The emergence of artificial intelligence (AI) represents a pivotal moment in the history of human endeavors. If you are not a software engineer, you are probably wondering what AI really is. It is a buzzword being thrown around by many. It is something every single company should understand and be mindful of. It is a new way for software to understand the world.
For years, engineers wrote software that was rule-based, meaning that software would understand that if a user clicked button A, then the software should perform action B. However, trying to write rule-based software to make a car understand the environment around it and autonomously drive you from point X to point Y was unthinkable. It would be impossible to do within the known software engineering framework.
At the most basic level, AI is a process where a software engineer trains software; the software is shown many examples and the software learns from those examples. For instance, a team of researchers at Carnegie Mellon wanted to teach a drone how to fly itself. Instead of writing software explaining how to fly, the researchers leveraged AI. In other words, the researchers put a video camera on a drone and had it crash itself into a wall ~11,000 times. From this footage, the drone learned what it meant to crash and reciprocally how not to crash by deciding to go straight, turn left, or turn right. See the crazy youtube video here: YouTube video
Another example of how AI works: software that takes and processes images in multiple layers, extracts features, and associates features with outcomes. In the below example, a picture of a face is input into the AI software. The first layer detects edges, the second layer detects a combination of edges, and the third layer detects object models. When training AI, one tells the software that this input image is a face and the AI associates each feature set with the feature set of a what a face should look like. Then, after training, the AI is able to go through these steps again with each new image and provide a confidence level estimate of whether the image features are the features of a face. This is a very high-level example, and in the real world, these models will have many different configurations, parameters, and layers.
You can also visualize this for yourself. In the following link, you can manually draw a number and AI will recognize what number you drew. You will see how the AI processes the image and the various layers it creates. In the end, the AI will estimate the highest and second highest level confidence estimate for what number you drew.
In the above instances, a software engineer provided a large number of examples from which the AI could learn. Humans also happen to learn in a similar way. If you put a cake with a candle in front of a young child, then they will likely reach out, touch the candle, and burn themselves. After doing this once, or perhaps a few times, they will learn not to touch candles. Both AI and humans learn from examples and from mistakes; both learn from experience.
More advanced fields of AI aim for AI to not rely on a predefined set of experiences or training data, and instead learn purely from trial and error. Google recently developed a simple computer game where an animated stick figure had to learn to navigate an obstacle course. The stick figure was not programmed for how to walk, run, or jump. The only thing the stick figure knew was the range of motion that each of its limbs could make and knew whether it successfully navigated each obstacle. Each time it tried to navigate an obstacle it learned from its mistakes and successes. See the video here.
The future potential of this is hard to fathom. The fact that AI can not only learn from pre-collected examples but also learn through experience plots a new trajectory of what software will be able to achieve. AI has large implications on the products we make, the experiences we sell, and the world in which we live. It is an opportunity to reorient one’s product and business for an unknowable future.
Also, check out the resource section for the: Best Entrepreneur – Books, Blogs, & Resources