In early June, OpenText told CMSWire its artificial intelligence platform Magellan would be ready by July. Today it made good on its promise with the announcement that the platform was generally available.
The announcement came during the Waterloo, Ontario-based company's Enterprise World conference, taking place through July 13 in Toronto.
Magellan: Logical Next Step in OpenText Strategy
While Magellan is a natural step in the evolution of OpenText’s enterprise information management strategy, according to Muhi Majzoub, EVP of engineering at OpenText.
More to the point, Majzoub told CMSWire, it also sticks it to IBM Watson by offering enterprises a product that OpenText believes is easier to use, more accessible and better placed to give it traction in enterprises by offering flexible AI and big data management.
Majzoub's words echo those of OpenText CEO and CTO Mark Barrenechea, who almost two years ago to the day took the stage at Enterprise World 2016 to talk up Magellan and talk smack about IBM Watson.
OpenText focused on three specific points with this release, but Majzoub noted this is only the starting point and to expect more development in the coming months and years. The three focal points for this Magellan release are:
- Open: Built on an open platform to take advantage of the flexibility, extensibility and diversity of an open product stack while maintaining full ownership of your data and algorithms
- Affordable: A lower price point opens the playing field to allow smaller businesses to use AI
- Cohesive: Magellan is an integrated platform, bundling technologies for advanced analytics, machine learning, data modeling and preparation, and enterprise-grade BI into a single infrastructure.
The history of Magellan goes back seven years to when OpenText bought Montréal-based Nstein. That purchase gave OpenText semantic analysis capabilities, which allow information to be easily found, packaged and sent to the right internal and external audiences.
“It also brought content analytics with contextual analysis and brought to us a series of crawlers. These crawlers can attach themselves to three or four different kinds of content. The first is structured content sitting in a relational database as part of a CRM or ERP application,” Majzoub said.
“It can also attach itself to social networks, to public websites, to communities, or a blog and enables users extract content from these areas. It can attach itself to any platform that has unstructured data.”
The crawlers leave the data in place but extract the metadata intelligence out of these original systems or data sources. OpenText has also modernized its technology to work on top of a Hadoop Spark Apache cluster to provide the scalability a cognitive engine or a AI system will need to grow over years.
According to Majzoub, the combination of Apache Spark with OpenText Analytics gives Magellan the ability to integrate across any EIM architecture to reduce the time, effort and expertise required to use advanced analytics in decision making and task automation.
Competing With Watson
Leaving aside the technical aspects of the release, one item that will definitely attract attention is the claim that Magellan offers enterprises the computing power of IBM Watson, but easier and cheaper.
“We play in the same area as Watson but we differentiate favorably in several areas,” Majzoub said.
“Watson runs on a mainframe and we run on standard X86 hardware architecture. You can buy it as an appliance from OpenText, or you can buy if from any other hardware buyer."
“Watson depends on mainframe programming to build algorithms, we run on standard sets of Python, R, Scala and others. The skillset of mainframe versus these standard tools are easier to find across the industry. This means that it is easier for us to support Magellan."
That is not the only issue that Magellan aims to tackle. According to Majzoub, there is a significant intellectual property problem.
“Today if you build an algorithm on Watson you don’t own the IP for the work you are doing on Watson. IBM will claim the ownership of that IP, or the Watson division will claim the ownership,” Majzoub said.
“We believe that knowledge is the customer’s knowledge. That means they own the IP of anything they develop. OpenText will only own the IP of any of the Magellan tools we provide to customers and then there are some open components — like the infrastructure, like the Hadoop Spark Apache, like the Python, like R — and those are open and available.
“There are many communities that support algorithm writing and the whole idea of the openness is the ability to cut and paste the algorithm that you picked up from a community and to cuts and past that into your Magellan notebook and to experiment using these on your data.”
In practice, Magellan will allow customers to apply predictive analytics to EIM. If the product itself is new, the idea is not. Majzoub pointed out OpenText has always tried to give its customers the ability to managed unstructured data and then bridge the gap between structured and unstructured data. This is just the latest phase.
Magellan will be deployable in three different models:
- As a managed service offered from OpenText’s data center
- Installed and offered locally from the enterprise’s data center
- Buy both hardware and software from OpenText and it comes ready-to-go
Over the next 12 months OpenText will continue working on Magellan by making the tools easier to use and making it easy to build and use algorithms. OpenText will also work on integrating it with its existing suites as well as building industry-specific applications.
“With OpenText Magellan, we are moving towards a more open, scalable and affordable future for AI and cognitive computing. OpenText is committed to ensuring AI is designed into its offerings from day one,” Barrenechea said in a statement.