The article has this to say about Google Translate's improvements in accuracy:
"Translate has hoovered up gigantic quantities of parallel texts into its database. A particularly fertile source of these useful things, apparently, is the European Union’s set of official publications, which are translated into all Community languages."
The first thing I thought was "what happens when the EU starts outsourcing its translations to Google Translate?"
Is this the future of machine learning? The learning algorithms start by mining a corpus of human output. Once they get good enough, they replace the majority of humans that generated the corpus. We then enter an echo chamber of machines largely feeding off their own output. Consequently, improvement of the machines stagnates, but the machines are still doing a good enough job to keep humans out of a job. We then have a future of "good enough that the cost of improvements can't be justified, but bad enough to be irritating"?
Humans have a sense of pride in their work, and will strive to improve for their own edification, even when the cost outweighs the benefit, or they have been told not to. A machine will just continue to deliver the level of service that the committee in charge tells it to.
My point is that those new algorithms need something to learn from, and the humans that used to do the job are no longer in the game. The original corpus could be reused, but then performance will be bounded by that corpus. If better algorithms are trained on the output of worse algorithms, presumably they just emulate the performance of the worse algorithm. Where do the better algorithms get their input from, if a large scale human effort no longer exists?
But why would every translator stop working or creating new works just because machines can do the job too? I don't think computer written novels will mean people stop telling stories.
Translation is an interpretation of the best phrase to use, and has a subjective element. Imagine trying to translate jokes - it depends on your sense of humor too.
Automation rarely replaces 100% of human workers. What tends to happen is that it replaces 99% of humans, does work that's almost identical to human work, and the 1% of humans left fix the machine work so it is identical to what 100% human workers would have done.
It's already happening in translation. Instead of a human translator translating an entire text. They simply first feed the text through translation software. That does a pretty decent job, but still makes mistakes. However, the mistakes are obvious to the human translator. Who now just fixes the translation. Most of the work is done by the machine. And a firm with 10 human translators can get rid of 9 of them, and still be as productive as it was it was with all 10 people.
This pattern of automation replacing almost all, but not exactly all, human workers is seen in many industries.
Machine learning researchers aren't going to stop looking for ways to develop better translation algorithms, learned more effectively from the vast volume of data that already exists.
While I don't think diplomats and legislative drafters are going to be put out of a job by Google any time soon, the problem is that if research seems to be yielding diminishing returns management will often cut off funding, even though the algorithm might just be stuck on a local maximum.
Maybe we will be content to have "good enough," but I'd think that there is enough value in good quality translation that Google and others will pay people to "train" their algorithms and data set. Instead of waiting till someone tediously translates things, and then trying to learn from it, it would have skilled translators skim over the output, and correct it where it deviates from perfection, while providing concise feedback as to how it got things wrong, how bad it is, why it is wrong, etc. Which is one example of the type of skilled job that I see opening up in the future....the training (not "programming" per se) of robots.
The same thing can happen for, say, training a self driving car how to merge. A good driver can let the car attempt it itself, but slap it into manual when it is failing, then provide some sort of additional feedback -- not so different than a driving instructor teaching a kid to drive.
As robots do more and more things, there will be more and more opportunities for people to train them.
The only scenario where the economics of that doesn't work out is if everyone is employed so they can't afford to pay the trainers enough....but the rest of the article doesn't support that.
You're assuming there is a single perfect translation, or even a single perfect translation algorithm across all domains. I'd wager that translating legalese and translating prose will need completely different algorithms trained on them and in the latter case you won't find two translators that fully agree on the best way to translate a given text.
I know translators who use Google Translate for a first pass and then correct the results. Translators time is limited so anything that allows them to focus more on the nuances of language is a good thing.
This article was fantastically well written and comprehensive. Worth reading if you're interested at all in the topic of automation and its potential effects on the future of employment.
The narrative that machines will "eat" all the jobs leaving the mass of humanity unemployed is an easy to understand story, but it hasn't ever come to pass and probably won't for reasons that don't need to be detailed in this comment. The nobel laureate Robert Solow compared it to worrying about the earth being struck by an asteroid. Possible, even worth consideration, but highly unlikely.
The idea that machines will dominate labor to the point where a few rich people gain all the profits from labor and the rest of us are left jobless and effectively under their control is equally unlikely. A gem from the article:
"Capital isn’t just winning against labour: there’s no contest. If it were a boxing match, the referee would stop the fight."
Income inequality is probably the biggest problem facing our society today. But it can only go so far. If the majority of people don't have jobs, no one will be able to pay for new iphones or driverless car service fees. Even B2B companies depend on B2C customers for revenue at some point in the chain. The basic, common sense economics of the world dictate that there are limits to inequality. The idea that machines will create a truly dystopian scenario is still the realm of science fiction, but that doesn't mean that we shouldn't take steps to address the problem.
