To answer this question, we looked at the fundamental drivers of the space to put together Machine Intelligence. The work is a primer that explains how artificial intelligence works, what is motivating the revolution in neural networks, provides examples of technology and top players, and highlights the potential future of the software. If you're yet not a client of the firm, we've combined the primer and financial services analysis into a single comprehensive document for free download. If you are a client of our research, then we offer two separate analyses that go even further than the public version.
One key takeaway is that the science behind today's software is half a century old, seeing several peaks and valleys of excitement, investments, and disillusionment. We've been here before, but not with the needed hardware (there are 20 billion smart devices) and software to make it work (cloud computing is a $100B market). Many of the math concepts underlying current advancements come from prior academic work, powered by the massive computing power and data sets with millions of data points across various types of human activity. All courtesy of the web.
Further, designing software by automating a process top-down is fundamentally different from using machine learning and neural networks, which create probabilistic models that change in response to new data. What is most surprising is just how creative the outcomes can feel. In this way, AI can also be used in a creative capacity to explore a space of ideas quickly or to do emotional tasks.
The growth and potential of Artificial Intelligence is a massive challenge for the traditional economy, and its development is likely to only accelerate. There are several sources of exponential growth: open academic archives, open source code, some form of Moore's law, increasing interest from students in AI, and ample venture capital.
But it is important to be grounded -- today’s narrow Artificial Intelligence is not a panacea and does not have general reasoning capacity. Still, there are many practical applications of automated human judgment. And those applications are subject to myriad ethical and existential concerns with which we must engage before it is too late.