The hype around artificial intelligence has reached a fevered pitch right about now, but the reality is that computers that can think like people are still a long way off. There is great value being offered by AI today, but it's the form of human augmentation, in which humans train the AI systems, that can then do complex tasks on their behalf.
Don’t misunderstand: There is an enormous amount of value unlocked by these new AI applications. Traditional software is powerful but requires significant amounts of configuration and setup to provide value, and even then it's is only useful if the rules they were given to follow hold true. AI systems are flexible, adaptable and require less time to set up, because they can learn from humans instead of having to be told everything they need to do.
This human-aided AI, or “artificial artificial intelligence,” is fast becoming a competitive advantage for early adopters. In the coming years, these competitive advantages will transition into table stakes, and organizations who don’t adopt AI will be less efficient and less competitive. To be ready for this shift, Sean Byrnes, co-founder and CEO of Outlier, offers eWEEK readers six important data points about AI you need to know to help you prepare.
Data Point No. 1: AI is only as good as the data on which it is trained.
AI systems are trained on your data, so they are only as effective as what your data can teach them. If you have incomplete data, then the lessons your AI tools learn will be incomplete and the results unreliable. For example, if you are selling clothing but have a bug in which your systems do not properly save data about sales of shorts, then an AI system trained on your data will believe no one likes to buy shorts. Ensuring that your data is complete, representative and accurate is critical before training an AI system, or else you will be left with an system that will propagate the mistakes in the data.
Data Point No. 2: AI systems can be more biased than people.
It is tempting to think of AI systems as cold, unbiased machines that make decisions based solely on the data. However, the data used to train these AI systems is a byproduct of human actions and decisions that themselves likely contain bias. If an e-commerce company primarily stocks products that are blue because the founder loves the color blue, then all of the purchasing data will be biased toward blue products. An AI system trained on that data can easily be biased to believe blue products sell better, even if the opposite is true. Identify and isolate potential bias in your business and your data, so that you are ready to train AI systems accordingly.
Data Point No. 3: AI systems can do some jobs better than people, but differently.
While these human-powered AI systems are powerful, they do not act or behave like humans. At their core, they are simply highly advanced mathematics that cannot reason or exercise judgment. Even so, there are many tasks they can do better than humans, but they do those tasks extremely differently. For example, when the AlphaGo system defeated a human player in the game of Go for the first time, the human game analysts did not even understand its strategy, since it was playing the game of Go in a way entirely unlike the way humans play. The fact that these systems do jobs differently is neither good nor bad, but if you expect them to do jobs as a plug-in replacement for humans, you will be disappointed. You need to open your imagination to new ways of thinking and operating to really understand how AI tools can improve your business.
Data Point No. 4: AI adoption is an organizational challenge.
New technologies such as AI will change our definition of work, and as a result they will affect the jobs of everyone at your company. This is both exciting and threatening to many, and without proper preparation, your organization may reject AI out of a sense of self-preservation. We saw this during the outsourcing boom, in which employees feared training their replacements; in this case, they fear they are replacing themselves with robots. Planning for AI deployments and educating your organization accordingly is critical to prepare them for the transition and avoid any conflict.
Data Point No. 5: AI increases the value of your data.
One of the amazing parts of AI is how it changes the economics of data. The industrial revolution made it cost-effective to mass produce items that had previously only been luxury goods due to the manual effort involved in building them. AI has the same potential to take data that is too expensive to analyze manually and make it efficient and easy to utilize it in your decision making. As a result, more and more of your data that today is of nominal value will become extremely valuable in the coming years. For example, your customer support records can become fuel for marketing growth campaigns when analyzed in conjunction with marketing and sales data by AI systems. This means that you should not treat data as disposable, but you should begin collecting and storing as much as you can for the day when you’ll use AI to turn it into a competitive advantage.
Data Point No. 6: AI changes the economics of many jobs.
Just as AI will change the economics of data, it will change the economics of jobs as well. Many corporate jobs involve a growing amount of data collection and reporting to simply keep everyone informed as to the state of the business. When AI systems can automate those data collection and reporting tasks the humans in the organization will spend more time making decisions and taking action, which means their individual impact on the business will increase. This means the economic productivity for each human will increase, improving efficiency across the entire organization.
Conclusion: The time is now for human-aided AI.
The AI-powered future has already arrived, and it will only become more important in the coming years. Understanding this change and embracing it is critical to being competitive during the next decade.
If you have a suggestion for an eWEEK Data Points article, email firstname.lastname@example.org.