Thought Leadership

An algorithm by any other name

By Spencer Acain

Algorithms are everywhere. From Google deciding what the top result will be for a search, to the systems that keep a plane steady in the air to even the recipe for a humble grilled cheese. As with many things, algorithms come in a wide variety of types with one of the most interesting in recent years being artificial intelligence, a class of algorithm which differs widely in both form and function from its predecessors. As AI grows increasingly complex and integrated with the world, it’s important to understand where the lines are drawn between traditional, familiar algorithms, and the new AI algorithms of the future.

What is an algorithm?

In the simplest sense, an algorithm is a set of instructions, usually executed on a computer, although a human working through a set of directions could also be considered to be executing an algorithm. These types of algorithms are highly precise in nature and form the basis for the logical structure of conventional programs and computers. Programs built like this follow simple steps such as “if a then b else c” or “while a do b until c” until they reach a final result.

While implementing algorithms of this type has its benefits, there are also drawbacks. Take, for example, a set of directions from a house to a nearby convenience store. The directions themselves represent a simple algorithm: follow this street for 2 miles, turn left, your destination is on the right, etc. Executing this algorithm will always precisely navigate from that house to that convenience store and in the case of directions, this exact, repeatable nature is what anyone using them would want.

But just as this reliable and repeatable nature is an asset to traditional algorithms, so too is it their greatest weakness. A traditional algorithm can never handle anything that it is not explicitly programmed to, meaning every input, output, instruction and edge cases must be painstakingly planned out and implemented into code. In the above example, the algorithm would fail if the starting place was anywhere other than that house, the destination not the convenience store, or if any of the multitude of conditions surrounding the trip where to change in the slightest.

AI: Building a smarter algorithm

At a basic level, AI is still an algorithm but thanks to a fundamental difference in design, one that, for a price, is able to overcome some of the shortcomings of traditional algorithms. Compared to a traditional algorithm, an AI algorithm must learn its abilities rather than having them programmed in directly. Artificial intelligence, which seeks to algorithmically emulate human cognitive capabilities, is able to process all types of, often unstructured, input data and by applying the skills it’s been trained to have, reach an answer much the way a human would.

When a human receives a piece of information, they attempt to classify it using their past knowledge and experiences as context. If the information is something they are very familiar with, the classification will be highly accurate. On the other hand, if it’s something they haven’t encountered before, the accuracy will likely drop accordingly. This is exactly the way AI algorithms work as well, with majority of modern AI systems built using artificial neural networks (ANNs) or deep neural networks (DNNs). Artificial neurons receive input data, either directly or from other neurons, then activate (send the data onward to the next neuron or as an output) based on whether the input value reaches a certain threshold. These thresholds, called weights, are fine-tuned during the training process as the network is tuned to better categorize and understand the data it is being fed.

While AI seeks to overcome many of the limitations of traditional algorithms, it does make some sacrifices along the way. Where a traditional algorithm is perfectly repeatable and predictable, an AI algorithm is not. In its bid for human-like intelligence, AI has lost its perfect precision, with results possessing a qualitative element that can, and very often does, change from one iteration to the next. Not only does AI possess a level of variability in its answers, it can also be wrong. Where a traditional algorithm would fail to provide any answer an AI algorithm can attempt to offer one even when it is woefully untrained to do so, potentially making it a poor substitute for both humans and the algorithms it seeks to surpass.

AI: a matter of intent

At the end of the day, AI is simply a collection of algorithms, some simple, some unbelievably complex. The dividing line between where complex algorithms end and true artificial intelligence begins is largely one of semantics and intent. AI, seeking to emulate humans, is far superior at processing raw, unfiltered data and extracting patterns and meaningful insights from it compared to traditional algorithms – even at the cost of pinpoint accuracy. As mankind continues its quest to better understand the root of “oneself” there may come a day when AI transcends the framework of being a mere algorithm and achieves true sentience however, at least for the foreseeable future, AI will remain what it is today: a powerful algorithm, and a useful tool.

Siemens Digital Industries Software helps organizations of all sizes digitally transform using software, hardware and services from the Siemens Xcelerator business platform. Siemens’ software and the comprehensive digital twin enable companies to optimize their design, engineering and manufacturing processes to turn today’s ideas into the sustainable products of the future. From chips to entire systems, from product to process, across all industries. Siemens Digital Industries Software – Accelerating transformation.

Leave a Reply

This article first appeared on the Siemens Digital Industries Software blog at