What's Lost (And Found) In Machine Translation

New videos DAILY: bigth.ink
Join Big Think Edge for exclusive video lessons from top thinkers and doers: bigth.ink/Edge
----------------------------------------------------------------------------------
Social media upended traditional media by cutting out the middle man. No longer were "gatekeepers" needed to vet popular opinion: now, any Tom, Dick, or Harrietta can go online and write what they think. But traditional media-and some could argue, certain political parties that rhyme with Schmemocrat-seemed to have stopped listening to what their their potential readers are saying, creating a gap between the two. Michael Slaby worked with Obama on two election campaigns to figure out what groups of people are thinking. It's not easy work. But he posits that with the right kind of ears, some publications and brands (and, yes, even politicians) could one day make themselves 'of the people' once more.
----------------------------------------------------------------------------------
MICHAEL SLABY
Michael Slaby is a global leader in digital and social media strategy, technology and data analytics, and explores how together they can elevate mission-driven organizations. For Michael, it is not only about developing the necessary technological platform, but understanding what it is that brings people together to take action online, as well as offline. Currently, he is Managing Partner of Timshel-a new company working to help solve social, civic, and humanitarian problems via better technology, engagement capabilities development, and creative capital. Previously he was a Fellow at Shorenstein Center at the Harvard Kennedy School of Government.
In 2012, when the Obama administration began gearing up for re-election, one of the first calls went to Michael Slaby, who was chief technology officer in 2008-when the historical campaign leveraged the internet and social media to raise funds and organize volunteers in ways that had never been seen before.
Michael helped lead Obama for America as Chief Integration and Innovation Officer in 2012, overseeing the CTO, CIO, and CAO, in order to ensure effective implementation and integration of technology across the entire campaign.
----------------------------------------------------------------------------------
TRANSCRIPT:
Michael Slaby: Over the last decade or so I think the biggest shift we've seen is sort of the final breakdown of the traditional channel-based structure, the way we tend to think about communications as paid media, earned media, owned media, this sort of traditional silo-ed approach to communications that we still hear a lot from marketing textbooks in business school and that kind of thing.
I think what has happened is: the process of the 20 years preceding the last decade was about fragmentation, new types of channels, satellite TV and mobile networks, and just this incredible proliferation of types of content and channels within the silos that left people feeling like they had infinite choice but left communicators feeling like it was impossible to reach an audience.
What social media has done, particularly Facebook and Twitter, have created a glue to knit all these fragments back together into something that feels like one big graph, one big network of content moving between channels in unpredictable ways of engaging with people in ways that we can't necessarily predict, of creating more two directional conversational dialogues and communication between individuals and the people that we're trying to reach and inspire, which is new behavior for us.
It requires marketers and communicators and publishers to develop new skills like listening that we didn't used to have to do, we just picked the channel and we picked the right message and we said something to an audience that was mostly pretty passive, and I think this is the big shift in thinking and the real challenge for a lot of organizations is we are now part of a graph with the people that we're trying to inspire rather than them being a stable audience that we're trying to reach and us being in a stable position as inspirer or publisher, we now have to participate in this system.
People create content, we create content, they share content, we share content and that means that we have to think differently about how we communicate, how we tell stories, how much content we have to create. We can't reliably predict where that content is going to get consumed so it's really easy to get into a situation where if we're not really clear about our values and our identity and who we are and what we're trying to achieve where sort of our values and mission sit as an organization it can become really easy to sound schizophrenic to the communities that we'...
For the full transcript, check out bigthink.com/videos/want-to-s...

[@mchandler2112]

This is an amazing video for not only physics and modern electronic translation, but also philosophy. "The way aeroplanes fly and the way birds fly don't resemble one another at all. That doesn't matter. What you want is the flight." That is a profound statement. Truly.

[@Freigeist20789]

thank you for focusing on the really important things!

[@Alpinex105]

See the problem with language is that it isn't static. We create new words, phrases that reflect our changing/evolving society. We're also very creative when it comes to language. The computer could estimate, but what happens with new creations? That would make it operational only for a certain time period.

[@Truthiness231]

*notes that two out of the three people in this conversation are software engineers* "How could they adjust to an established human social construct better?" Because it just needs to take in new info and/or be recalibrated? For example, google's search engine itself recognizes "recalibrated", whereas the Chrome browser (made by google but clearly with a smaller word list) doesn't.

[@Harabeck]

There are programs that claim to be able to do exactly that, but their accuracy is questionable.

[@goose1077]

I once spent time with a girl who barley spoke English. She was a friend's girl friend. She had a text translator on her phone and would type into it to get words to say. There was also a lot of gesturing and showing.

[@aluisious]

If you showed someone a smart phone with translation software 100 years ago, they'd have called it a miracle. Hell there are people alive today for whom many technologies now would have been simply miraculous when they were growing up. Can you imagine explaining GPS to someone in 1920?

