They are everywhere! You simply cannot apply for a new job in the tech sector without reading something about AI, neural networks or deep learning. Here’s the tea: In many, many cases, they are used incorrectly. Let’s talk about artificial intelligence!
Oh well… if you believe the advertising of companies in 2021, basically everyone uses AI-driven tools. But don’t be tricked, not everything that involves a computer involves AI as well! Using an AI-driven tool means that the decision making process is powered by an artificial intelligence. A machine can show intelligent behavior, but will, unlike humans, not show emotionality or consciousness.
Intelligent in this context describes the ability of a machine to execute cognitive functions, such as problem solving or decision making. The level of complexity on which these functions are considered intelligent behavior, and thus part of an AI-driven machine, is influenced by the time in which it occurs. Something that was considered AI-driven in the 90s may not be included in today’s definition of AI-driven tools because it is too simple. Technology has evolved massively over the course of the last years – yesterday’s groundbreaking news are standard capabilities today. This phenomenon is called the AI effect.
So, what is an AI-driven tool now? Today’s AI-driven machines typically involve a neural network, whereas machines that operate on statistical algorithms alone are not considered AI (anymore!). So if one of your vendors or clients advertises their AI-driven tools or machines, this is what they usually mean. The AI typically makes a decision or solves a problem based on a neural network that was trained on a large amount of data, possibly by involving deep learning as well.
Are AI-Driven Tools Better At Solving Problems Than Humans?
The answer to this is a clear maybe. If you have a large amount of data already, for example, in your Sales database, an AI may analyze this data in different ways than humans can, find hidden relations or patterns that are not visible to the human eye. Using AI may help you to identify changes that may make a large difference. Maybe you should focus on different industries, call at different times, or have less or more time go by between emails and calls. AI is without doubt very helpful in these cases! However, relying on AI alone means that you omit all the consciousness and emotionality that was previously used in your decision making process (which was powered by humans), so be aware that you might miss important cues.
“Tested And Verified” AI Tools
Humans make mistakes, and AI makes mistakes, too. One behavior that you can see everywhere is people blindly trusting the outcome of an AI-driven tool. This may be the machine translation they got from Google or DeepL, or the recommendation they received via Netflix, or even the decision on whether or not a patient needs further investigation after an x-ray. Especially in the medical sector, blindly trusting AI is a problem and using AI is a highly sensitive topic. So, can you trust AI at all in these areas? All algorithms that are released have been tested thoroughly. Meaning: With a confidentiality of X percent, they choose the right answer from a previously curated test set. There is always a potential for mistakes! The testing data could be skewed. The training date maybe introduced bias to the AI (so it happened to Amazon’s HR algorithm that frequently chose men over women with no particular reason). If the input data changes, the outcome also may change drastically, or the AI needs to be re-trained.
In the real world with real people, we have built a cultural consciousness of trustworthiness: When do we trust decisions, when is something suspicious. AI decisions are so new to everyone that we simply do not know yet when the decisions are trustworthy – and when they are not. No matter how tested and verified an AI tool is, it would always be the best to combine it with human knowledge, emotionality and consciousness to reach the highest results.
This does not mean that humans are perfect by all means. Their decisions may be as biased and wrong and an AI-driven one! Combining both and carefully evaluation the results is a way to improve on both sides. AI may show us ways to interprete our data that we have never thought of – on the other hand, AI can only be as good as the data it received for training.
The Lack Of Standardization
The sad truth is: Only because you cannot understand how something works, does not mean it’s AI-driven. Something that was considered part of AI years ago, may not be included today. In general, we refer to operations using cognitive functions when we talk about AI tools, such as problem solving or decision making.
With all the hype around AI-driven tools, one should never forget that AI is not per se intelligent. It simply is an algorithm trained on a bunch of data. It may derive decisions from this data and will not simply reproduce previously seen results, but also combine or alter them (that’s the magic of a neural network – another interesting buzzword) – but if the data was skewed, the algorithm will be as well. On thing we are still missing today is a definition of trustworthiness for algorithms or AI tools. Most tools are proprietary, meaning the source code and training process is only known by a company or group of developers. As there are no standard thresholds or tests on whether something can be considered an AI or not, or how ‘good’ it needs to be, there is a large discrepancy between tools. In conclusion, the decision of whether it is trustworthy or not always is in the hands of the humans who use it – that’s you.