Machine learning. Neural Networks. Artificial intelligence. In 2019, these three buzzwords were the most frightening to my friends and colleagues outside of the tech industry. Their fears stemmed from the constant news that they would be phased out of their workplaces within the next decade and replaced by a band of hyper skilled, unfeeling robots. To calm their fears, I like to remind them that using calculators never stopped them from knowing how to count, and in a similar way, AI won’t stop them from working. In fact, in the next, decade AI will make working even easier than ever.
First it’s important to establish that automation is less about Skynet style shenanigans and more about making humans more efficient. In its simplest sense, artificial intelligence is a system that uses human input to generate responses more quickly and accurately than humans can. Calculators, our smartphones, your Netflix suggestions are all powered by these systems which use data points provided by users, run through a system designed by people, to get optimal results. Machine learning is a core concept of modern AI and speaks to a systems’ ability to “learn” as it completes a task. Using neural networks, which mimic how human brains process information and develop connections to learn new tasks, AIs are able to complete complex tasks and generate sophisticated outputs.
One of my favorite examples of how far we’ve come with AI is the news from a university research team in Germany that “taught” an AI system to paint in the style of Van Gogh, Picasso, and other art masters. By feeding the system information (in this case pictures) from these great artists, it learned the elements that differentiated each painting style. When prompted, the machine was able to “separate and recombine” elements of the images it was taught to create new images in the specific art styles. The AI essentially “learned to paint like Van Gogh” and could apply the “Van Gogh style” to other images it was taught.
With developments like that, the next great frontier for AI developers is creating an AI that can rationalize independently. IBM’s AutoAI platform automates data preparation and ML model development. Startups like H20.AI offer companies enterprise-level machine learning systems without the need to deploy a dedicated AI/ML team. The suite of machine learning products technical and non-technical users to extract information from data inputs without having to develop their own algorithms or machine learning models. This is about as close as we can get to “sentient AI” with the limitations on technology currently. The goal of such tools is to make AI accessible and usable by wider audiences, hopefully allowing it to shed the stigma that surrounds it.
All of these developments won’t lead to the eventual replacement of humans by machines but rather change the way we work. Currently, these systems are best for completing rote, repetitive tasks, which if applied correctly can free up workers to learn other more complex skills. The proliferation of AI in an industry does not mean the displacement of the people there, it means that companies need to better empower workers and develop their skills to increase innovation and productivity. Humans are critical to the growth of AI, as it needs humans to build it, teach it and correct it’s course every so often. These systems, though powerful, are easily fooled by a simple change in data, so it’s critical that we pay attention to the conclusions the system renders. By teaching workers about their new tools, and how to work with them employers can do their part in normalizing AI in the workplace.
One of the key areas for improvement for AI in 2020 will be increased accountability and security as the industry becomes more transparent. There have been many public cases where AI has been deemed to make poor or biased decisions, such as the gender bias issues Amazon experienced with their recruitment matching AI. This will see the creation of trustworthy workflows interwoven into the structure of the technology giving us more control over our “autonomous” partners.
Read the full article in the Nasdaq