I do industrial automation for a living, and I just want to point out that automating things that exist purely in the digital domain is far easier than automating things like ship breaking.
Cant imagine how it even could be automated without advanced robotics. Those ships are freakin HUGE! Maybe a collection of robotic snakes with cutting lazers attached to their heads and some little scuttle bots to pick up the pieces the snakes knock off? Just cut the whole thing into 1’ disks or maybe hexagons is better
Maybe a collection of robotic snakes with cutting lazers attached to their heads
Upvoot for the matrix
What do you think snakes are
Just make a huge version of those supermarket bread slicer machines and feed the ships through it.
Or better yet, build a bigger ship and use it to smash the smaller ship to pieces
Ngl a ship eating ship is metal af
I just read The Three Body Problem, and I have some ideas on how it could be done.
Oh! One of my favorite books, have you read all three?
The three book problem
Not yet, but I will!
Oh, there is actually a fourth one by a different author but with Liu Cixin’s approval. It goes deeper into the trisolarian biology than the main series. I havent read that one yet
I’d suggest skipping it. While it’s a fun continuation, the writer is not nearly as good as Liu Cixin. It’s definitely a fanfic
This kind of makes me curious! I didn’t finish the third book of the original trilogy …
Definitely not terrifying
[…] I just want to point out that automating things that exist purely in the digital domain is far easier than automating things like ship breaking.
Not that you’re saying otherwise, however isn’t that even more of a reason more developers and resources should be allocated toward automating complex and risky physical processes?
Honestly, I don’t see how you would do it without general AI, which is something that will be solved in the digital domain first anyway.
Eh, it could be done with non-general AI. There are a finite number of different types of things to handle, so as long as it’s not thrown off by some bent steel or some missing consoles, I’d be amazed if they couldn’t automate at least specific ship designs.
Automation requires very high precision/consistency in the parts you want to work on. I seriously doubt that after many years of wear, tear, and impromptu repairs, those ships would be anywhere near consistent enough.
That’s why I said, “eventually with non-general AI”.
Even a well written algorithm could work with something that’s mostly in expected shape. How in the flying fuck is everyone so brainless that they cannot understand non-general AI can still adapt to things? Fucking hell.
I’m not talking about current industry practices. I’m talking about combining existing technology with unlimited bidget to create a factory that could kinda’ do the task.
“Possible” and “practical” are two extremely different things, and you goons pointing out that most obvious basic fact are adding nothing.
A single repair or modification would ruin the entire automation process. One single screw off by a single mm type thing.
Why the flying fuck do you think I said, “non-general AI”? Even a well written algorithm could handle things coming in not in perfect shape, yet everyone pretends “non-general AI” means, “execute instructions repeatedly without any input what so ever.”
Use your brain. Even basic dumb algorithms that can run on an Arduino can respond to input. Machine learning can easily respond to dynamic input, so stop failing to imagine the most basic of basic things I say.
Removed by mod
Removed by mod
Not that you’re saying otherwise, however isn’t that even more of a reason more developers and resources should be allocated toward automating complex and risky physical processes?
You’re solving for the wrong problem from the perspective of people with money investing money to solve these problems.
-
Shipbreaking, while dangerous for the workers, isn’t expensive because it is done in far flung countries with workers that have low wages, few protections for safety, and long term health consequences.
-
Art and writing (for western consumption) requires educated and talented people which are expensive to employ.
People with money, looking for a return, want that return their spending, not reduce human suffering.
-
Processing the digital world is just the first step. You can’t just build a safe autonomous ship disassembly robot without making sure your algorithms are actually sound. Look at self driving cars, they’re far from being safe and acceptable. Jumping straight into this problem without testing the shit out of your code in a virtual world is a mistake.
I mean automating it would certainly be a challenge but the first step would be building tools and robotics to allow human operators to more safely and effectively manage the tasks. Then you streamline the industrialized processes. Then you think about automating things.
But this is all really an economic problem, not a technical one. Software tools have minimal resource costs (compared to building/destroying a ship) but require skilled (expensive) laborers to operate. So to cut costs in any digital field you need to get rid of the expensive laborers. Thus the push for AI to replace any computer-bound work. Physical labor is already considered dirt-cheap in our fucked society, and no one is rushing to add expensive tools in fields where disposable people will suffice.
I sympathize immensely with the OP image’s final point, but “working for the right company” isn’t going to fix it. Reorganizing society is necessary, rethinking what we culturally value and uphold.
I think the solution for ship breakers is for the job to be a highly paid respectable job with protections. In other words the technology that desperately needs to disrupt this industry is probably… unions
Unions protect against automation that reduces labor hours.
