stays on the belt. Throw in a magnet and make eddy currents to trap metals.
They do use eddy currents for metal sorting. Take some of your home recycling to a recycler instead of leaving it on the bin. It's worth going once just to see the small scale solution. Get yourself a bag of soda cans and throw a couple soup cans in there. Not the cleanest place you'll ever be, so don't wear your nice shoes and prep for some odors, but better than some.
When I was a child I saw one system that expected all aluminum that just used an electromagnet to lift any steel out of the stream. That was quite a ways up in the air so I suspect the owner would have gotten pretty pissed if we threw a bunch of nails or soup cans in the bag for fun. But most of the other ones I've seen have two conveyors with a gap, and they use an AC magnet to fling the aluminum onto the second conveyor (via eddy currents). Most of the ferrous stuff just drops into the first bin, or funnels off to be sorted again.
For your plastic or glass suggestion, that may take breaking it up first, since large items might not lift or go in weird directions. And then some plastics are pretty dense. I have the vaguest notion that I might have seen a system that sorts plastic from glass while washing it, since the glass sinks and some plastics don't, but that could have been a prototype or a misfiring neuron for all I can remember.
I think it is harder than just blowing them with air… different plastics have to be recycled in different ways. The number in the recycle arrows tells you what it is and how (or if) you can recycle it (see https://www.thoughtco.com/recycling-different-types-of-plast...). I don’t believe these plastics can be sorted by density or level of wind resistance.
Why? Make the investment at centralized locations. No human should have to do a mind numbing job like that. Can’t even relax, it will be stressful too.
ML was not the focus of the company, we didn't really have any qualified devs to pull off a project like that. It became a prototype for a sales pitch and I don't know what happened to it after. Though I disagree about it not being a complicated/difficult endeavour.
This is a solution without a problem? The global recycling industry has collapsed. It never broke 10% btw, and is currently at 4%. China doesn’t want our trash anymore. No one is buying, regardless of marginal cost.
Our world will so quickly come to resemble the fantasies of Star Wars etc. With multi-armed intelligent robots sorting recyclables, perhaps in the bowels of Jawa scavenger ships...
What I find sad is to see startups here in Switzerland where we already separate and recycle extensively that offer a service where you just toss everything that can be recycled in the same bag. So batteries are mixed with plastic and cans. These bags compete with government collection facilities and are "reprogramming" people to be lazy. Regular trash is very expensive here and people who want to safe are forced to separate and recycle what can be.
The hardest task of recycling is the separation of items. Centralizing this does not scale as you need to hire more and more underpaid worked to do this task. The task that was done for free before by the person throwing the item out. All the facilities had to do is sort the remaining small percentage of miss sorted items instead of all items.
Centralizing this does not scale.
Sure it does.[1] What doesn't scale is manual sorting. It complicates collection, because there are all those bins to empty,the trucks are more complicated, and humans aren't good at separating the different types of plastics. That's easy with multispectral imaging, but humans lack that capability.
Bulk Handling Systems seems to be the US leader in this. Robots are used but do under 10% of the job. Most of the sorting is done by screens, shakers, drums, DC and AC magnets, and air-jet optical sorters. The robots only do "quality control", pulling out stuff that the heavy machinery mis-sorted.
San Francisco has one of their plants.
Aluminum is easily separated and recycled, and is the only profitable part of the operation. Steel gets separated and melted down, and about 40% of US steel is now recycled. Which is why Nucor, which started as a recycler, is the biggest US steel company. Some kinds of plastic bottles recycle well, but it doesn't really pay. CarbonLite, with three big plants, went bankrupt in 2021. The buyers kept the plants running, but the investors didn't get much. CarbonLite was caught between high fixed costs and not enough market for their product.
There's not much use for mixed paper. Offices used to generate large amounts of good quality paper, and newspapers were a consistent source of uniform product. But now, most paper is discarded packaging, with too much else mixed in.