Finally, I loved this quote:
"Robert Gordon, an American economist who in 2012 published a provocative and compelling paper called ‘Is US Economic Growth Over?’ in which he contrasted the impact of computing and information technology with the effect of the second industrial revolution, between 1875 and 1900, which brought electric lightbulbs and the electric power station, the internal combustion engine, the telephone, radio, recorded music and cinema."
Worth noting that four of those seven inventions were created by one man. Not relevant to the topic of automation, just truly freaking amazing.
Automation won't destroy jobs because technology and society change faster than automation can keep up. Think of all the job titles that didn't exist 30 years ago (social media marketer, seo specialist, etc). By the time those are automated away, they will be replaced another set of jobs that we can't even conceive of. It's tempting to think that it's different this time, but it probably isn't.
Secondly, the number of jobs and job titles is entirely arbitrary. We often think of the world as though there are a fixed number of jobs and once machines take them all then we're out of luck. But there is no limit or constraint of any kind on what a human can be hired to do. If you've ever worked at a growing startup you'll understand the impact of Parkinson's Law: Work expands to fill the time and resources allotted. In practice, companies tend to use the savings from automation to hire more people. And they should, it usually turns out to be a smart investment in human capital.
Ultimately, very few managers will say "I have enough people working for me and I don't need to hire anyone else". Those words have been uttered maybe 5 times in all of human history, including when I typed them just now. People like hiring, and the numbers just don't support the hypothesis of technology destroying jobs. The US unemployment rate now is similar to what it was in 1920. Despite automation in factories, Ford employs a similar number of people now to what it did in 1970, albeit no longer in factory jobs. If the advance of technology was truly putting people out of work, don't you think it would be evident by now?
Technology has always reduced the total amount of work people need to do. That's why we now work 40-hour work-weeks in service outlets, factories, or offices instead of 100-hour work-weeks on farms.
The difference is that, once upon a time, we redefined "job" and "full-time" to balance between the economy's actual need for labor and humane living standards that allowed for active citizenship. We reduced the work-week and redefined a worker's life as involving more leisure, because our technology allowed us to do that.
These days, unemployment is high and labor participation is falling because technology has advanced, but we still expect the same work time. The result? A bifurcation of the labor market into a broad section of underemployed, unemployed, or just plain low-paid unskilled and semi-skilled workers, and a small cadre of overemployed professionals now working more than they ever previously did.
Stop trying to pretend the problem isn't there, and let's just fix it.
Think of all the job titles that didn't exist 30 years ago (social media marketer, seo specialist, etc
I like that you picked those two because they happen to be two that are already disappearing.
Search engines have been fighting SEO for years. And they have pretty much won. SEO was a hot industry just a few years ago, and is already disappearing.
And social media specialization of marketing and PR was taken on by very young people and thought to be a special skill, until recently. But already businesses have figured out that there is nothing special about it. And every marketer is expected to be a social media expert. It's not creating more new jobs, it's just adding those skills to existing jobs.
It's tempting to think that it's different this time, but it probably isn't.
I am sorry, but to me that argument sounds just like the future will be the same as the past. It's not exactly wrong, it's just an intellectually lazy argument to make.
Just one example of this time it's different is the income split between labor and capital. Historically it had been fixed at 70/30. This is across all industries, and all nations, since the industrial revolution. But now labor's share is down to just 62%. And that's happening in places like China and Mexico too, so it can't be simply due to offshoring. Source: http://www.economist.com/news/finance-and-economics/21588900...
Secondly, the number of jobs and job titles is entirely arbitrary.
That's not at all true. Automation has always created unemployment. There was no such thing as 3rd world before the industrial revolution. The economies of countries like China and India, which realized on a lot of textile production, collapsed when the industrial revolution took off in the UK. The UK exported both goods and unemployment. And what we call the 3rd world was brought into existence. We are seeing a very similar process today with Germany and the Southern belt of the EU. Germany too is exporting both its product and its unemployment.
If the advance of technology was truly putting people out of work, don't you think it would be evident by now?
Yes, I do. But I don't think we'd see the unemployment rate go up. I am sure we'll first see labor's share of income drop, and wages should stagnate. Which is quite similar to what we are indeed seeing right now.
>The narrative that machines will "eat" all the jobs leaving the mass of humanity unemployed is an easy to understand story, but it hasn't ever come to pass and probably won't for reasons that don't need to be detailed in this comment.
It is already happening, just take a look at graphs of labor rate participation against productivity. Will it happen to 'all the jobs', no, but will it leave masses of humanity unemployed? Yeah that is already happening, EU unemployment for those under 30 is staggering.
These articles always exaggerate the disruptive potential of technology, the pace of change and the consequences, particularly ignoring the concept of wealth creation (the reference to Piketty was telling).
