Building artificial intelligence: staffing is the most challenging part

Every company worth its weight is set on achieving practical and scalable artificial intelligence and machine learning. However, it’s all much easier said than done — to which AI leaders within some of the most information-intensive enterprises can attest. For more perspective on the challenges of building an AI-driven organization, we caught up with Jing Huang, senior director of engineering and machine learning at Momentive (formerly SurveyMonkey), who shares the lessons being learned as AI and ML are rolled out.

Photo: Joe McKendrick

Q: AI and machine learning initiatives have been underway for several years now. What lessons have enterprises been learning in terms of most productive adoption and deployment?

Huang: “Machine learning projects are much more complicated and bigger than ML model algorithms, so be ready to build a robust team to take care of machine learning operations. Staffing a world-class machine learning team is extremely hard. The ML talent with experience are in high demand. One option is to provide training and build a culture that fosters internal transfers; sometimes, growing the team internally can be the key to building an effective ML team.” 

“Before building anything substantial, make sure you examine where the bottlenecks of the machine learning production pipeline are. While deciding on build versus buy, when you shop for a solution to speed up your AI/ML capabilities, make sure the solution you choose can be adapted, scaled up, and easily integrated with your product applications.”

Q: What technologies or technology approaches are making the most difference?

Huang:  “From a broader industry perspective, machine translation and information retrieval, in general, have improved dramatically after adopting deep learning. For example, at Momentive, we see a big difference in ML solutions that are helping customers find relevant and actionable information through massive amounts of response data effortlessly.”

Q: Are most AI initiatives being run internally, or supported by external services/parties (such as cloud providers or MSPs)?

Huang:  “Depending on the use case and organization, the requirements for AI initiatives are quite different. Some of them make more sense to leverage external services, some of them are required to be run internally. In general, we see more adoption of third-party services for use cases that are independent and don’t need to closely integrate with production systems. Whereas, we see more successful homegrown solutions for use cases that need to be tightly integrated with production systems.”

Q: How far along are corporate efforts to achieve fairness and eliminate bias in AI results?

Huang:  “The field as a whole is still learning 00 nobody has all the answers. With that said, the awareness of the impact of bias in AI has risen in recent years and progress is being made. There are increasing efforts to find solutions to mitigate the risk of bias in AI and discussions of bias and fairness in ML have become a new norm in both research and industry.”

Q: Are companies doing enough to regularly review their AI results? What’s the best way to do this?

Huang:  “There will always be human biases – there’s no getting away from that — but one thing we have done is make sure that the people working on this are from a variety of backgrounds to provide a breadth of representation and also feel included.  That means inclusion, not just diversity, in order to highlight all the different kinds of concerns that might be at play.”

Q: Should IT leaders and staff receive more training and awareness to alleviate AI bias?

Huang:  “The research of bias in AI and mitigations of it is fairly recent compared to the history of computer science, not to say compared to human history. Universities like Stanford and MIT started incorporating topics of ethical AI in their AI courses. The general assumption is that the more senior the IT leaders are, the more they can benefit from training that covers the latest development in this field. We have invited AI experts and practitioners from academia and industry to share their experiences and knowledge with our leadership team and all employees in a quarterly cadence.”

Q: What areas of the organization are seeing the most success with AI?

Huang:  “It depends. Normally it’s the areas where historical data are stored and can be easily accessible. Things started changing after deep learning technology was more widely adopted, with synthetic data and adversarial training playing a more and more important role.”

“There are many different parts of an organization that may implement AI successfully. For example, the IT org within the enterprise may use ML/AI technology to improve the efficiency of business processes, the finance org may leverage ML/AI to provide more accurate forecasting, the enterprise might build ML/AI solutions into its product offering to improve customer experiences, and so on.”