[@seag0d123]

The closed caption youtube option is strangely accurate for this video......

[@BogdanA74]

I agree with your objection about "miracle". What a computer can do nowadays might seem like nothing short of a miracle to a 19th century scientist or a 20th century non-scientist, however for a linguist it shouldn't seem too miraculous the fact that a computer excels at organizing and finding patterns and processing raw data.

[@Bender4411]

I agree some people use the word "Miracle" too much. A Miracle is something that cannot be explained and we can explain how computers do what they do.

[@Ttoxcs]

Add to that, that it would require the person to be knowingly lying. Politicians who think what they are saying is the truth would not be filtered out, even if we could get it to work.

[@irchristo]

hahaha, I know what you mean. Except that I was really interested until he finished. Then I realized he had said nothing and I felt punked.

[@Alpinex105]

I respect your profession, but keep in mind that there are huge social elements we're missing here. Software engineers are actually very knowledgable in these respected fields and acknowledge that these tools need constant reinventing. I think we both need to make our arguments clear. What does computer sophistication even mean anyways when it doesn't take part of the most fundamental language processes?

[@Frogmood]

That depends on how you define miracle. A miracle can be defined as an amazing product or achievement. By that definition, the fact that we have created machines with ability to understand and reproduce human speech is certainly a miracle.

[@tyhoying]

anyone else turn on captions for this video just to follow along with the machine's interpretation of his speech?

[@zxul767]

What some people posting here seem to forget, or even realize, is where human sophistication comes from, or where all the complexity we observe in the universe comes from, for that matter. It's a common anthropomorphical view which is further supported when you also happen to believe in Creationism.

[@xanokothe]

"Any sufficiently advanced technology is indistinguishable from magic." Arthur Clarke

[@Run.Ran.Run1]

Miracle? Did I hear him use the word 'miracle' in explaining technology?

[@Run.Ran.Run1]

I agree, and it doesn't ever belong in discussions about science or technology.

[@dlbattle100]

Computer scientist here :D. In order for machines to do a better job at translation they will have to actually understand what is being said, and have a model in their little electronic brains of the part of the universe that is being talked about.

[@Zekian]

You assume that the behavior of a computer program also remains static. I am purely talking about the advances in how we process and search large data sets will affect how we approach a task like translation. While the algorithms used may not change, the data can. You are right, we are certainly creative in our use of language and a computer will often lack context, as a result I believe it would 'read' very literally. However I think the biggest challenge would be the ambiguity in languages.

[@ninjaspion]

I completely agree! Awesome video

[@Mattio_]

It's clearly possible we can have translations as perfect as a human translator, probably more efficient by calculating it faster. If we look at grammar, words & sentance meanings. They all connect through a logical order. This text right now is all logically placed in an understandable order to make sense. If there's logic, there is maths. If there's maths, computers can do it. & the same logic applies to all languages. It's not like we need to imagine a scenario in order to communicate.

[@Frogmood]

In common speech, though, the word miracle is almost never used literally or to the exact definition.

[@EoverMC2]

I don't know how you would describe it otherwise? Considering the black or white discernment capabilities of computers. Undoubtedly "Miracle is a strong word but considering the Herculean task described I am convinced it is an appropriate use of the word.

[@ninjaspion]

David Bellos looks like Carl Fredericksen from the movie UP!

[@thelordopie]

The complexity in the universe comes from the infinitely complex interactions of matter under the laws of physics. Physics can get pretty weird on such a scale, but it is possible to describe the formation of every object in the universe. Human sophistication likely exists because it's beneficial to humans when others like you and there was a point where human inelegance was more important than strength. It's a common fact-based view which is further supported when you recognize science.

[@vrd333]

Computational Intelligence plus Machine Translation will put all translators out of the job for sure..!! Even though, natural face to face communication will never be replaced by a machine..!

[@theedgeofoblivious]

I think it will be interesting when machines are capable of reading human thoughts and conveying them to impulses in others' brains without the use of verbal or written language. I expect that within the next few decades that will be commonplace, something akin to telepathy. Written and spoken language will barely exist as a barrier to communication. It will likely still exist for entertainment and for reading older materials, but communicating and learning in this way could be instantaneous.

[@TheStarman70]

I agree with this one. We humans will evolve much faster then computers ever could. What we humans seem to forget is that a computer is a gigantic light switch. It can not think, It could be told to preform a program that fools some people into concluding that it thinks. But It ultimately does not. I can make a gigantic 64 bit processor in Minecraft, using nothing more then dirt, red-stone torches and red-stone dust. No matter how large you build that processor. It can only ever be a calculator

[@Zekian]

Due to such a large volume of data being available, mainly due to the internet, a lot of big companies (Investment banks etc) are investing a lot of research in analyzing very large sets of data. A lot of this data is text and as a result there is a desire for computers to get an estimation of the meaning of text. Applying the innovation in this area along with some of the advances in more complex searching of large data sets, I expect we will soon see approaches to get better translations.

