The complexity of artificial intelligence goes beyond technical wizardry. Man may have created statistical models that emulate human thought processes, but are we qualified to deploy them in real life?

Unrepresentative and flawed data sets skew results. Companies are begging for qualified data scientists; yet students graduating with specialized degrees often aren’t prepared for real-life job scenarios. And companies are deploying AI in siloed environments, without properly considering the true questions that need answering.

Even the terminology is confusing: AI, machine learning, deep learning — which is which, and what is what?

“Deep learning, machine learning, AI … have been used more or less in the same way, but they are not really the same thing,” said Francesca Lazzeri (pictured), senior machine learning scientist and cloud advocate at Microsoft Corp. “You have to think about deep learning as a type of machine learning, and then we have AI … on top of everything.”

Image: Francesca Lazzeri, Microsoft

Deep learning is a subset of machine learning, which in itself is a level under AI, Lazzeri explained. “AI is a way of building applications on top of machine-learning models, and they run on top of machine-learning algorithms,” she said. “So AI is a way of consuming intelligent models.”

Lazzeri spoke with Stu Miniman and Rebecca Knight, co-hosts of theCUBE, SiliconANGLE Media’s mobile livestreaming studio, during the Microsoft Ignite event in Orlando, Florida. They discussed the challenges facing mainstream adoption of AI and how Lazzeri and Microsoft are working to smooth the path to AI-enabled applications that run on verified machine and deep-learning models and representative, clean data sets (see the full interview with transcript here).

This week, theCUBE spotlights Francesca Lazzeri in its Women in Tech feature.

A study of man and machines

The philosophy of technology has been an underlying theme in Lazzeri’s professional life. She graduated summa cum laude with a master’s degree in economics and institutional studies from LUISS Guido Carli University in her home town of Pisa, Italy. Lazzeri then was awarded a doctorate degree in philosophy with a focus on economics, technology innovation, and statistics from Italy’s top-ranked Scuola Superiore Sant’Anna, where she did research in economics and innovation management.

A post-doc research fellowship in economics from Harvard University brought Lazzeri to “the biggest technology cluster in the world,” as she describes Boston, Massachusetts. After completing her time at Harvard Business School, she joined Microsoft’s machine-learning Cloud and Enterprise Division.

Five years later, Lazzeri is a senior machine-learning scientist at Microsoft, building cloud-based machine-learning solutions. She is part of an international team of cloud advocates, data scientists, and developers who work with both big enterprises and academic institutions.

In data science, academic study needs to be matched with hands-on experience

The bottom line is important, but the new Microsoft culture emphasizes social consciousness alongside driving profit. To balance the equation, Lazzeri’s team are involved in mentoring doctorate and post-doc students at Massachusetts Institute of Technology, Harvard, and Columbia University.

“What I try to help them with is changing their mindset,” Lazzeri said.

Although the students are among the top emerging talent in the data and machine-learning science field, they lack real-world experience. “These are all brilliant students,” Lazzeri explained. “[They] just need to understand how they can translate what they have learned doing their years of study … into the real world.”

Lazzeri provides the students with data, then works with them in a lab environment to build end-to-end solutions. This gives them practical application building skills they do not receive from their academic curriculum. “I prepare them for their industry, because most of them want to become data scientists, machine-learning scientists,” she said.

The flow of information is two-way, with Lazzeri learning from the students as she teaches them. “The beauty of this is also that we see how other people are using [these solutions] to build something even better,” she stated.

Azure ML platform supports AI deployment

Democratizing AI is something Lazzeri is focused on. “My goal is to make developers like you awesome at applied AI and machine learning,” she states on her blog intro page.

This objective is in sync with Microsoft. The company is positioning Azure as an intelligent, hybrid-cloud platform ready to operate in a multicloud world. And the Azure ML service has been updated to help businesses solve the challenges of deploying AI in the workplace.

“I say [Azure ML] is a very dynamic and flexible tool because it’s a tool on the cloud that is targeting more business people, data analysts, but also pro data scientists and AI developers,” Lazzeri explained.

While drag-and-drop tools in the Azure Machine Learning designer democratize access to machine learning models for non-coders, the Azure Machine Learning Python SDK allows pros to build models from scratch.

Algorithm cheat sheets offer businesses a standardized route to identifying core business needs and avoid the pitfalls of deploying AI at random.  “[They are] a really nice map that you can use to understand –based on your question, based on your data — what is the best machine-learning algorithm, what’s the best designer module, that you can use to build your end-to-end machine-learning solution,” Lazzeri said.

Problem + data + question = business value

“Everything starts with the business problem,” Lazzeri stated. “It is also important to be able to translate the business question in to a machine-learning question,” she said.

The algorithm cheat sheet is a useful tool to help businesses do just this. But there are still two more steps before machine learning can drive value.

Understand the data is step number two, according to Lazzeri. Data headaches are “not an issue that you can really fix,” she said, but there are ways to check that the data sets are the best possible. Use external and internal data to make sure it is representative of the entire population are important factors, Lazzeri advised. Also, data scientists need to collaborate with others in the company to avoid errors.

“Always make sure that you check your model with a business expert,” she said.

Step number three builds on the idea of breaking out of silos and working in collaboration: the deployment of the model.

“This is really the moment in which we are going to allow the other people, meaning internal stakeholders, other things in your company, to consume the machine-learning model,” Lazzeri said. “That’s the moment in which you are going to add business value to your machine-learning solution.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Microsoft Ignite:

Photo: SiliconANGLE