It's hard to make money at this. Separation, though, is not the problem.
[1] https://bulkhandlingsystems.com/wp/
I saw one of those how they do it shows where the host went to an aluminum recycling plant, and this is how I learned there are lead alloys out there. They use an xray machine to sort those.
My favorite thing about some of these sorting systems is that they are online but not instantaneous. For any reasonable rate of objects passing by the scanner, they can determine yes/no (or maybe/no) in a fixed amount of time, say 10ms for illustration purposes. So the picker is 10ms down the conveyor from the sensor, and the signal arrives at the precise moment to activate the picker. More computers or faster computers let you increase the flow rate by increasing either the density or the speed of the conveyor, but it's always a realtime system.
For instance I don't think that lead aluminum alloys care too much if you get a little bit of soda can mixed in, but I sure care if you put lead alloy in my window screens or into my aluminum can (although the lining protects a lot of the can, it doesn't protect the lip or your sweaty hands, or the PR nightmare of discovering lead in the cans).
The cost of recycling needs to be built into the cost of the product, for example via a portion of the sales tax.
This is likely the only effective solution!
It would revolutionize the packaging industry and many others, making everything easy to automatically disassemble and recycle.
Yep, this is why I only recycle aluminum, steel, and glass. Not even sure the glass is worth it. Everything else goes in the trash (saves the recycling facility the bother of doing that instead).
Does sorting done by the general public count as "free"? It takes time and effort to route garbage to the appropriate bin --- maybe not much, but it adds up over the whole population. The desire to avoid this effort doesn't strike me as "lazy" any more than hiring a housekeeper or landscaper is "lazy". Economies of scale and division of labor make our society more prosperous. Why shouldn't we apply these principles to trash disposal and recycling?
From an efficiency perspective, it seems "cheaper" to presort each item upon ingestion rather than mixing everything up and sorting later. For most people, placing an item in the right bin is a subconscious action on "autopilot".
It does not scale (at least not yet with current AI). It does cost everyone a bit of time but it also depends on you and how much trash you generate. Recycling is for the benefit of all of us and we should all do our part irregardless if a CEO or a house cleaner.
It's not noticeably more difficult to drop an item in 1 of 4 bins vs 1 - and it reduces spoilage.
mixing items causes contamination, that is rgen harder to separate.
The consumer sorting is often not very good and requires multiple bins with a higher capital and operating cost for storage and collection. It's a pain for the users. Single stream trash is the way to go with robotic sorting is the way to go. No reason it can't scale. Amazing work, one of the best uses of labeling and machine vision I've seen to date.
I have experience in commercial recycling and was about to post something with dim details, but this comment says what I was about to say roughly. It is hard to read comments from people with good intentions but zero experience in commercial recycling. Single-stream versus household sort is one of the most basic choices of any municipal recycling program. Every operation has waste, flaws and mistakes, single stream or not.
Excellent human sorting as you find in Japan or Switzerland is rare. Of course it is desirable. Pre-sorted inputs are not perfect in those jurisdictions, and rare in most other places.
The comment above lauds the technology here and I agree. Sorting a recycling line is a terrible kind of job for the majority of adults, and economics make the implementation worse in many cases. De-facto, it is the handicapped, children, prisoners and starving people that sort trash in most of the world. 1st world machinery with the right economics are an improvement in almost all cases, especially where food waste might be involved.
I'm wondering what happens to the woman who collects bottles in my neighborhood for her son's wheelchair when this technology comes to the third world.
That said the local garbage dump is encroaching on a refugee settlement, if it was something that could help it encroach less on that maybe it's a good thing.
Lots of factors here.