[@sivadfa]

I agree, also if you chose the Oxford Dictionary which is considered by most as the most reliable. It will also have definitions that don't require the idea of devine intervention.

[@Anarchy4Angels]

Just one more step towards making Star Trek a reality... we have the communicators, we're developing the universal translator, when are we going to get the transporters and spaceships?

[@Alpinex105]

Well to be clear the new data input to accommodate changes makes it static in comparison. If we look at it as a process than you're totally right. Computers have made our lives a hell of a lot easier. They could do a way better job at analysis large data, but when it comes to innovating or new ideas/language it's impossible. That's because it is not a part of the social framework (well at least in this century)

[@aluisious]

I'm pretty sure it would squelch the word "news."

[@SatanistSin]

I can't help but tune this guy out.

[@MrGnometron]

In the thumbnail, he is pointing to the Bigthink logo, hehe

[@IG88films]

"Oh, machine language. How interesting, and not even a little bit controversial. I wonder what the comments look like." ...

[@l3lackcloud]

At that point there would only be machines between them and not, say, people.

[@eurohim]

"Miracle" is a strong word.

[@josephk87171]

Anyone else have the impulse to press the "cc" button to this video? Works very well in this case actually

[@stuntyannick2]

It is possible, but very unlikely, because the people programming the machines would have to be able to spot such things.

[@spartascapefuntimes]

Gary Johnson 2012 ---spread the word---

[@Bender4411]

This makes me miss Microsoft SAM on Windows XP

[@Alpinex105]

See the way language evolves is interesting. Algorithms have to change to accommodate the "new". Now when you consider this process as a whole, the computer has to play catch up. Until the computer becomes a part of our social construct, it's impossible. Also this reminds me of this ( watch?feature=player_embedded&v=ltelQ3iKybU ) Keep in mind that overtime English will be an entirely new language.

[@Truthiness231]

I'd think a computer could adjust to changes in language faster than a human can; modern translators already have to account for multiple definitions. Ultimately (and sooner than one might think), computers will beat us at this and any other non-creative feat, and really all creativity needs is a slight malfunction in a computer combining two unlike things and seeing if it can come up with a new concept, so I figure it won't be long before a proxy of a computer will be doing that eventually too.

[@archangel1996j]

@RobertjBrown88 umm do you have any idea what singularity is ?

[@fredsk8x]

i think ,if your raise by believer this word is incorporate in your vocabulary and take another meaning ,like unexpected ,lucky or unbeleiveble . i'm pretty sure he doesnt mean it litteraly. (please someone re-structure my point ,english is not my native language) thank you

[@Zekian]

As a computer programmer and someone who studies computer science, I disagree. Perhaps I have my head in the clouds and I'm optimistic, But I believe we will see a very different approach to how computers translate text in the near future.

[@ericpa06]

People are always saying things like "machines will never do X, as good as a human would". And.. through history we've been proved to be wrong over and over again.

[@Alpinex105]

Why?

[@GalryZ]

Damn you!

[@falsehope8949]

Is he referring only to statistical machine translation?

[@MrVivstanshall]

I saw an article in which they are begining to use "virtual customs agents". They have already installed one on the Mexican border. As you pass through you simply answer a few questions at a computer and it can tell if you're lying. Apparently it's much more accurate than humans.

[@Frogmood]

Plus, that is only one of the definitions of the word.

[@TheMrymerej]

its a metaphor, all there is is metaphor

[@andrius0592]

It won't work if the person believes his own lies.

[@KoboldCommando]

Came expecting a discussion of programming languages, compilers and machine code. Left kinda meh.

[@ciknay547]

For me, it gives it a more authentic, realistic view on the subject. This guys is rehearsing off memory, or a general idea of what he is talking about, and relying on his knowledge to affirm his point. He may not be the best speaker, but you can tell you know what he is talking about.

[@7StringSeven]

This is never going to happen... It is impossible... Due to individuality.

[@Run.Ran.Run1]

It doesn't belong in this context, particularly when talking about machine translations that pick the most inaccurate words possible. A talk about exactitude should not contain loaded words.

[@ciknay547]

I suppose its down to personal preference.

[@101Xemnas101]

Anyone else think this was more of a "medium think" topic?

[@RobertjBrown88]

It'll happen with the singularity.

[@WhitentonMike]

What this guy is saying isn't of much use. It turns out that people don't all think the same way because our internal representations are different. Richard Feynman talked about this in one of his famous talks. Starts at 0:48. watch?v=lr8sVailoLw So what if computers don't think like people. It doesn't mean they won't. It also doesn't mean they won't translate better than people do.