Honestly more unions should fight for company stock for employees or similar stake programs. As we hopefully get more automated having workers interests aligned against it seems like a losing fight.
Finally, some fucking sense into all of this.
Yeah exactly, I work in AI and robotics for medicine, and im so goddamn sick and tired of these people and their absolute god-awful uneducated takes on AI.
Shh. Just give one of them dancing robot dogs an impact driver attachment. They’ll figure it out in a week.
Once we perfect doing it in software, then we can graduate to hardware. Today, digital paintings; tomorrow, real paintings; next year, tear down a fucking ship!
And yet, people do… https://www.leviathan.eu/
That is really cool job description I haven’t seen pop up before! Would you mind sharing what type of things you need to automate? It sounds so interesting, I never really understood why factory line jobs should exist for example * because the work is dangerous, the opposite of stimulating/engaging (works for some sure), and just generally overall depressing unpleasant places to work. We SHOULD be striving for a world where humans don’t have to do such menial unfufilling work.
*very superficially, all the nuance that makes it continue to be necessary and exist I understand)
I work in the auto industry, so programming the machines that make the car parts. Humans are still involved because getting machines to handle changing conditions is very slow, expensive, and still winds up unreliable in a lot of cases. The simple process of picking a randomly oriented part up out of a bin and placing it accurately on a fixture is actually very difficult for a machine to do, when compared to how easily a human can accomplish the exact same task.
“AI” researcher here. The only reason there are models that can “write” and “create art” is because that data is available for training. Basically people put massive amounts of digital text and images on the Internet and the companies scraped all of it to train the models. If there were big enough datasets for ship building, that would happen too…
Besides, what the guy is yapping about it is 80% a robotics problem not an AI problem. It’s apples and oranges.
He’s essentially saying why can Will Smith finally eat pasta normally while we still don’t have the robotic workforce from the 2001 Will Smith movie “I, Robot”.
Not really. You would still need to, you know, build drones or automated factories to actually perform the salvaging. But the point is that nobody DID, because capitalism values profit over human life. Nobody who “matters” is interested in solving that problem.
Not true. We have capable robots now. See Boston Dynamics like the other commenter said. Plus we have had industrial robots making cars and stuff forever now. To make robots that can handle a wide variety of things (every ship is bound to be different) is hard and we don’t have data to train such models (see reinforcement learning, imitation learning, “sim2real” problem etc)
This is kind of a dumb argument, isn’t it?
I have to imagine someone centuries ago probably complained about inventors wasting their time on some dumb printing presses so smart people could write books and newspapers better when they could have been building better farm tools. But could we have developed the tractor when we did if we were still handwriting everything?
Progress supports progress. Teaching computers to recognize and reproduce pictures might seem like a waste to some people, but how do you suppose a computer will someday disassemble a ship if it is not capable of recognizing what the ship is and what holds it together? Modern AI is primitive, but it will eventually lead to autonomous machines that can actually do that work intelligently without blindly following an instruction set, oblivious to whatever might be actually happening around it.
The argument isn’t against the technology, it’s against the application of that technology.
Path of least resistance. It is harder to build a robot who can disassemble ships with its hands than it is to pattern match together pictures.
This XKCD comes to mind: https://xkcd.com/1425/
Someone doing it automatically doesn’t stop you from doing it manually. It’s still a meritless argument.
I get the sentiment, but it’s a bad example. Transformer models don’t recognize images in any useful way that could be fed to other systems.They also don’t have any capability of actual understanding or context. Heavily simplifying here, tokenisation of inputs allows them to group clusters of letters together into tokens, so when it receives tokens it can spit out whatever the training data says it should.The only actual things that are improving greatly here which could be used in different systems are natural language processing, natural language output and visual output.EDIT: Crossed out stuff that is wrong.
Well, this is simply incorrect. And confidently incorrect at that.
Vision transformers (ViT) is an important branch of computer vision models that apply transformers to image analysis and detection tasks. They perform very well. The main idea is the same, by tokenizing the input image into smaller chunks you can apply the same attention mechanism as in NLP transformer models.
ViT models were introduced in 2020 by Dosovitsky et. al, in the hallmark paper “An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale” (https://arxiv.org/abs/2010.11929). A work that has received almost 30000 academic citations since its publication.
So claiming transformers only improve natural language and vision output is straight up wrong. It is also widely used in visual analysis including classification and detection.