It's not necessarily a pain for the users. It's a cultural thing. If you can change the culture, it's easy to do it with humans at the source. I'm in South Korea now and they sort all the garbage before even throwing it out. Different bins for different types of plastic, styrofoam, tetra pak containers, cardboard, glass, organic waste, etc. My brother-in-law, his kids, and I will take the garbage down and put it all into the right bin. No complaining, everyone thinks it's the proper thing to do. Makes it super easy to process each type of garbage correctly. https://en.wikipedia.org/wiki/Recycling_in_South_Korea
South Korea didn't do it this way in the beginning. When I was in China a couple of years ago, they were just starting to process their garbage this way, though it'll take a while for them to really do it well. The interesting question is why it's so hard to get the west to handle things this way. I don't know Europe, they might be better, but I know North America, and wow, the people in North America really don't care about this kind of stuff in general.
Of course it's a pain. Just because they're used to it doesn't mean they like it. Having that many different bins is obviously inconvenient.
Again, I'm not sure of that. They feel that it's the proper thing to do, so I'm not sure they view it as painful. I hate to say it, but it feels like a disparity of maturity. Some kids do their chores because they're responsible kids, take their responsibilities seriously, are proud of being responsible, and view kids who aren't similarly responsible as immature. Other kids hate their chores and do their chores in spite of their hate. Both kids are facing obligations, but their emotional reactions are different.
It's nothing to do with maturity, don't be patronising. They feel that it's the proper thing to do because it is the proper thing to do if you don't have good centralised sorting.
If you do then centralised sorting is clearly better all round. It's less work for people, collection is simplified and more efficient, you don't have issues with people missorting, you can sort into more categories, etc. etc.
So nobody in the world has good centralised sorting. That's our reality. So what does that imply then as to people not wanting to deal with the pain of presorting? You say that it's the proper thing to do in that reality. I'm welcome to using a word other than maturity to describe the phenomenon if you can suggest one. I'm not married to that word.
> So nobody in the world has good centralised sorting. That's our reality.
Yes. That's what people are trying to solve.
But that's irrelevant. Your original comment was claiming that manual sorting isn't a pain because it's the proper thing to do. That's just nonsense. It's the proper thing to do (for now) but it's also a pain.
Would you have said washing clothes wasn't a pain in the 1800s? It's basically the same thing. Something annoying that you have to do because the technology to do it automatically hasn't been invented yet.
I think this is overly getting into semantics now. But if we must, when I was a child, cleaning the floor was a pain. As an adult, not cleaning the floor creates the pain, having a clean floor is great, and the act of cleaning itself is just a natural thing to do, just as breathing and sleeping is. Natural acts aren't painful. I think a border gets crossed for those who can make something natural. For people just getting started exercising, exercise is painful. For people who exercise regularly, exercise is enjoyable. In fact, I recall one Olympian Simon Whitfield saying that his body would experience pain if he didn't exercise.
Perhaps different people also have different thresholds for pain. I'm not sure people in the 1800s would have thought, "Gee, washing these clothes is such a pain, I wish something could be invented to do it for me." I think they rather would have thought this is just a natural part of life. Perhaps the super rich who could afford servants could have the luxury to afford such thoughts. I'm betting many people in the 1800s wouldn't have even been in that headspace.
Again, you said that presorting is the proper thing to do in a reality without centralized sorting. At this moment, I can't really think of anything that I'd consider both "proper" and "painful" if I have managed to seriously make my mindset consider that thing to be "proper". Before making the switch, I would consider it "painful", yes. Looking at the before and after, I can only see maturity being the dimension that really changed. Maybe commitment is another word that could be used too. But if I really think about it, I think commitment would be so heavily moderated by maturity to the point where it might as well override commitment in terms of effect. Again, I'm open to alternative words if any can be suggested.
Airbnb guests mean that separating recyclable items at the home doesn't work, you just end up with some incorrect stuff in every recycling bin.
I used Airbnb for years traveling to Zürich, and there were always very explicit directions about the recycling with warnings that any fines incurred would be passed along to me. I think the whole system seems fairly simple and reasonable.
I am not the host, it isn't working in my apartment block.
