Artificial Intelligence (AI) is poised to have a transformational impact on the way we live and work. The AI for Business series aims to provide business and IT leaders with the tools and understanding they need to make smarter decisions about AI.
Science fiction author Arthur C. Clarke famously noted that “Any sufficiently advanced technology is indistinguishable from magic,” and few advancements have demonstrated this adage better than the recent introduction of generative AI tools like Dall-E and ChatGPT. Along with the ability to “create” original work, seemingly out of thin air, AI-enabled tools are being tipped to upend entire industries in the not-too-distant future, so this is not a topic that business leaders can afford to ignore.
Yet, as with previous buzzwords du jour (RIP Metaverse, we hardly knew ye), discussions around AI can come with an overwhelming amount of hype, making it hard to distinguish between the trivial and the truly transformational.
In this article, we set out to demystify AI and establish a foundational understanding of the technology to help business leaders separate facts from fads. To help us do that, we turned to Michael Littman, author and professor of computer science at Brown University and the director for the Division of Information and Intelligent Systems at the National Science Foundation. His book, Code to Joy: Why Everyone Should Learn a Little Programming, will be published later this fall.
Littman says one thing to remember about AI is that it’s a great tool, but a poor solution for business problems. “Your job is not actually going to get much easier. It’s just that you’re going to be able to go further because you’re actually working with this, this powerful tool. I think a lot of people forget that. They think, oh, well, these machines are smart. We can just turn it over to them and they will do that job. And that’s not proven to be the case at this point.”
The many layers of AI
When it comes to understanding artificial intelligence, one of the first major challenges can be found in the name itself. “Intelligence” can be a bit of a loaded term since it is a characteristic primarily associated with consciousness and human beings. Yet, any scholar of history or reality television can attest to the fact that “intelligent life” does not always result in intelligent outcomes, so it’s helpful to start by defining a few key terms associated with AI:
- Intelligence: In relation to AI, Stanford University defines intelligence as “the ability to learn and perform suitable techniques to solve problems and achieve goals, appropriate to the context in an uncertain, ever-varying world.” It is this ability to learn and adapt to changing conditions that separates intelligent systems from even the most complex machines.
- Artificial Intelligence (AI): Originating in the mid-1950s, the term “artificial intelligence” is described by IBM as “a field which combines computer science and robust datasets, to enable problem-solving.” Though AI is often portrayed as some singular, monolithic technology, it is best understood as a broad field comprising individual subfields, such as computer vision or natural language processing, that focus on synthesizing one component of the larger intelligence puzzle.
- Machine Learning (ML): Machine learning is a subset of the AI field that focuses on using data and experience to allow programs to “learn” and improve themselves based on past results. ML, which contains the subfield of Deep Learning, draws from disciplines including computer science, statistics, psychology, and neuroscience to create a broad array of learning models that are utilized throughout the various subfields of AI.
- Generative AI: Driving the current AI boom, generative AI tools such as ChatGPT and Midjourney utilize deep learning models and massive amounts of training data to generate content based on user prompts.
The moving pieces of AI
Aside from confusion around language, the recent advances in generative AI technologies have blurred the lines between computational output and creative expression. What is actually going on behind the scenes to produce that image of a breakdance battle between Bigfoot and Abraham Lincoln?
The full answer to that question can be a bit complicated, but you don’t need an advanced technical degree to realize the benefits that AI can bring to your organization. Even if you, like most users, will primarily use AI on an end-user or application level, it is still helpful to understand a few of the basic elements that go into creating AI models, along with the potential implications they can have for you and your business:
- Lots and lots of data: As you might expect, creating an AI model requires a massive amount of data, but where that data is sourced from and how it is used can have wide-ranging implications, from copyright issues to perpetuating the same biases that have long plagued human society. The AI-powered tools you implement within your organization could have access to both your company's and your customers’ most sensitive or privileged information, so ensure you are able to secure that data when evaluating your options.
- Computing power: To make accurate predictions, AI-powered tools must be trained on a huge number of examples. Artificial intelligence, along with most other modern technology, has been made possible by the exponential growth in processing power and efficiency over the past several decades, yet for some business applications, it can grow to be a major cost center, especially at scale. The environmental impact of the amount of computing power these tools need is also an important issue.
- Pattern recognition: A lot of AI work is inspired by the human capacity for pattern recognition. Computers can intake more data than a human could in a lifetime, but unless they’re very carefully programmed, generalizing those patterns with the ease that humans do it is incredibly hard. In fact, subfields of AI, such as image recognition and natural language processing, illustrate that even the most basic human actions often involve surprisingly complex acts of pattern recognition. AI tools can reference vast quantities of data to identify patterns and make connections that can deliver valuable insights and innovation into how you do business. However, remember that, as with all statistical analysis, correlation does not always equal causation, so don’t assume that every correlation you find is meaningful.
- Predictions: Having an excellent predictive tool gives you the ability to optimize a system to account for variables you might not have been aware of previously. From healthcare to hospitality, predictive analytics are being used to analyze data and predict future outcomes help businesses strategize for almost every eventuality. However, this data should be viewed as an additional advisor, not the “end all be all” when making major decisions with your business.
- Optimization: AI-powered tools employ a range of machine learning models to improve the quality of their predictions over time based on experience and additional data. This ability to adapt and improve can produce huge benefits, though ML is not always the best tool for every job, so be mindful when implementing AI-powered solutions and in the way you interpret and communicate results.
- Guardrails: Creating an AI model that can generate any image imaginable is no easy feat, but the greater challenge may be in defining what images that model absolutely should not create, even if prompted, due to moral, ethical, or legal concerns. Ensuring AI tools are suitable for end users requires a surprising amount of human supervision, and there is a growing chorus of lawmakers and industry experts cautioning against the risks of unchecked AI. So, when implementing AI tools in your organization, don’t neglect the need for ongoing human oversight.
From theory to practice
Given the dominant role that AI is poised to play in the future of business, understanding these building blocks can help business leaders make more intelligent decisions when it comes to artificial intelligence. Littman says it’s important to not try to go it alone when using these systems. “Don’t make it up because we’re starting to actually gather some real information about what makes a difference when you put these things out into the world.” A practical place to start is the National Institute for Standards and Technology’s AI Risk Management Framework, or AI RMF, Littman suggests.
Coming soon: Additional articles in GoTo's AI for Business series, where we'll build on this foundational understanding of AI and help business leaders learn what they can do to stay ahead of the curve.
