🛰️ Discover Kayrros ERCOT Impact Monitor at ERCOT Market Summit 

Get the most timely and accurate information on capacity additions to the grid,

with cutting-edge satellite imagery and AI.

Can Machines Learn? Kayrros Explains AI

macro_excerpt

Can Machines Learn? Kayrros Explains AI

Artificial Intelligence (AI) seems like the way of the future, a piece of a world invented by Philip K. Dick or Isaac Asimov and inhabited by humanoid robots and rampant with computerization. But, it isn’t. In fact, it is part and parcel of modern day life. Even if we don’t recognize it, AI is already widespread and becoming increasingly easier to use, but as it grows it can be more difficult to grasp the basic concepts. That is where we come in.

The use of artificial intelligence has been around already for over half a century. In World War II, American bomber planes needed AI in the form of autopilot, to stabilize the course, attitude, and altitude of the aircraft. Later developments, building on the old models and creating new, more efficient systems in aircraft led to automatic flight and route monitoring.

Now, AI is omnipresent. Old models are changing into new concepts, influencing everything from email marketing, which aggregates consumer data to develop the most compelling customer interactions, to the Mars rover, which traverses an alien landscape far from human control. By incorporating AI, these applications can build on each other, share common codes, learn from previous mistakes, and solve problems beyond human capacity.

AI-based machine learning operates from a few simple concepts. First, data must be fed into an algorithm, like coal into an engine, to determine and estimate an accurate model. A metric is then developed for future inputs, which tailors information and speeds up later analyses. The algorithm works to analyze the data and creates outputs based on its initial design. Similar to a locomotive, whose efficiency depends on the more you feed it coal and the better you grease the wheels, AI machine learning becomes more efficient the more you input data and tailor the results to help it “learn.”

The applications for machine learning are wide and varied. The algorithm can group data, compress data, quantify or categorize data, which is useful for everything from devising an efficient trash compacting system to protecting computers in a hospital from malicious data hacking. By grouping data over a time period, AI can yield actionable results, such as the optimization of crop yields for a farmer or predict future supply movements at storage facilities.

What makes these algorithms extremely useful is their ability to adapt and react, evolving as data is fed into them to develop processes for more targeted outcomes. They are also forward-looking and can postulate future movements, like a player at a chessboard devising techniques to prepare for future actions. These differentiations from traditional processes allow for growth, helping researchers create new algorithms to solve increasingly complex problems.

In essence, machine learning is about extracting knowledge from data. As new tools create different processes for analysis, research is becoming faster and more accessible. Current technology has created an interconnected web of communication, opening up new avenues for innovation.

Though AI is an old concept, and has developed compatibility with many different sectors of daily life, there are still spaces that will benefit from development. Driverless cars are the hot topic of the day, but whether that will create a more efficient system or just work at all, is still up for debate. The concept that Kayrros sees and wants to work towards is driverless energy production.

Imagine a London power plant operated manually by engineers to determine the output necessary for a weekend, taking into account everything from weather to football games on TV. Human-run systems are prone to error, and a change in weather patterns could throw off a predetermined model. If a well-built and adapted algorithm is designed to let the plant do the projecting, the plant would be able to adapt to fluctuations in weather or use in real-time. A one percent increase in power plant efficiency will create a staggering effect on consumption and production.

Energy is an overlooked topic, and the power that lights our home and delivers WiFi to our computers is often taken for granted. At Kayrros, we talk about energy because it matters.