Given how often ‘artificial intelligence‘ is discussed in the press and at conferences, one would think that any business will use it — or at least has it. But the adoption of more advanced machine learning techniques that have revived AI research is dishonest at best.
What is preventing companies from adopting this more advanced approach?
Artificial Intelligence (AI) will provide an additional $13 trillion in economic output by 2030, which, by McKinsey’s latest estimate, could increase the global GDP by 1.2% per year. These numbers highlight the untapped potential of machine learning: any large society, any industry can be transformed by artificial intelligence. AI is the science of building computers that behave intelligently; some of this concept comes from knowledge derived from data, which is the primary purpose of data science.
I’ve worked at AI in many technology companies, including Google and Baidu, and I’ve spoken to countless CEOs who are trying to invest in machine learning and AI but aren’t sure how to do it. From the interviews, I identified the biggest obstacles that companies face in implementing AI:
1. Talent: AI talent is a serious shortage. Hiring AI experts is important but companies that are willing to add to this hiring by investing in their talent in the pipeline and self-training can maximize their artificial intelligence capabilities. MOOCs are a great resource for business teams (I offer step-by-step instructions for Courser; there are many other great options online).
2. Choose the right projects: Like other technologies, AI projects should be chosen because they are: (1) business improvement, and (2) possible. It requires business judgment and deep technical judgment. Work with trusted partners from the outside as well as with trusted partners inside to launch small lighthouse projects and build momentum.
3. Purchasing a CEO: All large companies using artificial intelligence have support at the CEO level. Commitment at the level of the CEO and board of directors is required to create an artificial intelligence strategy that equips and influences the entire organization while helping you create defensible business benefits.
4. Artificial Intelligence Workflows and Processes: For many companies, establishing a central CIO is an important milestone in adopting new technologies. A senior AI officer, like any other central AI manager, can serve as a point of contact for all teams, create new workflows, and collaborate to integrate technologies that are not created internally.
5. Infrastructure (including data): Every company has valuable legacy data, but identifying the right issues to address this data requires deep judgment. In addition, creating a scalable strategy for collecting, tagging, and enhancing your datasets so that they can be easily applied to your AI efforts can make your organization an industry leader.
6. Sectoral regulatory barriers: Companies in highly regulated sectors (automotive, healthcare) face unique compliance challenges. Timed, controlled starts often prove to be the most effective.
7. Education: Real AI societies are built, organized and modified. Training an entire organization in AI, from product managers and salespeople to operations and human teams can help multifunctional leaders decide how to go through changes in the AI industry.
8. Fear: AI forces change in many organizations. Companies will be most successful if they communicate clearly how they plan to adopt new technologies and how they can help employees navigate these changes. Those who dive without explaining it to employees can resist. Those who can resist can avoid creating an artificial intelligence strategy that risks being abandoned.
(Andrew Ng is a senior contributor to the AI fund and CEO of Landing AI. He is also a former Chief Scientist at Baidu, the founder of Google Brain, a co-founder of Courser, the founder of deeplearning.ai, and an adjunct professor in the Department of Computer Science at Stanford University.)