but it turns out all that cash was going toward a workforce of over 700 Indian engineers, rather than an AI.
I doubt much of that cash was going to their workforce. Should have though.
On the internet no one knows you are 700 Indian engineers.
"Welcome back to tonight’s episode of “Is it AI or 700 Indian Engineers!!” 👏🏾👏🏾 👏🏾
I’m picturing a room of people with protractors ray tracing Doom.
I’ve got so indian answers from perplexity, and from chatgpt too… so this is why!
“Sachin! Quick, look up the best way to impress a first date who is vegetarian!”
What do you mean the new llm I invested in is just 700 Indians in a trench coat?!
AI stands for “actually indians”?
All Indians.
I’m waiting for “generalized” actually Indians
You mean A Guy Instead?
IP is “Infact Pakistanis”
And what’s TCP/IP?
Todays Chosen Pakistani/Instant Pakistani
Always Infosys
Anyone remember Cha-Cha?
Charming!
Crazy that 700 professionals in india is cheaper than a compute/data centre.
a data center isnt cheap, probably cost billions, a year to operate or millions. plus all the side effects, like power requirement and water waste.
I can run a small model on my home pc, if i had access to those kinds of resources, i could run a AI data centre for profit, probably more rssource efficient than those 700 dudes as well.
700 professionals in India probably make more coherent software than AI.
100%, but i dont think that will stay true for too long.
Of course. Just one more nuclear reactor bro, just one more trillion dollars investment bro
Oh is that what itll cost? /s
Yeah. Trust me on this, bro
In the meantime. This is a valid business model.
Depends what you mean by “valid”. If you mean “profitable”, sure: Fraud has always been a profitable business model.
But if you mean “valid” in terms of what Microsoft got out of their $455M investment, not so much, as they wanted a great new AI model, not the output that the “human-powered” model produced pretending to be an AI.
in a trench coat
Next do “self driving cars”
Check the trunk for an Indian guy with a street map.
You mean the 40 horsepower is actually 40 Indians under the hood??
I hope this isn’t part of a larger trend of human labor being devalued because companies pretend it’s just machine labor. I hope that’s literally impossible.
A lot of companies have been doing this for years. AWS literally sells this as a service: https://www.mturk.com/
i like that they named it after one of the most notorious instances of fraud.
Amazon SageMaker Ground Truth
Who names this shit? I want to have a serious talk with their mother.
OracleMaker was too problematic.
Amazon web services are incomprehensible. The names aren’t good either
Yeah the whole AWS ecosystem has a super shitty naming. Everybody knows S3, but what kind of name is that? All the other services are no better
Still better than EC2 and the likes. Where is EC1? And what about 3?
EC3 is wrestling in NWA now.
Simple Storage Service
E: I agree that their names are shit though. But S3 makes sense (once you know what it means). Just like
boto3makes sense (once you know what it means)Probably inspired by the Minnesota Mining and Manufacturing Company, or 3m for short
Next I’m going to find out ChatGPT is 700 thousand Indians typing really fast.
Hurry! We need a photoshop of what a Col Sanders action figure would look like.
But all sharing the same keyboard? I need to understand the logistics behind that feat
They wouldn’t be able to type fast enough if they weren’t using the same keyboard.
It would save electricity
That would explain why it sometimes gets sluggish!
Oddly between 5pm and 6am Delhi time.
Removed by mod
Removed by mod
Isn’t this exactly what was exposed at the Amazon “Just Walk Out” stores? Turns out all the cameras and sensors weren’t good enough, so they paid thousands of people in India to watch videos and correct checkouts. They basically just outsourced the position of cashier, while pretending it was all done automatically!
https://gizmodo.com/amazon-reportedly-ditches-just-walk-out-grocery-stores-1851381116
I built some of the components that went in to the test locations. Amazon had absurdly tight tolerances for the parts they were buying. They effectively wanted a shelf that was also a scale, and the tolerances they demanded weren’t really necessary. So it was an insane expense but they paid it and wouldn’t hear otherwise.
My company also made most of the lockers they’re using in places like Whole Foods, and Amazon insisted on controlling the entire design process themselves. They sent us prints, we made parts. They made it very clear that that was the relationship they wanted, so we complied. No test runs, THAT would be too expensive. Let’s just make ten thousand parts and put them together.
I would like to be very clear that in an industrial setting, this is unusual. You need something specific, you call a company that makes things like it and see if they can make what you need. You have a conversation about what you need it for and how many you want. The relationship is personal, you get to know the people around the region that you need stuff from.
Amazon swooping in with a heavy purse and a list of demands is weird, when someone kicks in your door with a stack of prints and enough money to keep the entire plant in overtime all year, it’s hard to say no to that.
So the first batch of prints they send is wrong. Parts do not line up right and the doors don’t even fit. We didn’t discover this until 70% of the components had already been painted.
Second batch they assure us addresses the problem, we need to start over.
My friends, it did not address the problem. Half the changes they needed to make they didn’t. The doors still did not fit.