For example: I've been to China and Japan countless times in the last decade. On my last trip to both, I finally had Google Translate on my phone. The difference it made in communication - particularly in Japan - was phenomenal. Both ways - both in scanning local characters with the camera and getting an instant translation (you lose the fun of the restaurant roulette, but you no longer end up with mapo tofu by accident), and in communicating with the other side by typing your English meaning and showing the translation. I had a 20 minute conversation near Chengdu with our hotel manager, she would type her side into Translate, wait for the slow internet to send it over, we'd read, reply... we both wrote each other a long thank you letter before leaving!
What is relevant about this example: the job of "translator for tourist" has been created, not automated. Before I had Translate, I made do with sign language and the few words hastily scribbled or learnt before a trip. I didn't hire a local to walk me around and talk to everybody. Conversely, I take the MRT in Singapore because it's 80 cents a trip. If an automated taxi comes along for $2, I might take it, but doing the same trip for $10 is not something I'm interested in. This is added productivity for me, just as the flying shuttle enormously increased the quality of life of hundreds of thousands of housewives in the 18th century, instead of creating the "mass domestic unrest from idleness" allegedly feared by the ruling class.
Another point missing from the conversation is manager laziness. In my experience as a DBA/data person, most companies do not care about automating away their reporting function (amongst other things). I've seen people doing "manual joins" (yes, that's a manual lookup then copy paste each value one by one) as late as last year, in both tiny "modern" tech startups and enormous corporations with several hundred thousand employees; for the latter, 50MB files were "big data". Today, you can run a DWH for a fairly large business with an employee working part time, provided he knows what he's doing - with AWS and modern tools like Postgres, it's incredibly easy to be high level - but few companies do it. Maybe it's today's relatively permissive high capital low opportunity environment, which tolerates very low IRR.
Why it's different this time: computers are so cheap and so general-purpose.
For decades, there have been many jobs that could be automated, but weren't, because the machinery wasn't cost-effective. It might have to be custom-engineered for the job, and if you didn't have the volume, it didn't pay. An automated hamburger outlet was built in the 1960s by AMF. There are machines for almost every picking job in agriculture, but much picking is still done by hand. There are still hand car washes.
Now, if a computer can do it, the computer will be far, far cheaper than a human doing it. The computer can also provide 24/7 operation, and, even better, once one computer knows how to do something, a million computers can be doing it tomorrow. Deployment is very fast in this area.
The list of things humans can do and machines can't keeps getting shorter. That's not going to reverse. But what gets checked off next? More desk jobs.
Actual physical robots are still rather inept. That's getting better, but progress is slow. What makes robots work? Money. For decades, robotics R&D was under 100 people in the US, mostly at CMU, MIT, and Stanford. Then came the DARPA Grand Challenge, when DARPA told the universities to get results or robotics funding would be cut off. Suddenly entire CS departments were devoted to automatic driving. After that success, DARPA tried throwing money at Boston Dynamics. It took about $125 million to get the fieldable version of BigDog working. Now Google is in the game, spending who knows how much.
A key point in robotics, and AI generally, is that there's now enough known that spending money gets results. There was a false dawn in the 1980s; look up the Fifth Generation project and the NASA Flight Telerobotic Servicer, notable failures. This time, though, many of the old ideas work, powered by four or five orders of magnitude more compute power, and lead to new ideas which also work.
Advanced robotics right now is about at the Xerox Alto level - there are impressive prototypes that work, but they're not cost-effective yet. Robotics has not yet had its Apple II or IBM PC. (The Roomba is too dumb. The Dyson robot vacuum, though...) It's going to take a while to break through the cost barrier, even once the smarts are there.
The implication for jobs is that manual labor is, in the near term, less at risk than intellectual labor. If your job is to do something where the inputs and outputs are through phone or computer, be afraid.
Fantastically written article. I found the most interesting question near the end. I've never been a proponent of socialism, but in a world where work can be filled by the push of a key by unquestioning automata, could it become viable?
The torrent of money and talent going in to self-driving cars seems to be only increasing. One of the strong points of this smart and level-headed article is pointing out how disruptive this will be, and how fast the disruption can come. I'm aware that there will be a lot of benefits, but the problem is, they are not distributed evenly. Some people will lose.
"Translate has hoovered up gigantic quantities of parallel texts into its database. A particularly fertile source of these useful things, apparently, is the European Union’s set of official publications, which are translated into all Community languages."
The first thing I thought was "what happens when the EU starts outsourcing its translations to Google Translate?"
Is this the future of machine learning? The learning algorithms start by mining a corpus of human output. Once they get good enough, they replace the majority of humans that generated the corpus. We then enter an echo chamber of machines largely feeding off their own output. Consequently, improvement of the machines stagnates, but the machines are still doing a good enough job to keep humans out of a job. We then have a future of "good enough that the cost of improvements can't be justified, but bad enough to be irritating"?
Humans have a sense of pride in their work, and will strive to improve for their own edification, even when the cost outweighs the benefit, or they have been told not to. A machine will just continue to deliver the level of service that the committee in charge tells it to.
Edit: fixed spelling