This month’s hot events and news from the data, analytics and computing world.
This month AWS (Amazon Web Services) launched its cloud-based business analytics solution, Quicksight for the public after running it in the preview mode for more than a year. It lets employees quickly analyze data, create visualisations and reports. According to AWS, the point of differentiation is the ability to run ad-hoc analysis which is usually not available with traditional BI solutions. Just like any other cloud solution it also takes away the hardware and software maintenance cost along with the cost associated with database scaling. While the preview edition had 1500 customers, the official website features Hotelbeds, Infor, and MLB Advanced Media as some of the present customers.
It connects with AWS data sources like RDS, Aurora, Redshift and S3. Data can also be imported via spreadsheets, flat files and on-premise databases like SQL Server, MySQL and PostgreSQL. Apart from all these options, third party SaaS integration includes Salesforce.
Pricing starts from $9 per user per month for the standard edition and $18 for the enterprise edition with annual commitment. Free trier is available for a single user account. According to Gartner, with this kind of pricing (competitive with Power BI and substantially lower from others) Amazon can make a mark in the $15.2 billion BI and analytics platform market.
Google has introduced Zero Shot Translation – an artificial intelligence system which can train itself to provide translation between languages for which it has not been trained previously. This is powered by GNMT, which is a neural network introduced in September. Neural networks are a type of computer system that act like human brain and develop the ability to learn from past actions to solve new kind of problems without any specific programming by humans. This means Google Translate can now translate, Greek into Mandarin without even receiving any training data on this particular pair of languages.
Google Translate team had previously announced that it had developed the capability to read the whole sentence instead of breaking it down to several phrases. The GNMT system helped them add more cohesive, human-like and natural elements to the translator. But this system could only work on the languages that were tested and majority of the 103 languages supported by Google Translate couldn’t benefit. Zero Shot Translation came into existence to solve this particular problem.
Here is how it works:
If the GNMT has been trained to translate English to Korean and Korean to Mandarin, then all these languages will be used to form different pairs and share the translation with each other. This will create clusters of phrases with similar meaning taken from different languages. Thus the system would be able to use the modelling power to translate English to Mandarin.
Back in April, 2016 Facebook had launched its messenger platform and allowed companies to build bots on top of it. Now there are more than 34,000 Messenger bots on the market handling wide range of activities — shopping, entertainment, customer service and more. This month Facebook launched an analytics tool that will help bot developers monitor consumer interaction and the same can also be matched with the person’s Facebook profile data to derive deep insights. The Messenger bot analytics platform will help the makers visualise the complete customer journey by integrating data from the web, mobile app and the bot.
Here is an example: Messenger bot analytics can be used to track people who unsubscribed from messages and create audience segment based on their Facebook profile data — like age, education, job title, gender, location, purchasing cycle and whether they have liked the official Facebook page of the business. All these can be used to analyse the type of message the user received before opting out of message. Facebook is also collating and anonymizing the user data in order to restrict bot makers from viewing specific action performed by the individual user. Bot makers will have the option to track different types of messages via the analytics tools and using that data they will be able to figure if certain type of call-to-action is working well for particular type of user. This will help them build customer segments and improve the relevance of the message that needs to be sent to the users.
When the smartphone wave hit the world, Intel was caught unaware. The company which is known for powering servers, laptops and desktops simply couldn’t compete with SoC makers like Qualcomm and Samsung. Now the demand for hardware to support Artificial Intelligence projects is increasing and they don’t want to miss the opportunity.
Intel has laid out their AI focused strategy based on the new technology they’ve gained from the acquisition of Nervana Systems. The new portfolio will span across products and services ranging from fog computing to usage in data centers for an accelerated growth in AI space. According to CEO Brian Krzanich, Nervana’s breakthrough innovation will be applied precisely to neural networks to produce the highest performance for deep learning and generation of superior computing power for model parallelism (splitting the model among GPUs and using the same data for each model) by providing high-bandwidth connection. He also said that Nervana’s technologies will increase the performance by 100 times in the next three years when it comes to training complex neural networks, helping data scientists rapidly solve biggest AI challenges. To supplement the development of AI product and make training and tools accessible to software developers, Intel has also announced the launch of the Intel Nervana AI Academy.