Microsoft aims to coax AI developers
Microsoft has traditionally had strong ties with developers. It now wants to emulate its success, using a platform to simplify AI development
Microsoft is fleshing out its AI platform for developers creating AI-enabled applications. The platform includes open source machine learning frameworks such as Tensorflow and Pytorch, integrated with an Apache Spark service called Azure DataBricks.
Microsoft has also bundled in support for Onnx,the open ecosystem for interchangeable AI models. Developers can also use pre-built services to create speech and image recognition applications and there is a bot service for creating digital assistants.
Analyst IDC’s latest Semiannual artificial intelligence systems spending guide has forecast that AI systems spending will reach $5.2bn in Europe in 2019, a 49% increase over 2018.
As a result, there appears to be plenty of competition in the AI platform market. Along with Microsoft, Amazon Web Services, Google Cloud Platform and IBM are all keen to sell AI-enabled cloud services to businesses looking to use AI to gain a competitive edge.
For AI, Microsoft appears to be putting a lot of emphasis on the developer community. Its head of Cloud + AI business is Scott Guthrie, who has a developer background and co-invented Microsoft’s original ASP.NET web application technology. Guthrie is the Microsoft executive vice president with responsibility for a large chunk of Microsoft’s business, including Cloud, Windows Server, database, CRM, ERP, and the AI platform.
He regards Microsoft’s AI platform strategy very much as one to support and aid developers building AI-powered software. As such, he is happy to feed suspicion that other tech giants may end up competing with their customers.
“Our business model at Microsoft is focused on delivering a B2B [business to business] experience and helping our partners and customers deliver to their customers, and that’s a differentiator that increasingly comes up versus Amazon as well as Google, given that their business model is ultimately about reaching users.”
How are customers using Microsoft’s AI platform?
Deschutes Brewery in Oregon is one of the businesses that is now using the Microsoft AI platform. Kyle Kotaich, operational tech lead at Deschutes Brewery, said: “Three to four years ago, we were looking at being able to present real-time data to our operators, brewers and quality technicians. One of the tools was machine learning models that fill in the gap between data from manual measurements, which are critical for our processes.”
Joel McCune, a solution engineer at Esri, said machine learning plays a large role in its software. “We are working with a large government project, looking at the imagery following a natural disaster. When the hurricane hit down in Florida Panama Beach was hit hard, the next morning we collected aerial imagery.
“We’ve built a model that is able to recognise damaged structures. That allows first responders to know where to search for survivors. We’ve trained another model to assist in understanding where the responders can and cannot go, locating blocked and damaged streets. It is a solution that really does save lives.”
Icertis Solutions, a specialist in contract management, is another business running the Microsoft AI platform. “Artificial intelligence now crawls through hundreds of thousands of contracts in multiple languages,” said the company’s product management expert, Vivek Bharti.
“Contracts are largely unstructured data. We use the text analysis in [Microsoft], Cognitive Services and build context-specific skills on top of that, a pipeline on top of the Microsoft services. This becomes like any other financial or supplier data that you can go in and search like structured data.”
For Jean Lozano, chief technology officer at digital asset management company MediaValet, Microsoft’s AI platform enables machine learning in the cloud. “We realised that the problem is not content creation but content lifecycle management,” he said. “We decided to build on Azure. I told myself we would never hire a systems engineer. Cognitive services is one of the services that we use, and if you combine cloud and AI, what you have is machine learning as a service.”
Read more about AI platforms
- With so much artificial intelligence, machine learning and deep learning in development, we look at the kit programmers might find useful in their AI toolbox.
- Devising compute strategies for AI applications can be challenging. Find out about the hardware, network and software frameworks available.
Ethics in AI
Microsoft sees ethics as an aspect of AI the industry needs to address. Michael Philips, assistant general counsel at Microsoft, said: “We don’t debate so much about can computers achieve these things. The real question is, should they?”
Microsoft has an AI and Ethics in Engineering and Research (AETHER) Committee. “The technology is out there and it is moving very quickly,” said Philips. “Some fantastic things have happened at the EU and in Singapore. But there is a real gap still between these concepts and getting to a framework and then translating to actual practices.”
Microsoft, and others including Google, are “calling for policy makers to become more engaged”. Regulation will be focused on sector-specific applications such as facial recognition, he said.
“There are calls to put a moratorium on the technology. We hear the concerns, but the promise of the technology is so significant that just putting it on ice really isn’t the way to get through the issues.”
A history of extrapolating developer complexity
Microsoft’s developer platform used to be characterised by end-to-end Windows, .net and SQL Server. But the world of machine learning and AI is dominated by open source. This is something the company now embraces.
“Our focus is on how do we build powerful composable pieces that can be used independently and that embrace open protocols, whether it’s HDFS [Hadoop Distributed File System], whether it’s Spark, PyTorch or TensorFlow, Python, and the rich variety of open source. So customers have maximum flexibility,” said Guthrie.
These tools are challenging for organisations without data science skills, so the company also provides pre-built AI services. Azure Cognitive Services packages AI in the fields of vision, knowledge analysis, speech, language, anomaly detection and finally search, making it possible for non-specialist developers to build AI solutions.
Democratisation of technology.
“If you look back at Microsoft’s history, that’s always been our approach. Once, building anything with a UI was super hard. Then Visual Studio and Visual Basic came along and changed it,” said Guthrie.
Given Guthrie’s background and Microsoft’s track record in the developer community, from an AI-perspective, it is more evidence that under CEO Satya Nadella the company is becoming less interested in consumer technology.
From an IT strategy perspective, the question is whether developers will buy into its developer-first approach or opt for new, exciting technologies coming out from the likes of Google and Facebook.