[@wQQter626]

Automatically...AUTOMAGICALLY*

[@BogdanA74]

Engineering applies science where science must solve practical, everyday problems. Science solves all the problems mankind can formulate, no matter how practical they are. Therefore, science is always ahead of engineering, but engineering is extremely needed to bridge between science and useability of science. As for machine translation, the online translators cheats. To translate means primarily to understand, and a machine cannot understand... and I don't think it will ever.

[@HunterRodrigez]

i honestly wonder what a person like you is doing on a channel like that

[@Aizacc84]

¬_¬... I was offended by the word "Miracle". There are no miracles in any machine.

[@HelpmelamEnglish]

I feel that it is you that has an issue here and not Mr Bellos.

[@VellianoRosso]

You sir clearly do know, so fill me in. By the way why is it that ignorant people on youtube always make empty statements from authority.

You are making the assumption that Fox doesn't lie about what's coming next as well.

[@Skolotoi]

Would you trust a machine like that?

[@joaoqftito]

sorry, it's just so hard to control yourself when the opportunity presents itself. I usually fart in the general direction of the people who do that, but now i understand them.

[@Alpinex105]

In a way it's organic.

[@Truthiness231]

Fox News isn't lying, they've reassure their viewers that by stating that they're fair and balanced before every show.

[@Magneira]

Well if you showed this to any of the people that "created" divine in the first place, like stephen king speaking though one of these machines, they wold call it a miracle, and yes sir it is a miracle.

[@mariwanmt4639]

My major is Machine Translation. Is there anyone who has the same field?

[@JoshuaMathias31]

+Mariwan MT That exists as a major? I don't know of any machine translation degree. Computational linguistics is what I'm planning on studying.

[@MumblingMickey]

well technology until rather recently used deterministic methods of evaluating text/speech. But the latest tts and srs uses non deterministic AI. So where he's accurate in some things he stated there historically, thats not how machinee intelligence works any more. Although it is still different to how a brain works.

[@z4k4z]

"Miracle" is the wrong word.

[@irchristo]

Miracles -- If we draw an imaginary line between science-engineering and miracle, we will see that line perpetually pushed back throughout the history of man. Miracle will always be encroached upon by science-engineering.

[@goose1077]

CNN and MSN too

[@GalryZ]

I believe I have commented on this fine video before anyone else has had the privilege. First like a Sir.

[@loopuleasa]

For those of you who are butthurt from hearing the word "miracle", just replace it in your head with "miracle of technology" and stop being immature.

[@Pastor_virtual_Robson]

muito atual...em 2020

[@agemoz]

Engineering is science... Also, the flight of an airplane does reveal key features as to how birds plane. Bellos way of thinking is a bit outdated.

[@TheMohawkNinja]

Yeah, I doubt we will get any computers to infer puns and cultural references anytime soon.

[@GameplayTwist]

The war with robots has already begun. All shooter video games are made or translated by humans and they program code which defy's reality for hit detection. So when 1 player kills another and the other can't do the same due to the coding of the developers; is that not a war against robots. Simply a war on coding. It already occurs.

[@GuillePuerto]

All I'm hearing is Toupee, toupee toupee... Toupee

[@tombradford7035]

You often get fixated by male hair....

[@DaBrainFarts]

holy crap. there are 4 haters. how can you hate on truth?

[@aluisious]

Computers are not sophisticated. They are very simple, and they do very simple things very fast. That's it. Humans are sophisticated. The moment machines catch up to us in sophistication, though, we will be relegated to zoos if lucky, and museums if not.

[@Alpinex105]

Sorry if my replies are similar, I just saw your comment. KZbin says I got one comment .

[@wofebofe]

??? only 303 views and 188 likes??? oh wait....

[@Alpinex105]

The key word was creation. How could they adjust to an established human social construct better? You have to be human. Also the definitions are not always exact or accurate. Many words me and you use or will use are not recognized officially so how do we takle this? The language can also be specific to a certain region or group. It's easier for humans and computers to work with each other rather than against.

[@boanergesza11]

I disagee with a comment made. I dont think its engineering vs science. I believe engineering bridges the gap between art and science. Thouhts?

[@mathor]

so much time and so little said

[@zoriusth]

Attention, all yoctograms! You fractals have no future! QED. Class is dismissed!

[@KoboldCommando]

Coming up next on Fox News! (dead air)

[@VellianoRosso]

Human thought is not that complex, all we need are potential verbs and differentiation between words, that are close in meaning. And your argument that everyone would interact like computers is irrelevant. Have you ever seen Mathematicians solve a problem, there is not just one route. The goal is to interchange information, as optimal as possible. What u are arguing for is street language, which will still exist.

[@VellianoRosso]

I can't tell if you are being sarcastic. xd