Live data from Hacker News

Amazon grows to over 750k robots, replacing 100k humans

finance.yahoo.com

201–210 of 400 posts

Re: Amazon grows to over 750k robots, replacing 100k humans

#201
post #52

It's clear that technological development creates a shift in jobs, i.e some jobs are lost, but new ones are created as a result. Whether the total #jobs increases or decreases is debatable. The issue that I see addressed less often is that the new jobs require by definition, a higher skill set. (You wouldn't displace 100 manual labor workers with a machine that requires 105 workers to maintain). So by definition, the…

When intelligence becomes truly abundant and cheap, then human intelligence won't be valued by the market. You only get paid well because your intelligence is scarce, but that won't be the case in a future with ASI.

It's hard to predict what will be valued, maybe personality traits that lend themselves to jobs like social media influencer, politician or actor, because people may still value real humans in these jobs.

Re: Amazon grows to over 750k robots, replacing 100k humans

#202

This is self-reported: remember the 1000 Indians that were watching people shop at the Amazon Go stores? Amazon told us it was all "sensors" because it fit their company narrative to do so. I am not saying that Amazon doesn't have 750k robots and hasn't laid off 100k people... but they usually have some seasonality, plus, the quoted number is from 2021, the height of at-home shopping. "The world's second-largest priv…

The 1000 Indians story was completely incorrect though. They has 1000 people in India tagging the videos after the fact for improvements to the machine learning algorithms. They weren't watching in real time.

I have a question: was the customer billed based on whatever the machine learning algorithm determined, or based on what humans decided after the fact? In other words, was the humans' job to just provide training data for future versions of the machine learning algorithm, or to manually fix the algorithm's errors and prevent over-billing?

Re: Amazon grows to over 750k robots, replacing 100k humans

#203
post #99

Earlier quoted context omitted.

The chain of logic is falsified by the Whitney cotton gin: it was a labour saving device, which saved enough labour to make cotton much more profitable, which led to the growth of the cotton plantations in southern USA, which led to increased slavery, and those slavers actively prevented their slaves from learning to read. https://en.wikipedia.org/wiki/Cotton_gin https://en.wikipedia.org/wiki/Anti-literacy_laws_in_th…

IQ is sort of useful to measure humans because we have roughly comparable skills. It’s not applicable to AI at all in terms of measuring job fitness. If I need someone to move some furniture it might only need an IQ of 85 but that doesn’t mean AI is doing it any time soon.

> If I need someone to move some furniture it might only need an IQ of 85 but that doesn’t mean AI is doing it any time soon.

The article and discussion are literally about machines moving stuff.

Re: Amazon grows to over 750k robots, replacing 100k humans

#204

This is self-reported: remember the 1000 Indians that were watching people shop at the Amazon Go stores? Amazon told us it was all "sensors" because it fit their company narrative to do so. I am not saying that Amazon doesn't have 750k robots and hasn't laid off 100k people... but they usually have some seasonality, plus, the quoted number is from 2021, the height of at-home shopping. "The world's second-largest priv…

The 1000 Indians story was completely incorrect though. They has 1000 people in India tagging the videos after the fact for improvements to the machine learning algorithms. They weren't watching in real time.

There are reports that 70% of the sales had to be manually reviewed by the team in India (https://www.businessinsider.com/amazons-just-walk-out-actual...), so no, they weren't just training algorithms.

Re: Amazon grows to over 750k robots, replacing 100k humans

#205
post #204

Earlier quoted context omitted.

The 1000 Indians story was completely incorrect though. They has 1000 people in India tagging the videos after the fact for improvements to the machine learning algorithms. They weren't watching in real time.

There are reports that 70% of the sales had to be manually reviewed by the team in India ( https://www.businessinsider.com/amazons-just-walk-out-actual... ), so no, they weren't just training algorithms.

That article in particular was the one that was debunked. It was based on a false rumor.