The UK doesn't have a single standard on how to do recycling, it is up to each town council to choose how things should be sorted as well as the type and colour of the bin for each class, someone from outside that area would not know the system without being told.
I’ve been automating processes and systems (with software, not robotics) since my early days as a sysadmin. I would later mentor many juniors over the years and was often challenged by them about how good for our careers doing such a thing was. I would always respond that I felt our jobs were to try to automate ourselves out of being necessary, because then we could move on to bigger more challenging problems, now that we had gotten common repeatable issues out of the way. “There will always be bigger, better work for those with the skills required to automate their jobs out of existence”, I would tell them.
Now 25 years later I find myself saying similar things when discussing the topic with people who fear automation displacing workers, and my initial response to this article was much the same.
Then, a memory of a commercial I used to see as a kid popped into my head and for the first and only time I’ve felt myself being pulled to an anti-automation sentiment - and this is an isolated use case of automation I am feeling this way about, and it is all because of this commercial that apparently still occupies a place in my mind probably 40 years since it aired.
https://youtu.be/CSiz6kbIZkw
I had a ‘special’ mailman at NASA Ames in 2007-2009. This guy was the best mailman ever; he was 100% focused on his job and did it with the seriousness of handling CBRN materials. You might wonder how someone can go from good to great at mail handling, especially to someone that doesn’t send or receive much mail. There were lots of unexpected things. Address errors always got to me. Mail was sorted by ‘importance’ in the box. Hand delivery of government mail and warnings about anything anomalous. I got the feeling that you could give this guy any small job and he would study the minutiae intensely, finding ways to get better long after any ‘normal’ person would have gotten bored and started thinking about something else. I see a lot of problems when people don’t employ them well. They often live institutionalized and do odd jobs part-time like grocery bagging. At Stop’n’Shop, they are switched around and perform poorly. At the Westford Market Basket, there was a special bagger that had a rough first few years, but learned everything about the store, and ended up managing it, superlatively imo. 100% focus on whatever they are told to do for years at a time. My mailman would have made a great project manager, a set of well-defined human tasks that are almost always ignored and subverted because of the power that comes with the position. Imagine a person in a power role that was actually focused on doing their job right.
In my home-town there is a garage/factory which mainly recruits all kinds of disabled people. They produce e.g. wooden chairs, cup holders and the-like but also purely artistic decorative stuff. All on very different levels i.e. some people nail some wooden sticks together as they have been told, some have complex jobs (if they are able).
For sure, all of the tasks they do during the day can be automated and in-fact are automated at the big facilities. However, people like to buy their products - not only because they like to support them and to give them the chance to contribute to the community (like it was said in the video: they pay taxes and everything) but also because there is a general market for hand-made stuff. There is always a small niche of consumers who prefer hand-made stuff because of its individual charm etc. over the soul-less mass-manufactured alternative. I believe that this demand exists precisely because of the rise of automation (and thus is unlikely to vanish is automation is pushed furhter).
I'd also argue that working in a woodworking shop - being able to actually create something and (if the handicap/IQ allows for it) even be creative - has a much better effect on overall quality of live than working on a assembly-line bullshit job. I don't know of a handicap which does not allow you to work in any of these kind of jobs but does allow you to assort plastic from paper within reasonable amount of time (but I'm sure someone can give me an example; in that case I'd argue that its up to us to find or "invent" a suitable job or some helper-device to enable them to do so - we have the money to do this).
So yes, maybe getting rid of that particular job is the best thing that could happen to the guy in the video - provided there is another job-alternative available that does not only let him add value to the consumer-society but also to the intellectual and creative parts of it as well as of himself (relative to his level of disability).
Yeah these are great insights in the macro-scale, but on the individual level automation can screw your life completely. Specially if you are older.
What you as a 50 y.o. trucker are going to do when Waymo takes your job instead? There is no way you can get the same 6 figure job somewhere outside the truck-driving business.