Thank you for the correction. So hypothetically, with millions of hours of GoPro footage from the scuttle crew, and if we had some futuristic supercomputer that could crunch live data from a standard definition camera and output decisions, we could hook that up to a Boston dynamics style robot and run one replaced member of the crew?
And such is the march of progress.
Huh? Image ai to semantic formating, then consumption is trivial now
Could you give me an example that uses live feeds of video data, or feeds the output to another system? As far as I’m aware (I could be very wrong! Not an expert), the only things that come close to that are things like OCR systems and character recognition. Describing in machine-readable actionable terms what’s happening in an image isn’t a thing, as far as I know.
No live video no, that didn’t seem the topic
But if you had the horsepower, I don’t think it’s impossible based on what I’ve worked with. It’s just about snipping and distributing the images, from a bottleneck standpoint
No live videos
Well, that’d be a prerequisite to a transformer model making decisions for a ship scuttling robot, hence why I brought it up.
Describing in machine-readable actionable terms what’s happening in an image isn’t a thing, as far as I know.
It is. That’s actually the basis of multimodal transformers - they have a shared embedding space for multiple modes of data (e.g. text and images). If you encode data and take those embeddings, you suddenly have a vector describing the contents of your input.
I get the sentiment, but that is a really dumb take. Software automation is a hell of a lot easier than creating robotic automation to disassemble ships of all shapes and sizes. That’s why art automation has been done, and industrial freighter recycling automation has not been.
How would that even be possible? Presumably, you’d need to break the ships down into pieces first, and even then, you’ll be dealing with huge numbers of oddly shaped and sized components of varying materials. It makes a lot more sense to have people do that, though it is likely very dangerous.
Seems more like a job for unions and workplace safety regulations than for robots
more like a job for unions and workplace safety regulations
Yes. That’s why they do these things in third world countries. The people there are cheaper than robots will ever be.
3th
threeth.
When teaching about programming languages that are zero indexed, I avoid the word “first” because it is ambiguous and instead use “zeroth” and “oneth.”
I get the sentiment, but that is a really dumb take.
$13B invested in OpenAI feels more and more like malinvestment and graft, incentivized by our disastrous energy policy and enormous tech subsidies.
This isn’t purely software automation. Its also an investment in physical media and machines, new or renovated energy infrastructure, and enormous volumes of potable water.
Seems more like a job for unions and workplace safety regulations than for robots
In 2020, a leaked company memo detailed Amazon’s use of a new technology — the geoSPatial Operating Console (SPOC) — to analyze and visualize data sets pertaining to threats to the company, including unions. Reported by Jason Del Rey and Shirin Ghaffary at Vox, some of the data points related to unions include:
Amazon-owned Whole Foods’ market activism and unionization efforts. Flow patterns of union grant money. The presence of local union chapters and alt labor groups.The approach is an obvious attempt by the company to use more passive means of identifying and neutralizing union sympathizers in the company.
“Amazon’s tracking of workers’ micro-movements, decision points and searches and then linking all of that data to that of unions, community groups and legislative policy campaigns is union busting on its face,” said Stuart Appelbaum, President of the Retail, Wholesale and Department Store Union (RWDSU) in a statement at the time.
That is very true, but my critique was more focused on the difference between automating software tasks vs mechanical tasks, especially with non-uniform inputs and not the economic investment required. Some tasks are better suited to automation - and plagiarizing art is far easier than deconstructing and recycling massive industrial freighters.
Not on the side of the AI art generators here - that was just low hanging fruit compared to something like was suggested in the original post. Definitely need extremely strong labor law to protect against AI union busting (and union busting generally)
my critique was more focused on the difference between automating software tasks vs mechanical tasks
Somewhat paradoxically, we’ve been much more successful automating mechanical tasks than digital ones. We’ve had steam looms and automotive assembly plants far longer than server farms and super computers.
And I might argue this kind of automation has been far more fruitful. I can point to a lot more in my daily life that has benefited from the industrialization of steel and plastic fabrication than what I’ve received from Google Search Results.
To say the millions of man-hours and trillions of dollars sunk into the advertisement and entertainment industries couldn’t be put to better use… Come on, man. The latest Marvel movie wasn’t so good that I wouldn’t have traded it for a globalized 1980s British NHS.
deleted by creator
I think you absolutely nailed the analysis. Another small point to keep in mind is that for Microsoft, all the investment in OpenAi comes back as a revenue figure when the system works operating on top of the Azure platform.
Software automation being easier seems like a reason to not have so many people doing it, then? Like, the harder problem is the one that could really use all of the focus?
