With the growth of AI not destined to slow any time soon we thought it made sense to share this solid list of predictions and guidance from Dr. Madhu Shashanka, Founder & Chief Scientist, Concentric.ai. His company focusing on keeping organizations secure via zero trust data security and helps you remediate risk.
1) The COVID-19 crisis has been a reckoning for all businesses, both small and large. As we tackle the current challenges, what is your top prediction for the potential application of artificial intelligence in 2021?
It’s clear AI will face a lot more external scrutiny, and potentially regulatory oversight, in applications with social equity and justice implications. When the AI models get personal (e.g., face recognition in certain contexts, ethnicity or gender identification), they can inform policing or resource access decisions. Those types of applications will face strong pushback in 2021. Less “personal” applications, such as cybersecurity, industrial optimization or agriculture, have limited potential for creating social justice issues and will face less friction in 2021.
2) In your view, what is the main roadblock faced by companies when it comes to AI adoption?
Unrealistic (or, perhaps more accurately, uninformed) expectations. Organizations often think AI is a drop-in component that’ll improve existing processes and tasks. An AI project isn’t just another IT initiative. Rollouts often have more bumps and it’s easy to underestimate the ongoing investments needed. Managing model bias, dealing with missing or incomplete data, and degradation of models over time are new and unexpected challenges that can derail adoption. This inability to realize the complexity involved, coupled with the fear-of-missing-out attitude, leads companies to rush through AI adoption as a new shiny hammer looking for nails. Disillusionment inevitably follows.
3) Any tips or suggestions for firms who are on the path of AI adoption?
- Don’t rush it. Recognize that AI is a means to an end, and understanding your use case, in depth, is essential to a successful journey.
- Stay grounded. Identify specific narrow use-cases that can deliver quick wins if successful. Andrew Ng’s advice is still spot on if you are starting out with AI: manual tasks that take about a second to complete (e.g., recognizing cars or people in security footage) are excellent AI candidates. Start with well-defined, industry-specific uses cases as opposed to generic functional use-cases in HR or Finance.
- Be humble. AI is rarely a linear path. You need a plan, and you need a plan to change your plans. Think of your execution strategy as a series of experiments with clearly defined gates. Be prepared to make major changes or even stop. Don’t let sunk-costs bias get in the way of making the hard decisions.
- Build momentum. Demonstrate quick wins, no matter how small. That’ll set the stage for successful adoption across the company.