Oh my Google, what have we done?

Posted    |    Comments 0    |    Rating: not yet rated

Captcha if you can…

The news roll has finally addressed a nuisance that we’ve all surely come across. Captcha tests, the “I am not a robot,” tests, have become embarrassingly hard. ‘Select all the traffic lights,’ says one test and I, among others, find myself really scrutinizing little, tiny thumbnail images of blurry portions of blurry scenes. A few might even have traffic lights. I have to guess.

Years ago, Google bought the Captcha stuff. It was about ten years ago as of this writing when Google acquired a small company with a new product, reCAPTCHA – Completely Automated Public Turing test to tell Computers and Humans Apart. The joy of this was how difficult it would make things for robots to spam online comment forms since text-readers couldn’t handle the warped images of letters and in some cases, those warped letters were from old books and documents Google’s own OCR technology couldn’t read. So, as the WIKI put it then:

The process usually involves one computer (a server) asking a user to complete a simple test which the computer is able to generate and grade. Because other computers are unable to solve the CAPTCHA, any user entering a correct solution is presumed to be human.

Optical Character Recognition today

“Point, tap, brush and listen,” instructs Google to people using Google translate on their Android phones. Decypher that foreign menu – point and tap to take a picture of it. The app highlights what the OCR recognizes as text and whichever you brush your finger tip across, one of Google’s voices will read aloud (and the English translation will print on the device). Pretty hot stuff for 2012!

Fast forward six years to Summer 2018 and Google’s unassuming Translate Blog lets everyone know that then two-year old technology, neural machine translation (NMT), was moving offline to live on Android phones. NMT deciphers entire sentences rather than translating each word directly. Google’s wish to create the Babel Fish is pretty much complete in less than a decade.

Pixel 3 phone

Out last week as of this writing is the new Google phone, Pixel 3. It’s a big step up from the other Pixel phones and reminds me that I’m wrong every time I say Google is a software-only company. The phone is made by Hon Hai Precision Industry Co., Ltd. Foxconn. Ouch. So the off again, on again jobs in Wisconsin company is guaranteed some USA business. I don’t have any Pixel because of my personal $400 dollar self-imposed poverty limit on phone purchases. My LG was so cheap, the provider gave me a dollar after I signed up after calculating incentives. :)

Anyhow… The Pixel is slick. Just about everything it boasts in YouTube spots and commercials comes down to image recognition through its front and rear-facing cameras. The phone’s slogan? “A new way to see the world.” I swear, I’m not making this up. Granted, Google’s browser is spell-checking this post right now as I type it, catching fake compound words like everytime [sic] and my really enthusiastic use of apostrophe’s [sic], and here and their, my grammar. For this snooper-vising, as with image recognition, we have ourselves to thank.

The CAPTCHA. We hate being watched but it not only benefit’s us, we taught it. We taught systems like these NMT systems the baselines of how we communicate while expecting we would be keeping robots out of our chats, website comments and communications in general. Now the A.I. enhanced OCR technology is so fancy, an ‘image’ of it can run offline on a phone for about 50MB (according to the Google Translate blog).

So is that a red light?

Back to the tougher CAPTCHA tests like the traffic light question – It’s not too peculiar if you’ve never seen what I’m talking about. You have to be out and about online, signing up for new services and actively creating a presence. If you only ever visit the same dozen or so funny and news and food websites, you’re not likely to have encountered these crazy-hard picture turning tests.

I’ve ID’d license plate digits, traffic lights vs brake lights, store fronts vs house awnings and painted vans. I’ve been training the A.I. unfortunately and the take away, as these turing tests get nuttier and tougher, is teaching the server giving the questions what the limits of my observation are. And meanwhile, in A.I. news, chipmakers like Nvidia are taking devastating financial hits like Softbank pulling over 3 billion dollars of support days after TheStreet considers Nvidia a “hot stock with upcoming earnings.”

The A.I. revolution has started an arms race for processors states the title of a July 2018 Ars Technica article. And we’ve provided the bullets with CAPTCHA. Just this year, a new kind of figuring for A.I. It’s the power to infer; the power of guessing.

LINKS:
Google acquires reCAPTCHA in two-for-one deal
Google Translate Blog: Point, tap, brush and listen
The AI revolution has spawned a new chips arms race
Intel To Rival NVIDIA In The ML Market With Its Latest AI Chip

Author
Categories ,

Comments

There are currently no comments on this article.

Comment

this -- Enter your comment below. Fields marked * are required. You must preview your comment before submitting it.