It's only a matter of time until automation replaces humans with normal IQ in a way that they're permanently out of jobs.
So far automation has only created more jobs. Why would this suddenly change?
Because eventually all the tasks that a normal human can do can be done cheaper by a machine. Whenever more automation enabled economic growth in the past, a (large) part of the new economic activity had to be done by humans. We were lucky that the growth was big enough that more new jobs were created than old jobs were lost. In the near future the fraction of new tasks that can be done by humans will shrink. It is plausible that economic growth as a whole will shrink too.
And of course there are the short term effects to consider too. If truck driving is automated, a fifty year old truck driver is unlikely to find a similar job. Changing careers gets progressively harder as you grow older. Many people will be unemployed for the rest of their lives despite other new jobs being created.
I don’t disagree at all. The point is I never really stopped to consider those folks who really will have a hard time moving up to better things, or even laterally. The video I linked, a commercial about a human doing the very job the AI/robotics project aims to do so instead.
It’s short. I hope all commenters take 30 seconds to watch it, and maybe within the discussion in this thread, someone will make a salient point that explains away the nuance I referred to for this exact specific instance.
It just gave me pause, and opened my eyes to that the issue - and my beliefs - can be more nuanced, in light of such things as the PSA from the 80’s I linked in my original post. What’s that guy supposed to do? He Eve said he learned it slowly but now he’s good at it. When all the sun-menial tasks are automated, what becomes of them?
For the record I am a cynical technophile by any measure. I have been called callous (and much worse) for statements like “starvation is a self-solving problem. Eventually there will be enough food for all, as the population decreases”… don’t think me anti-automation; this edge case is a thing that makes me go “hmmm…”
I agree that it's generally been true up until now that we've invented new jobs to replace those which we've automated away, but I don't think that is the whole story. Going back to at least the industrial revolution with for example handloom weavers, people who have had their jobs automated away were often impoverished for the rest of their lives. I don't have any statistics to back this up, but I'd argue that this will be true for most displaced workers today -- any career change later in life will put you behind the curve, and it's not like someone working as a coal miner is likely to learn how to code or whatever. My thought is that these new high-tech jobs we invent have been and will be performed by new generations leaving the displaced to suffer.
I also don't think this trend of inventing new jobs will hold true forever. To the poster who asked why this would suddenly change, it's because we now have algorithms which can learn. We may be long dead (though perhaps not), but I am absolutely convinced that eventually we will invent machines which are able to perform any given task at or exceeding normal human ability. When that happens, and when we deploy it in mass across the economy, I think we'll have to face an existential crisis as a species.
Today we want everyone to have a job because we want everyone to have a means of independently supporting themselves. We want each of us to contribute to society, to not be a burden. We want people to have a job because it often provides a sense of dignity and a sense of purpose.
My thought is that, just as you say that normal people will find another job as automation advances, the attitude we will need to adopt instead will be that people in general disabled or not will need to find another means of providing meaning in their life. Because honestly, my thought when watching that commercial was not, "oh how nice that this individual is working." My thought was that the job looked terrible, is probably not good for health, and that it's not a job that human beings should even be doing. I generally think this about most jobs, but especially so about jobs where you're sorting through waste.
Real people will suffer from automation, but I think the process is inevitable, and personally wished I existed in a time where automation freed all of us to pursue what we were interested in rather than serving the purposes of making the economic machine spin round. Let it spin on its own.
The wealth produced by those robotic replacements will not go to the replaced humans, or it will go in small pieces as social services, doubling down on the frustration. This sounds volatile to me.
> For the record I am a cynical technophile by any measure. I have been called callous (and much worse) for statements like “starvation is a self-solving problem. Eventually there will be enough food for all, as the population decreases”… don’t think me anti-automation; this edge case is a thing that makes me go “hmmm…”
Here is the thing, this isn't as linear as you make it out to be and is seemingly not taking into account very obvious escalation of things as Society fails to deliver and fail it's populace: protests, strikes, peasant revolts, and eventually total war. You cannot exclude the lower rungs of Society and expect that because of 'progress' things will sort themselves out.