But the harder problems aren’t as obviously profitable for a large number of tech CEOs, and they’re not ripe for being a “winning glittery ticket” for a large number of comp sci students looking to be the next big thing in Silicon Valley.
Making art and writing just happens to be easy to automate with neural networks and machine learning, neither of which was originally researched for the purpose of replacing artists and writers.
Good luck disassembling a ship with a neural network. And maybe do some research about the difficulties of application-specific robotics.
Ah yes just write code for the ship fold itself neatly back into reusable materials.
Just build a grinder the size of a football stadium to shred battleships into pea-sized chunks, and sort according to metal type, how hard can it be?
It might be more cost effective to build a concrete bunker the size of a football stadium, use placed explosives to blow up the ship inside of the bunker, and then shred the exploded ship up into pea-sized chunks
Teach me how to code that and which compiler will spit out the football stadium grinder
Teach me to how to code a compiler to spit out fully realistic and accurate videos via text prompts.
I hate this take because I dream of a world where AI can assist any storyteller in bringing their story to life.
The rest is just capitalism. Capitilism is the issue, not the AI.
OP: “We’ve tragically gone down a path of quantifying and min-maxing every aspect of existence, including creativity and the value of human life.”
Comments: “OP clearly doesn’t understand the comparative efficiency of the ROI here.”
Irony so thick you can cut it.
He chose a poor way of conveying his message.
Disappointed programmer here. I thought I could automate farming so that people wouldn’t die of hunger. Now I realise that if you automate farming, it would just make some CEO more money because his company now makes corn syrup and destroys rural communities even faster.
I got my “contract not renewed”, for the Fortune 500 B2B CRM company I worked for.
I can try to bust my ass to make my 2018 laptop try to render images I can’t draw, which does give me some pleasure. It’s not the AI tool’s fault humanity sucks, it’s the goddamn people with money.
Oh no. You can’t do it for fun now because the computers are doing it.
This sort of ignores the fact that the advances in that technology are widespread applicable to all tasks, we literally just started with text and image generation because:
-
The training data is plentiful abd basically free to get your hands on
-
It’s easy to verify it works
LLMs will crawl so that ship breaking robots can run.
Second this.
We’re in the first days and everyday I add a new model or tech to my reading list. We’re close to talking to our CPUs. We’re building these stacks. We’re solving the memory problems. Don’t need RAG with a million tokens, guerrilla model can talk with APIs, most models are great at python which is versatile as fuck, I can see the singularity on the horizon.
Try Ollama if you want to test things yourself.
Use GPT4 if you want to get an inkling of the potential that’s coming. I mean really use it.
-
I remember years ago everyone was saying that art would probably be the last thing AI would be able to handle and menial jobs would probably be the first.
Now look at where we are!
This shit again?
The tasks AI is replacing only require powerful computers and internet access.
If you want to make that comparison, to scrap fucking ships using AI, you need a robot that the AI can control.
Or what else do you want to do? Putting a fucking computer server that is running some ship scrapping AI in the middle of a shipyard and see if it magically grows arms?
No, I’m not denying we have an issue with this fucking capitalism (with and without AI), but stop comparing “software” tasks with other tasks what would required specialized machinery/robots.
Isn’t the point that we don’t bother looking into those specialized machines and tools because why bother when we can just throw meatbags at it?
There are plenty of people working on automation for manual tasks, but it’s a really hard problem. Making machines that can move around freely and are compact as humans is really hard. Automation works really well on assembly lines where parts can move to the machine.
But this is total nonsense because those tools are getting developed and have huge budgets. Many of them are already on the market and in use, especially remote control cutting tools.
Far more money has been invested in self drive and ambulatory robotics than image gen, it just so happens image gen is far easier than walking or using a saw.
Gpt 5 is coming around October and I think it’ll likely be the version that is able to effectively create task based workflow so it’ll be able to set up simulation training to evolve kinematic solutions within a framework, basically the thing we need robots to be able to do. When that’s possible you can expect to see a big boom in multiuse robotics.
we don’t bother looking into those specialized machines and tools
Yes we do. It’s just a hell of a lot more expensive, and a lot more difficult.
The image generator AIs are a byproduct of image recognition AIs, they’re related.
Image recognition development is fundamental to advanced industrial automation. We’re getting there, the media is just not covering that part because it’s more fun to write an article about stupid computers thinking we have 6 fingers than about false positives dropping 2% because of some new development
Big reason why I just build cute little games as a hobby instead of writing spreadsheet software for a megacorp to optimize the lowest quarterly earners out of a job, or develop AI to optimize myself out of a job.
People are cheaper than robots, ergo they are more expendable
