AI has time and again failed to live up to the promise that surrounds it, despite periods of significant scientific advances. Technological progress has however gathered pace especially in the past decade. Machine-learning algorithms have progressed, owing to the development of reinforcement-learning and deep learning techniques based on neural networks. Ranging from language translators to facial recognition and assistants like Alexa and Siri, AI now powers so many real-world apps that it's become commonplace. Businesses across industries are utilizing AI to power their operations increasingly, along with these consumer applications.
Various factors have contributed to the recent advancements in the field. Compared to even a decade ago, we have exponentially more computing capacity to train more complex and larger models. The humongous amounts of data being generated and available now to train AI algorithms is another key factor. Now despite these advancements, many issues remain that will require more breakthroughs to make more headway. Till now, most of the progress has been in what is referred to as "narrow AI." Here machine-learning techniques are being developed to solve very particular problems, for instance, in natural language processing. The harder problems come up when dealing with "artificial general intelligence," where the challenge is to build an AI that can work on general problems in much the same way that humans can. Most believe we're decades away from this becoming a commonplace reality.
So what sets apart a successful AI startup in today's landscape?
Have a measurable and clearly defined goal that results in business value. As a business owner, you should look at AI with the lens of business potential rather than merely automation for small gains. From a broad perspective, AI can support 3 critical business needs today: building insight from data analysis, automating industry processes and engaging with users and employees. Ambitious projects that cater to the utopia of tomorrow are less likely to succeed than humble projects that enhance the business processes of today.
Work on a long-term AI strategy. A well-thought-out AI strategy will serve as a roadmap for your company towards building value while creating defensible moats too. Then start small. Once your team begins to see the initial AI projects succeeding and build a deeper understanding of the technology themselves, you will be able to identify the places where AI can serve best and re-route resources to those areas.
Embrace responsible AI. Don't move too fast and "re-invent" things. The project has to be technically feasible. Too many businesses are still starting projects that are impossible to realize with today's AI tech. Before kickoff, ensuring trusted AI engineers do due diligence on a project increases your board's conviction in its feasibility.
Acknowledge the internal talent gap. Identify the gap between what you are working to accomplish and what you have the internal strength to achieve within the time frame. No business today has sufficient in-house AI talent. To gain that initial momentum faster, partner with an outsourced team with strong technical AI expertise. Much of the optimization and construction of deep neural networks remains an art demanding real expertise. With the rise of digital content fortunately such as massive open online courses or MOOCs like ebooks and Coursera, it is more cost-effective than ever to train employees in new skills like AI.
Set up a pilot project. It's time to start building and integrating once your business is ready from a tech and organizational standpoint. The most critical factors here are to start small, keep the project goals in view, and be aware of what you don't know and what you know about AI. Bring in outside experts to help where need be. Then bring internal and external people together in a small team. A tighter time frame will also keep people involved focused on clear goals. After the pilot is completed successfully, you will be able to ascertain what the long-term, more ambitious project will be and whether the value proposition makes sense for your company. It's also important to ensure that expertise from both sides—the people who know about the technology and the people who know about the business—comes together well on your pilot project team.
Fact is, most of the winners in this artificial intelligence tech gold rush are likely to be ones with scale. Google, Microsoft, Alibaba, and other technology giants are racing ahead. They house some of the best engineering and research talent, have access to some of the biggest datasets, built widely used products, and have deep pockets. For a startup to succeed in this highly competitive market, it will need to be supported by leading engineers with deep domain expertise and have access to quality and ideally unique datasets.