Just look at the wide-spread amount misery we've seen since 2008, people creating tent cities and the mass homelessness has led to a Humanitarian crisis, and an epidemic itself.
With that said, recycle sorting is a MVP for AI/ML applications: it's a low paid, and hard job that requires a perpetual stream of exploitable underclass based labour: even if someone would want to do the job short term out of desperation, it isn't one that will be held long term so the churn, specifically turnover, is probably not worth the cost in retraining Human based worksforce.
Matanya Horowitz a grad from CU Boulder, and eventually PhD grad from CalTech, was working on this [0] when I used to work out of the engineer department, he eventually went on to start an AI based robotics company (AMP) and raised 75 million since it's inception.
I'm also of the POV, after having done Ag myself and seeing the complexity of supply chains and the inherent pitfalls and obstacles it relies on, that farming must go at least 90% automated: primarily it is based on the fact that ~99% of the population relies on the other 1% to feed itself since that is Globally the amount of people involved in Ag--and the average age of farmers Globally is well into late 50s an early 60s. This is an imperative issue that requires more talent focus on automated processes in the laborious and most physically straining aspects that require often short-term and poorly compensated migrant labour--it's essentially slave labour 'with more steps.'
Instead of creating scripts for IT departments you should perhaps you should stop focusing your energy on such low-hanging fruit if you really believe what you wrote and focus on the major existential issues before you start wielding your technophile iron fist and talking about social engineering on this scale. Moreover a month's work of seemingly meaningless manual labour would also give you some much needed clarity and perspective.
With that said, I went from fintech founder to working for a megacorp and quit because I felt my skills were being wasted and did my last stint in kitchens right before COVID when the 'burger flippers, and food delivery people are obsolete and will be automated' rhetoric was at ATH. After COVID (which is still a thing) it's pretty clear now that reality hit those people incredibly hard, and salaries in those Industries have finally caught up with the demand from the Market as more people refused to come back and the 'automate everything' crowds are licking their wounds with mediocre results.
I went back to school to study AI and ML because I want to focus on automating Ag, and even with the leap it has amongst most other Industries, there are still a myriad of issues to work on in regards to Computer Vision and engineering feats that are optimized for crop-specific harvests and weeding.
We're getting a taste of what broken supply chains and surging costs is having to Society, and the truth is as bad as it is still (barely) working; and when food stops being delivered is when Society starts to undo itself incredibly fast.
0: https://businessden.com/2017/11/02/cu-grad-raises-3m-recycli...
1: https://www.colorado.edu/coloradan/2021/07/02/cu-alum-create...
Thanks for the thoughtful reply.
One point I would like to respond to is that the only specific instance I mentioned where I said we should be “automating ourselves out of” jobs, was in the context of those systems/network/devops roles; roles I myself have held.(Incidentally, automation at the enterprise level involves far more than “writing scripts for IT departments” that you reduce it to, and that is somewhat what I was referring to when. I mentioned that there would be bigger and more important business processes to support as infrastructure becomes commodity, and some automation framework(s) become SOP. Sorry, I digress.)
I did not suggest anywhere - at least I didn’t mean to - to wield a techno “iron fist” (although that does conjure a pretty cool image in my head <grin>) and I don’t think I suggested that every job everywhere should be automated, and in fact the memory of an 80’s PSA coming to mind giving me an immediate feeling of sympathy for that guy in the video, was really the whole point of why I posted.
Also, I wasn’t born into high-paying tech jobs. I was once young, and indeed have worked some pretty shit jobs. I don’t however, think they were meaningless. They may have meant nothing to me, but obviously someone thought they were worth paying me for. So aside from “dig a hole and refill it” type work, I don’t view anyone’s job as not having worth.
