Google has mastered search. It has dominated the field in this area, but the company is making its way in cloud computing as well. Another area it seeks to lead? Artificial Intelligence (AI). Technology companies are selling AI as part of their cloud services, and profiting as a result. The capacity of data centers is sufficient enough to support it.
A technology prowess, Google has experience with the cloud. It is now looking to serve external customers in new ways, rather than focus all the attention on its internal operations. Customers are not only the end user consumers one sees every day on the street. Competitors are big names such as Microsoft and Amazon, as the company aims to draw in customers such as Netflix and Spotify.
Why does Google see its AI solution as the one to beat out of other business services? The answer may be in the techniques and capabilities the company is investing in. Machine learning is one of them, and has been an interest for a long time. Others include image recognition and also search and video recommendations.
Bridging the Gap Between Machines and English
- SyntaxNet: One of many AI software systems, SyntaxNet enables machines to break sentences into the basic parts of speech. It can decode the relationships between words and phrases.
- Parsey McParseface: The program can analyze relevant words, and enables complementary programs to use this function in searches and other related tasks.
SyntaxNet has been open sourced, allowing for much broader and more flexible development of the potential capabilities and applications. The system is capable of grammatical logic. It can tell the difference between nouns, verbs, and subjects vs. objects. All AI projects are run from the TensorFlow software engine, which SyntaxNet runs on top of. The development of the concept of translating language to machine operations is therefore speeding up. Open sourcing the project means more pairs of eyes are examining the complexities of translating human language and acting upon it.
The research community is being encouraged to work on it. Google has also said its language processing system, Parsey McParseface, is close to 94 percent accurate in relating words to sentences, based on testing that involved reading sentences from Newswire stories. Humans, by comparison, are 96-97 percent accurate. An increasing level of accuracy is one of the ways it hopes to beat out the competition, which includes Siri, Amazons Echo, and startups such as Viv.
An Advanced AI for Development
Its also widely available for developers, who can use it in their code. The Google AI system can be easily integrated into apps. Users can access it on their phone, via Google Now or Facebook Messenger, for example. People are therefore going to be very aware of the ability of the software to understand English and provide an accurate response.
One of SyntaxNets advantages is its flexibility. A neural network type functioning enables it to read a sentence, and create hypotheses of its variants. This enables the software to pick the most likely solution based on the context by ranking each hypothesis. The most highly ranked one is what drives the ultimate response the user hears.
Google has bigger ambitions. A spokesperson said its looking to not just enable the system to understand context in any language, but also learn world knowledge. The accuracy, flexibility, and functions of its rapidly advancing AI system is proving that Google can continue to outdo the competition. The market for AIs and personal assistants is heating up. A higher level of response based on evolving algorithms can keep the company on top. The open source SyntaxNet is already proving to be a desirable option for developers on many different fronts, from converged infrastructure to the cloud.