This is an article I co-wrote with Dudu Noy last December for a “coming in 2014″ guest post. Dudu is the CMO of Israeli language technology company, Ginger Software, which my company represents in the Japanese market (Ginger Japan website). So far the prediction has proved very prescient.
Ginger has developed a super-smart hybrid technology that combines NLP and statistical algorithms to process all the English that is out there on the web, and using this enormous corpus as a reference, corrects non-native English as the user types.
Like Japan, Hebrew-speaking Israel is an island nation in terms of language, and so they have similar challenges when it comes to communicating in English. They also have one of the most fervent technology development cultures on the planet. Ginger is a product of that context, and I believe is a ground-breaking technology that has the potential to lower the language boundaries between cultures globally through supporting accurate and expressive communication in English.
Ginger’s take on cognition is just one area of this burgeoning field, and the article discusses the bigger picture, as well as explaining more about Ginger’s approach to it.
The article first appeared in one of Japan’s leading tech news sites, The Bridge, in December 2013 in English and Japanese, where naturally, as my client, Dudu got the credit. The main audience was Japanese, so a bit more effort went into the Japanese script, and I have rewritten some of the more awkward phrasing for the English version below.
Cognition-as-a-Service will come of age in 2014
Here at Ginger we are predicting that 2014 will be remembered as the year that CaaS, or “Cognition-as-a-Service” platforms came of age. Cognition is historically a complex biological trait including skills such as decision making, problem solving, learning, reasoning, working memory and not least language, skills that today the computer sciences are chipping away at from various angles.
With each major evolutionary step in computing we have seen over the last 30 years, from mainframes to PCs, the internet, cloud and SaaS, and now ubiquitous smart mobile, the new realm has not so much replaced but augmented what was there before.
In the same way the promise of CaaS is to allow apps and services to function more intelligently and intuitively, allowing you to converse with them, ask questions, give commands and complete tasks more efficiently and conveniently.
Apple’s Siri is one of the most famous cognition-based services in general use today. And now Google’s recent innovations to its search product for mobile, incorporating more contextual conversation for queries, pits Google’s AI technology against Siri in the cognition-augmented search arena. In both cases, most of the technology itself is in the cloud, even though the device is in the user’s hand. Their main functions only work when there is an internet connection.
Natural Language Processing
The reason for this is that the two necessary tricks to make sense of a user’s speech input – speech recognition and natural language processing (NLP) – require cloud-based servers performing intensive processing of proprietary algorithms, and these processes are beyond the capabilities of handheld technology.
NLP-type processing is so intensive because of the sheer diversity and complexity comprised by a language like English. Old school NLP solutions were based on rigid rules that map inputs to a big list of known inputs. But the list can never be long enough, and the hard rules can never cover all the edge cases that appear naturally in language. This is why the experience of talking to a supposedly “smart assistant” has so far always left the user frustrated and feeling misunderstood, since they have up until recently been built using the hard wired rule-based approach.
You need more powerful, agile technologies that can figure out that in a sentence such as: “Yuko wants to eat an apple”, Yuko is something that can have wants, and can eat things, and that apples are things that can be eaten. The technology needs to be able to do this for the vast majority of sentences the app is likely to encounter. This is incredibly hard, but here at Ginger and a few other places, we are doing it.
The “Platform Model”
It is not just Apple and Google who are eyeing this new technology realm. IBM is now also a player with Watson, recently announcing that the same supercomputer-strength software that conquered the quiz show “Jeopardy!”, will be available to app developers through an API and software toolkit. This will allow cognitive apps that leverage cognition to be hosted in the cloud on Watson. This would obviously be a great thing for IBM’s cloud hosting service as well.
This “platform model” in tech business is nothing new of course. In recent years IBM did this with its Websphere application server technology, which went from an internal project to a software community of thousands of developers. Salesforce.com did this with its Force cloud-app development platform, as did Amazon with Amazon Web Services.
But what is different with CaaS platforms is that cognitive powers will be baked in to the operating system, and all the apps that are developed on that platform. That will bring intelligence to a mass public in a wide variety of as yet unimagined contexts.
Gingers Approach to Cognition
At Ginger we have not opened up our technology as a platform via an API yet, but we are already providing the benefits of its cognitive powers to a mass user base globally. Our technology uses statistical algorithms in conjunction with natural language processing, referencing a vast database of trillions of English sentences that have been scoured from the web. This allows us to work out what the users of our applications are trying to communicate, be it in Microsoft Office apps, Gmail, Facebook or anywhere else for that matter, and correct their mistakes and suggest improvements to their expressions.
One thing is for sure – this is a really interesting space to work, and it will be fun to see where computer based cognition will go in 2014.