It does beg the question, “what will the shit jobs be for future teenagers?”, which I will for the sake of simplification lump into the same category as the glass-sorting job in the PSA. I don’t have an answer for that, at least not yet.
Perhaps this is tangential to the existential crises you describe, but, like you, I have been learning ML/AI, since about 2016 (whenever that “A Neural Algorithm of Artistic Style” paper was published) and I do have some fears about AI, superintelligence, etc and what bad things will (not just may) come of it, though admittedly my fears of the negative are far outweighed by my excitement for the positive potential. This is why I own paperclipmaximizer.ai though admittedly my tongue was firmly in my cheek when I bought it. We can only hope our super intelligent AGI overlords are as optimistic as the GPT-generated quote from Maximo on that site.
I appreciate everyone’s input and time taken to provide it. I have been reading HN for a long time, but never registered or posted until last week. It is really great to have a community of intelligent people not trolling or shitposting and I was pleasantly surprised that you all actually watched the video and came back with real points of conversation. Thanks!
"No model, regardless of size, outperformed the best-performing human on any task. However, for some tasks, the best-performing model beat the average human."
https://read.deeplearning.ai/the-batch/issue-150/
To be fair, it's hard to predict the future.
Here's an idea:
Charge product manufacturers enough tax to get recycling of their packaging to 95%.
Aluminum and clear glass would immediately become more popular (they are mechanically sortable). I'm guessing the manufacturers would quickly standardize on a small number of plastics for everything else.
And the price of everything would go up in response. probably not politically feasible in this climate.
If externalities are being priced in, overall costs would go down. They are only cheap now because someone else pays that cost
There is no way to fully price in externalities. This problem is similar to measuring the length of a shoreline. The closer you look, the more externalities you’ll find.
That doesn’t change the fact that you’ll pay more at the till, which is where vulnerable people will feel it most.
A friend of mine worked at this company, and the bane of his existence was one popular Arizona green tea bottle. The plastic and glass versions looked identical[1], so no way to sort it visually. Great for brand identity, bad downstream.
[1] https://www.plasticstoday.com/packaging/arizona-beverages-sw...
Did your friend eventually find a workable solution? I wonder if they could reach out to the beverage company and ask for a distinguishing feature that’s easy to recognize for computer vision but wouldn’t affect the brand’s look-and-feel for human vision. Perhaps a machine readable version of a resin identification code [1] encoded as a pattern visible from any angle?
[1]: https://en.wikipedia.org/wiki/Resin_identification_code
Seems like we need similar legislation as tobacco packaging. Dedicate 40-60% of product to machine readable identification of what type of recycling it is.
Or just go Aldi's route and put enormous barcodes covering entire sides of the package. Scan rate goes up _immensely_ when you don't have to give even half a hoot about orientation.
Before we tried to use AI and cameras for this task, bottle-return machines have just used barcode scanners. They're fast, deadly accurate, and take zero intelligence to program. The achilles' heel is certain jerkface brands that use white-ink-on-shiny-aluminum, which violates the contrast requirements of many scanners. For those, image recognition would be a valuable second method.
Aldi seems to have figured out packaging the best out of ANY store I have been to. And the empty boxes for your groceries are always useful if you forget to grab a basket beforehand.
The article shows the solution to this being a glass breaker stage.
I wonder if echolocation could have done the trick
I think the normal thing is to drop trash down a tower with a forced air cross-current. Glass gets diverted less than plastic by the cross-current.
You can also use various densities of liquid, to catch different densities of plastic floating vs. sinking if you've already crushed things up to avoid air bubbles. However, then you need to later separate the liquid from the shredded garbage.
Seriously. Give that robot some more senses. Echoic, magnetic, infra-ultra, spectrographic analytic, nuclear magnetic resonant... the whole schmeer. It's certainly easier than making them smarter.
I suppose a law could be passed that versions of a product with different materials would need to be visually distinct? Or manufacturers could voluntarily do it.