Re: Amazon grows to over 750k robots, replacing 100k humans

#206
post #39

Earlier quoted context omitted.

How detached from reality are you that you think people can make ends meet while working 15 hours a week?

Some people do do that. Try stepping outside of your bubble.

I concur. They're not looking in the right places. Musicians and artists would know about these sorts of hustles, but not your typical HN bubble dweller.

Re: Amazon grows to over 750k robots, replacing 100k humans

#207
post #204

Earlier quoted context omitted.

There are reports that 70% of the sales had to be manually reviewed by the team in India ( https://www.businessinsider.com/amazons-just-walk-out-actual... ), so no, they weren't just training algorithms.

That article in particular was the one that was debunked. It was based on a false rumor.

Debunked where? The only response is from Amazon (in the same article), who are disputing the number but still admitting that the team did intervene when needed.

Re: Amazon grows to over 750k robots, replacing 100k humans

#208

Earlier quoted context omitted.

Per the articles I read, they did more than label training data. They also did other things to facilitate the checkout process. They didn't get into what those things were however. > "Associates may also validate a small minority of shopping visits where our computer vision technology cannot determine with complete confidence an individual's purchases," the spokesperson said.

"may also" still doesn't suggest that the remote observers were doing all the work and the computer vision systems didn't work and it was all a farce though. spot checking that the system works isn't the same thing as doing the job of the whole system

No one is suggesting that. The question is how automated was the system? According to the article 70% of all transactions were manually checked out because the algorithm didn't have enough confidence in its prediction. Amazon has disputed the number, but not the fact that the manual intervention was happening.

Re: Amazon grows to over 750k robots, replacing 100k humans

#209

I wish I could believe that all this automation would lead to more leisure time for all of humanity, but it certainly won’t happen on its own out of generosity from the robot owners.

We've been saying that for a long time. [Keynes predicted we'd be working 15 hour weeks by now]( https://onlinelibrary.wiley.com/doi/full/10.1111/ecca.12439 ) The thing is, people don't actually want to tradeoff their labor time for leisure time all that well. They value consumption too much. There are plenty of 15 hour jobs out there. You can get one right now! But you'd need to have a smaller appartment, an older c…

> There are plenty of 15 hour jobs out there. You can get one right now!

Try talking to recruiters and tell them you want to work 2 days a week as a software developer and see how far you get.

Re: Amazon grows to over 750k robots, replacing 100k humans

#210

Earlier quoted context omitted.

Per the articles I read, they did more than label training data. They also did other things to facilitate the checkout process. They didn't get into what those things were however. > "Associates may also validate a small minority of shopping visits where our computer vision technology cannot determine with complete confidence an individual's purchases," the spokesperson said.

The whole discussion around that headline has always been so confusing. It's honestly unclear. What I'd like to know is, what % of actual costumer checkouts required a human being vs what % was 100% automated. Some of the headlines and negative discussions make it sound like 100% of the checkouts required human approval. If that is true, then that's clearly terrible. If on the other hand it was anywhere below 5-10%,…

> Something like Waymo also has humans constantly monitoring and helping, but I don't see much discussion about that being "controlled by humans".

Constantly doesn't really make sense, riders can call for support and get a near instant response, but it's pretty obvious from some of the mistakes that humans aren't watching every ride in real time. There's video from inside a Waymo which is stuck in a lot repeatedly looping, discovering the exit it wants to use is blocked and then just driving to another exit before deciding hey, perhaps the exit I want to use is open (nope, it isn't) and returning again, and again, and again for example.

All the humans involved can see what's wrong here, but the Waymo driver software doesn't get it and so I think eventually a human (Waymo employee) has to drive the car out.

And yeah, there's definitely in San Francisco for example some people who believe it's all a Mechanical Turk again, even though that doesn't make any sense. In their heads, driving is another of those "uniquely human" abilities and so the only way to have AI cars would be fully general AI, and the real explanation has to be off-shore remote driving.

Post reply on HN