MT: Every Titan has its Achilles Heel

We’ve debated many times in the past about the “natural” advantages Machine Translation might have over our fellow translators composed of mere “flesh and bone.” Their lives decaying and finite against the colossal golem that threatens to take over the entire industry, eventually rendering us humans obsolete.

Will we become the outdated dusty VCRs of the future? Will the new generations of cyborg children look at us one day as today’s kids look at 8-track cassette cartridges?

I can almost picture their mystified, sad expressions, aimlessly pondering, trying to figure out what us “meat sacks” were ever for. “Error 404” flashing on their “mind screens” as an inconclusive result.

So what have we got going for us that may keep us going against the dreaded machine?

The answer: Improvisation, which is at the very core of our unpredictable nature.

The very human ability to extrapolate two non connected pieces of information by putting them together, achieving a new posible meaning. Such is the gift of creation, in a sense.

That’s why machines can’t be poets (or only dadaistic poets at best).

Yes, they can randomize, even make their own mistakes, but never on purpose. They’re not programed to make mistakes, although they do from time to time, but only while their codes are being perfected (from 0.1’s to 0.2’s), and never with a creative purpose “in mind”.

Improvisation, connecting separate pools of information in order to “understand” is, for now, an exclusive human ability. The machine either knows, or it doesn’t. When it doesn’t, it needs fresh input that includes that consolidated juxtaposition of ideas into a new concept or syntagm. (I’d like to see MT trying to make sense out of a classic soviet cinema montage. Sergei Eisenstein would LOL in his grave so hard).

It’s like trying to have philosophical conversation with Alexa or Siri.

They might have some pretty witty answers for pre-scripted questions, but when you ask them something completely off grid, they just can’t compute and fail to deliver. That’s where we humans still step strong.

Last year in Seoul, a team of human translators defeated an artificial intelligence MT engine in a competition hosted by the International Association of Professional Translators and Interpreters (IAPTI), proving us “meat sacks” are still holding our ground in the business. How did our underdeveloped species manage to achieve such an accomplishment, you ask? Well, it seems MT was unable to make sense out of specific collisions of words it didn’t have any input for in the first place, rendering nonsensical translation results where its human counterparts could deduce meaning through the use of imagination and some risk taking.

So it seems, fallibility, and the ability to question things, to feel free to risk absurdity, might be our God given gifts after all. What makes us unique, and perfect in our own imperfection.