I wish articles like this would just get to the point. We don't need yet another explanation of how recycling works, how it has low recovery rates, etc. Given the audience, take it as a given the reader has the basics down.
Further down the author finally starts to talk about their topic (trash picking robots), but then digresses yet again. It's a really terrible writing style.
State what your topic is up front with a brief high level, then you can mix in some more background while expanding on the details further below.
Hi, here's my life story.
[14 paragraphs of crap]
Ingredients list.
Recipe.
Aaaaaaaaaaaaaaaaaargh!
If you want to be taken seriously you need to pad everything out to 40 paragraphs. It's some kind of unwritten rule of the internet. Literally. They even have bots that cull succinctity or label it trolly. It's psychotic.
I remember in high school reading a Tale of Two cities. Absolutely disliked how all the language is over dramatic and flowery. Try reading some scientific papers from the same era. Many of them are similar in how they were written.
Then sometime after I read something by Mark Twain for the first time. I was blown away at how much easier and more enjoyable his writing style was.
You might enjoy Derrick Sivers articles then. He goes out of his way to make everything short and concise.
You all would enjoy the density of Twitter more
GOOD Point by my wife! Why can’t we have a blower on a belt to accomplish most of the sorting? Plastics blow one way and glass stays on the belt. Throw in a magnet and make eddy currents to trap metals.
They do use eddy currents for metal sorting. Take some of your home recycling to a recycler instead of leaving it on the bin. It's worth going once just to see the small scale solution. Get yourself a bag of soda cans and throw a couple soup cans in there. Not the cleanest place you'll ever be, so don't wear your nice shoes and prep for some odors, but better than some.
When I was a child I saw one system that expected all aluminum that just used an electromagnet to lift any steel out of the stream. That was quite a ways up in the air so I suspect the owner would have gotten pretty pissed if we threw a bunch of nails or soup cans in the bag for fun. But most of the other ones I've seen have two conveyors with a gap, and they use an AC magnet to fling the aluminum onto the second conveyor (via eddy currents). Most of the ferrous stuff just drops into the first bin, or funnels off to be sorted again.
For your plastic or glass suggestion, that may take breaking it up first, since large items might not lift or go in weird directions. And then some plastics are pretty dense. I have the vaguest notion that I might have seen a system that sorts plastic from glass while washing it, since the glass sinks and some plastics don't, but that could have been a prototype or a misfiring neuron for all I can remember.
I think it is harder than just blowing them with air… different plastics have to be recycled in different ways. The number in the recycle arrows tells you what it is and how (or if) you can recycle it (see https://www.thoughtco.com/recycling-different-types-of-plast...). I don’t believe these plastics can be sorted by density or level of wind resistance.
I wonder, if the AI could identify recyclable items, then just lighting up items so humans could sort would be eaiser.
Lights and hands cost way less than complicated grip machines.
Why? Make the investment at centralized locations. No human should have to do a mind numbing job like that. Can’t even relax, it will be stressful too.
I worked at a mid sized startup where this idea was floated for developing a prototype. Neat to see it has become production level software somewhere
I don't think this is a particularly complicated idea actually. Why didnt you guys pursue it at the time?
ML was not the focus of the company, we didn't really have any qualified devs to pull off a project like that. It became a prototype for a sales pitch and I don't know what happened to it after. Though I disagree about it not being a complicated/difficult endeavour.
This is a solution without a problem? The global recycling industry has collapsed. It never broke 10% btw, and is currently at 4%. China doesn’t want our trash anymore. No one is buying, regardless of marginal cost.
Sorting has always been our worst enemy. We can’t recycle everything but can recycle a lot of materials really, really well.
Our world will so quickly come to resemble the fantasies of Star Wars etc. With multi-armed intelligent robots sorting recyclables, perhaps in the bowels of Jawa scavenger ships...
Is the use of AI necessarily reducing the, say the carbon footprint though?