3rd try, we lied and said we needed some extra time because a different client had elbowed in with a large order while they were redesigning. We had an intern recreate every print in CAD and test fit it, we ran a single batch of test pieces to assemble one row of lockers and as we were doing that they sent a revision.
They finally got their lockers, and asked for basically book dividers but insisted again on insanely tight tolerances.
After the dividers went out we stopped taking their calls.
Amazon be smokin that meth again.
I worked as an associate for a public accounting firm that does not ever advertise itself, because we specialized serving ultra wealthy individuals and you could only engage us if you knew of us through such circles.
One day, our office got a call from the personal assistant to someone very wealthy who is known for abusing ketamine, asking for an engagement on a very unusual and complex tax situation. A call was set up to discuss the scope of the engagement, because the partners have always been very particular about what clients they will take on, because really wealthy individuals are often very unpleasant, stressful, & frustrating to work with.
Apparently during the call the assistant was patronizing, like we should feel flattered that we were chosen by m’lord, and demanded non-negotiable terms that we would conduct our work exactly as told with no questions asked. They had even sent their own engagement letter for us to sign with them ahead of the call, and it was completely absurd.
The partners patiently explained that is not possible, as that is not how this type of professional relationship works, and declined the engagement.
The assistant was losing their mind, shocked we would turn such an opportunity down. They offered even more money and even some compromise, but the way they initiated the interaction set the tone to expect throughout the professional relationship.
I was very impressed by the partners in the sense that I knew they were incredibly greedy people, but they are so fucking intelligent and had such a great instinct to avoid clients that were going to end up costing way more money than they brought in, because us associates would absolutely refuse to deal with bullshit because it was already a super stressful job, and we were way too talented and incredibly expensive to replace if we walked off.
The self restraint must have been legendary, and exactly the right call, because all the professionals that do end up accepting end up getting embroiled in costly lawsuits and getting thrown under the bus.
Anyway, I hated that job and I wish I that quit sooner than I did. I got such bad burnout, I developed PTSD and now I prefer just living like a hobo rather than go back out there.
PS: Fuck capitalism and fuck Amazon. I refuse to buy anything from them ever again. Cancelled my credit card and told them to go fuck themselves. Fascists.
Sometimes you have a run in with a customer that ain’t worth having-- no matter how much money they pay.
Peoole aren’t appreciating just how bad these things are because they’re misinterpreting it. The goal of what they are doing here and with Amazon was never to just fake the technology right. The goal was to fake that the technology existed by using humans to do an automated thing and then to leverage that into making it actually automated.
But essentially what that means is theyre inventing technology that hasn’t been invented yet and selling it to you and the reason for doing so is to replace you with technology before it can even technically happen.
It’s essentially like someone building a new automated factory and telling workers at their other locations that they can’t be hired there since it’s automated but then someone goes inside and finds out they’re just using child laborers until the robots are ready and also robots haven’t been invented yet.
They’re using blood to grease wheels that don’t even exist to turn yet.
Feels like it should be illegal to mislead people like that.
On the other hand, the only way to get good training data is to generate data indistinguishable from the real-world scenario and then have humans mark it up the way you want the system to do it. You might as well have the data actually be from the real world and recoup some of the costs with sales.
Sure, but you still shouldn’t be selling the technology as actually working, instead of developing.
Amazon bought whole foods a while back. What would have stopped them from just collecting the data in their own stores, and then developed the tech?
Hint: shareholder value.
What would have stopped them from just collecting the data in their own stores, and then developed the tech?
I won’t pretend that Amazon avoided that due to ethical concerns, but doing that would have almost the exact same ethical concerns.
All they had to do was run the tech alongside traditional cashiers. Make it known on entry, and your fine. No ethical concerns.
But what they did was sell tech they didnt have to shareholders to pump up the stock.
The lying is unacceptable, but either they hire temporary workers to obsolete themselves, or they force tenured people to obsolete themselves.
From an engineering perspective they didn’t want to do this since it’s not just about AI tasks. If you go watch videos of it they have camera arrays and special shelf layouts and all sorts of stuff.
Not to mention the engineers probably wanted to be able to test it privately and without disrupting an actual store and community.
So it’s what I would’ve done as well frankly
What are you talking about?
It was never AI. It was always cheap remote people working in foreign countries. But you would take that, and sell it as AI like they did?
Yes, it’s the exact same practice.
The main difference, though, is that Amazon as a company doesn’t rely on this “just walk out” business in a capacity that is relevant to the overall financial situation of the company. So Amazon churns along, while that one insignificant business unit gets quietly shut down.
For this company in this post, though, they don’t have a trillion dollar business subsidizing the losses from this AI scheme.
JWO hasn’t shut down. The system got polished enough for them to sell it to other companies, so they don’t need their own test-platform locations anymore.
JWO and similar systems do not reduce labor. The people working cashier become customer service attendants. These systems are valuable when the issue is throughput and sales are being lost at peak times. Airport convenience stores and stadium concession stands, for example, can get significantly higher revenue for the same footprint.
And many delivery robots are helped along by a remote worker.
They should have had 701
