Tuesday, December 27, 2016

The disappearance of Big Data

Quoting from my blog April 2016

No one will be talking Big Data in 24 month (November 2017)

It  is going to get so good, so fast, so in 24 month, I mean literally in less than 24 months, that no one will be talking about Big Data. We will just spit in data like using a fire hose, and then spit out patterns predictions, correlations, causation, that we could never  understood. These technologies are compressing things so hard and our brains are built linearly. That why we can not see it.

We will not bother with Hadoop 3.0. We will use "this thing"

I do believe that the notion of Hadoop 3.0 would  simply be, "we will not even bother with it."  

Yes, indeed. It did already happen, December 2016

Sunday, December 18, 2016

The myth of Google Translate

The article The Great A.I. Awakening is typical of the flamboyant style journalists from New York Times use to create an artificial suspense ex-nihilo
Late one Friday night in early November, Jun Rekimoto, a distinguished professor of human-computer interaction at the University of Tokyo, was online preparing for a lecture when he began to notice some peculiar posts rolling in on social media. Apparently Google Translate, the company’s popular machine-translation service, had suddenly and almost immeasurably improved. Rekimoto visited Translate himself and began to experiment with it. He was astonished. He had to go to sleep, but Translate refused to relax its grip on his imagination.
The highlighted text  is added to manipulates the reader. Profesor Rekimoto, went on and tested a Japanese version of Hemingway's The Snows of Kilimanjaro translated  back into English using Google Translate. The result is decent.

This made me experiment too. As I was born in Romania, I selected the classic writer -  considered impossible to translate,  Ion Luca Caragiale.

Here is the Romanian version of a fragment extracted from the short sory The Visit
După un moment, se-ntoarce cu cheseaua goală; o pune pe o masă, se apropie de mine, îmi ia de pe mescioară tabacherea cu țigarete regale, scoate una, o pune în gură și mă salută militărește, ca orice soldat care cere unui țivil să-i împrumute foc. Eu nu știu ce trebuie să fac. Mama, râzând, îmi face cu ochiul și mă-ndeamnă să servesc pe domnul maior. Întind țigareta mea, militarul o aprinde pe a lui și, fumând, ca orice militar, se plimbă foarte grav de colo până colo. Eu nu-l pot admira îndestul, pe când mama îl scuipă, să nu-l deoache, și îmi zice:
— Scuipă-l, să nu mi-l deochi!
Here is the version from Google Translation
After a time, coming back with empty cheseaua; puts it on a table, get close to me, I take the small table royal Cigarette cigarette, take one, put it in his mouth and a military salute me, like any soldier who asked to borrow a ţivil fire. I do not know what to do. Mother, laughing, winking at me, and I-ndeamnă to serve Mr. Major. I extend my cigarette, lights it on his military and smoking, as any military walking from place to place very seriously. I do not admire îndestul, while mother spat, not to deoache, and I say:
- Spit it, not me him the evil eye!
Although Google did a decent translation, with only four words kept in  original Romanian, there is no sense of style. Some of the words are archaic and suggest a comic perfume the Romanian reader captures and laughs, while the English reader still struggles to make sense of the text and can't understand the reason why the Romanians are laughing.

So no mater how advanced the AI translation engine is and no matter if the Natural Language Understanding has a high rating, they can't translate from culture to culture the  sentiments which make literature a pleasure to read.

In my humble opinion, we can improve language , but we can not replace the people completely. We still need very skilled human translators to take over from AI/ML. The leading companies in applied AI  understand this very well.

The suspense from the New York Times article has been created by a  human,  GIDEON LEWIS-KRAUS,  and not by a machine

Love me some Google translate (Go here for the magic http://translate.google.com/  ):

Monday, December 05, 2016

What makes Cooldimi bot different?

Written by Kuldip S. Pabla, co-founder and CTO of Cooldimi 

What is Cooldimi?

"Cooldimi is a stealth mode startup in based out of Silicon Valley. Cooldimi's bot platform is focused on enabling conversational commerce by using Chatbots. Cooldimi thinks that AI machine learning Chatbots will not replace human beings in the near future.

The  Yale University Professor David Gelenter's  view  is  The Danger is not Machines Becoming Humans, but Humans Becoming Machines

As a result, Cooldimi's unique approach involves creating super intelligent bots by blending artificial intelligence alongside human intelligence, which is irreplaceable in bots technology for the time being.

What is "bite-sized content"?

A bite-sized content is what a bite-sized snack is as compared to a normal sized meal. By analogy, the bite sized content came up with the advent of the millennials generation. They do not like long articles, long videos; they like shorter articles, shorter videos (5 minutes max versus 30 minutes).

One of the customers - a popular science center- shared that they have a website that they continuously update. However, the customer has experienced lower web traffic that has been  going down, though their site is continuously updated and has excellent contents. In my discussion, I suggested a messenger based interface, where the users can ask for smaller, shorter questions, and the responses are smaller, shorter and relevant answers.  The customer would love to use that service. More important for desktops, laptops, tablets you need to open the website. In Cooldimi solution,  all you need is to send a message, ask a question and then you get back the relevant responses. The customer marketing  representative also would like to know whether their marketing campaign are successful or not. In addition they liked this  idea, as is so focused to the point, because youngsters with a limited attention span would like to use messengers rather that browse entire websites looking  for a specific answer.

They want to understand what is the real meaning  of the  questions people have (Gap Analysis) - What people really mean? What questions they should really ask, and they don’t?

Why a Cooldimi bot is different?

Most of the bots - according to some sources they estimate close a million bots floating the Internet - are rule based. I would say between 80% to 90%  bots are rule based. They behave exactly how they are programmed to. They cannot learn and cannot respond to anything they do not know.

The next generation of bots is when a user can communicate in natural language, the way speak. This is also referred to as Natural User Interface (NUI), like GUI. So if I say for example: What is the price of apples? or How much the apples cost? , while the intention is the same, different users may ask it in many different ways. Using Natural Language Understanding (NLU), the second generation bots can understand intentions and can answer the queries even if they are not programmed to answer so. These bots can also be continuously trained to augment their smartness.

That said, these bots cannot do what a human can do - understand emotions and sentiments. Humans in a face to face interaction, can see the gestures, voice inflections, the body language. One can see whether the person I am talking to is happy or not, maybe she or he is frustrated or enthusiastic . These are important cues in sales interaction.

But when  talking to a broad audience of humans, how would an inanimate bot fine tune these conversations? It has to be done through sentiment analysis, and this is how a bot from cooldimi stands out. Not just we analyze sentiments, we use the cues to improve bot behavior to make it smarter.

Blog Archive

About Me

My photo

AI and ML for Conversational Economy