Demystifying Exponential Technologies
By Tony Saldanha
If you had to pick only one exponential technology to focus on (and that would be a mistake!), it would most likely be Artificial Intelligence or AI.
AI is essentially the imitation of intelligent human behavior by computers or machines.
It is the broadest way to think of computer intelligence. All other related terms are usually sub-sets of AI. So, Machine Learning is a subset of AI that ingests data to learn specific tasks. Deep Learning, which became prominent when Deepmind’s AlphaGo program beat the world Go champion, is in turn, a subset of Machine Learning. It is a way to solve complex problems using neural networks that simulate human decision-making.
You are already a major consumer of AI technologies in your personal life. If you’ve used Siri or Cortana or any other virtual assistant, you’ve already used AI. If you’ve had a credit card or bank proactively flag or block a transaction, you’ve experienced the fraud detection use of AI. If you’ve had Amazon or Netflix or a similar service provider recommend a product based on your profile, that’s AI too. Driverless cars use AI. An example that you might not know about – a huge number of simple stories that you read online on Yahoo!, AP and others on financial summaries and sports results come from AI tools.
Understanding the possibilities and limitations of AI:
AI may be all around us, but it is not omnipotent. The key to unlocking its vast potential lies in an innocuous but powerful term that’s well known in IT circles – use cases. A “use case” is an application of a specific tool to a given problem. So, credit card fraud detection and machine-generated news articles are two “use cases” of AI.
The reason why AI is suddenly so popular is that the number of use cases has hit a tipping point driven by affordable computing power. That phenomenon is true of all exponential technologies, but it’s just that AI happens to be at the peak of the cycle. The internet was one such exponential technology 20 years ago. It spawned off an explosion of use cases. In fact, much of the dot-com boom was about innovators creating new use cases on top of the internet. Most of the early use cases on the internet were about “access” to stuff – buy products online, check your bank account, transact with your Bureau of Motor Vehicles. As the dot-com era died, it spun off a second generation of use cases that were in themselves further platforms to build even more use cases. Cloud computing is one such example.
It essentially made computing server capacity available online to anyone with an internet connection, which in turn spun off a new generation of cloud-based software applications that were dramatically better than those that you had to physically install on your PC or your server. AI is experiencing a similar explosive growth of use cases.
Use cases that matter:
The good news is that one doesn’t need to be an expert in AI to understand the type of use cases that could help or disrupt your business model. Here are a few tips on AI that should help.
- You don’t just “do” AI, like you don’t “do” the internet. It’s the use cases that matter
- Therefore, beware of any vendor that’s trying to sell you AI as a panacea or as a platform; you’re in the “use case” race, not the AI race. Unless of course, you’re an AI developer
- Most of the AI algorithms are open sourced. Any vendor selling you an AI platform is often packaging free stuff and probably selling it at a premium. There’s some value in curated and packaged algorithms, but usually not as much as one would think.
- There are literally thousands of use cases possible in big enterprises. A small number of these will be critical to your future business model. Focus on those – follow the money.
- As a corollary, if your future business model depends on these few use cases, you probably want to develop some intellectual property in them to provide sustainable competitive advantage. Build enough AI and data science capabilities in-house for these.
Examples of use cases: AI can exist wherever human judgment is involved. I have highlighted only a few select categories of use cases broadly relevant to functions in most enterprises, for illustration and inspiration.
Practical information you need to know:
How to identify use cases
- Bring your best business experts and data scientists together to identify them. Even they will need to iterate and experiment frequently. Don’t try to survey for use cases – experiment!
Building AI capabilities
- Just do it. Hire a few data scientists. And most importantly, start gathering any and all types of data relevant to your business. AI is powerless without having data.
Using AI in old-world industries
- AI is still relevant there. With very few exceptions it’s the combination of old + new that creates disruptive business models. Amazon’s AI wouldn’t be worth much without its logistics capabilities. You need to complement your old-world assets with new-world capabilities.
The effect of AI on jobs
- As with any automation, AI will definitely repurpose people and skills. Your business plan needs to anticipate this.
Will AI destroy mankind?
- Probably not, because smart people are starting to raise the issue of boundaries for AI. But not investing in relevant AI capabilities for your business will sure as heck destroy your business!
This is an excerpt by Tony Saldanha from his best-selling book, “Why Digital Transformations Fail.”