Google’s AI building blocks, platform and solutions discussed at Google AI Master Class
11 Sept 2018
AI as a science has been around for decades. We heard a lot about it and wondered if it is only restricted to data scientists to be able to make full use of AI in helping the world, innovate and solve real world problems which is scalable.
In our last article, we already saw that AI is used medical science in retina screening and has been pretty successful. It is also used in agriculture to identify the good or bad crops by classification.
So, who can really develop such solutions? The data scientists definitely can do it but to understand the nifty gritty details about neural networks or classification might be too steep to comprehend for the million of software developers out there.
Mr. Oyvind Roti, Google gave us a run down of the building blocks and solutions that can help AI available to everyone, making it easy, fast and useful for developers and users to work gain access and build on existing ones without knowing the underlying complexities. Developers can come out with software solutions with API calls which return the results and can be presented for analysis.
To date, there are 14 thousand paying customers like box, GIPHY, Rolls Royce using the Google AI platform.
[nextpage title=”Cloud AI building blocks – Sight, Language and Conversation”]
Cloud AI building blocks – Sight, Language and Conversation
Cocado has personalised customer interactions by using AI to filter mails for urgent cases that require more immediate attention while box, the software that provides free storage space uses Google Cloud AI to enahnce it’s image recognition technology help its customers who need to identify objects in photos and pictures in their online files.
Google has the Cloud AI building blocks which helps developers get into speed with sight, language and conversation tools.
One such example is Google Translate which basically takes in a string of Unicode characters, detects the language and translate it into the target language. The query can be sent as a query string and the result is return as a JSON response to the call. This solution is scalable and it is used in “Clas of Kings” which requires language translation for the global gaming community. VICE, a media company use Google Translate which helps translates 18 languages in real time for writers. This greatly speeds up the editorial content in various language. It could be integrated with speech api to do real time language translators.
An example of Google Translate which can convert websites on the fly. In this case, this web page in English is translated into Chinese on the fly. Click this link to translate.
In Vision, it an recognise faces and is able to a meta data returned in JSON which can be indexed in a database. e.g. if the picture of the person is a professional based on his outfit. It can detect if the person’s mood, is the person happy or violent. Such a system may be used for surveillance.
Likewise, if you have a library of videos, an application can be developed to identify the objects within the videos. This is possible as videos are just multiple frames in motion. Thus, videos can be easily search for a bicycle, a cat within the video. The meta data can then be used as an index for searches or future use.
Google also provide Cloud AutoML which makes it easier for developer to do machine learning for any problem by providing data sets. The higher the model size the higher the accuracy.
Disney uses Cloud AutoML technology to build vision models to annotate their products with Disney characters, product categories and colours. These data are index and used in their search engine and product recommendations on shopDisney.
[nextpage title=”Cloud AI platform”]
Cloud AI platform
Google’s Cloud Machine Learning Engine is used by Airbus Defence & Space to filter out satellite images that are not blocked by clouds. Hermin Textiles uses TensorFlow to classify the textiles. To speed up, Cloud TPU v3 now has a processor performance of 420 teraflops which is 230% faster than it’s second generation Cloud TPU v2.
ebay uses Cloud TPU to make visual search faster and more accurate.
[nextpage title=”Cloud AI solutions – Contact Centre AI”]
Cloud AI solutions
Contact Center AI is a solution that provides call centre support which helps customers. When customers call in, their requests are cataloged and analyzed using Google’s Dialogflow Enterprise technology — for a more intuitive, seamless experience.
It can automatic determine the topic and as Contact Center AI can parse topics based on the conversation, getting the answers customers want, fast. There is no need to go through boring manual Press 1 2 3 for …
Live agent assistance
If a live agent is needed during a call, the Agent Assist AI automatically supplies the agent with articles and knowledge documents based on the conversation. No hold music required. Agent Assist will stay on the call to help the live agent with documents from your enterprise’s knowledge base.
This solution is definitely useful for tech support, banks etc.
In Conclusion, whether you are starting to develop applications or need solutions, Google has the building block and solutions which can integrated into the enterprise readily available.
Discuss