Comment by eksemplar | Hacker News Reader

Comment by eksemplar | Hacker News Reader

Parent Full thread

I work in a digitization departement in a multiplicity, which translates to me doing almost nothing but replacing jobs with software automation.

Things like going through 500.000 cases containing multiple documents and sorting them correctly can be done with Azure services in 5 hours after a 14 month algorithm training period. It takes 5 people working 37 hours a week 3-5 months to achieve the same result and they will have a larger margin of error than the machine. I know this, because we did the same process as a competition between man and machine.

Getting back to the point, however, everyone loves my department. Part of this is because people aren't really fired due to automation, we solve the freed up resources by not hiring people when someone retires, but the main reason is because we automate boring stuff.

Nobody wants to sort through 500.000 cases manually, so you're absolutely correct, automation can free up humanity.

Some people will spend their time being productive, but the truth is you don't have to. That is, if we manage to take back ownership from companies.

Today western society is build around income tax, and as long as that doesn't change, we're not heading for freedom but for oblivion.

> we solve the freed up resources by not hiring people when someone retires...

If we define technology as doing more with less, automation then because the greatest technological feat of humanity.

> Today western society is build around income tax, and as long as that doesn't change, we're not heading for freedom but for oblivion.

Nailed it!

>That is, if we manage to take back ownership from companies. Today western society is build around income tax, and as long as that doesn't change, we're not heading for freedom but for oblivion.

I don't understand. Ownership of what do you want us to take back and who is us?

Not OP, but I read this as "us", as in, working people in the lower or middle class with no notable investments outside of maybe a mortgage, and ownership being ownership of the world's companies (i.e. currently owned by stock- or shareholders depending on whether the company is public or private). I believe they were referring to the worst case scenario where automation removes most jobs (which are not just replaced with others), but because the automation mainly benefits the owners of the companies that own the robots/AIs, the majority of people would simply lose the ability to earn money (because they can no longer find work, and they don't own anything that can generate income for them in other ways). The fact that goods can be produced cheaper is little help if you cannot earn any money at all. The worst case scenario is a breakdown of consumerism, as most consumers stop being able to earn money, and all wealth accumulates with the owners of companies in a way that would make the current 1%/99% split look like a socialist utopia. Is this likely? No idea, but it is one of the possible futures that the current trends point toward, and it has a lot of people worried, and is one of the reasons why there's renewed interest in concepts like basic income among futurists.

That scenario would mean ruin for everyone, the 1% and the 99%.

How do you figure? If they can produce anything with automation what do they care if there's no one to purchase it? Value will be expressed in other ways. I could easily imagine the 1% or whoever completely controlling production and allowing everyone else to die.

Could you detail what type of processing are you doing to the documents? Merely OCR or some type of NLP?

Mainly OCR. We were looking for cases where a specific set of law required forms were missing.

It took 14 days to train the algorithm to recognize these forms, and maybe 1,5 months to learn how to use the algorithm and the Azure cloud.

I can't give you the exact technical details because I'm not the CS guy, but I can tel you that we used Googles OCR and a common tree algorithm suggested by "the internet" for machine learning.

The machine vastly outperformed humans in most cases, especially with 50+ page compilations, but got beat when documents had been scanned extremely poorly.

Reply on news.ycombinator.com