kromkrathog - stock.adobe.com
Trends in AI pointed toward substantial progress for the technology over the course of 2019. It was the same case for RPA and 5G. But the past year also showed us a lot about what some of the most hyped technologies can't do. As we head into 2020, let's take a look at how the outgoing year's progress sets us up for the year ahead.
A new network on the horizon
5G is expected to become widely available in 2020. The super-fast, next-generation communications network opens new opportunities for how we consume technology. AI, especially, will be impacted by this, as 5G enables edge computing at scale, where smaller devices can offload computationally heavy tasks to the cloud and instantly act on the received results in real time.
5G will help autonomous vehicles get one step closer to reality, as self-driving cars will be able to communicate with each other to reduce crashes and operate more efficiently. This is because 5G not only enables fast communication, but it also enables more devices to connect to the network in a specific area.
Pete Zimmerman, North American software sales manager at ERP development company VAI, believes this could be the "biggest breakthrough" for AI and robotic process automation (RPA). "With 5G wireless connection, businesses can more easily leverage technologies like AI and RPA because of the influx of available data and ability to connect to a network of devices," he said.
The real use of RPA
There were high hopes for RPA in the past year -- and rightfully so. The technology has the potential to enable the automation of low-value tasks, freeing up the workforce to be more productive and creative. But we also learned its limitations. As Miguel Valdes Faura, CEO of digital process automation platform Bonitasoft, put it, "Quick wins using RPA may be just baking in mistakes."
Companies need to evaluate and optimize the tasks they assign to RPA bots with a hard look at how they fit into the overall process. "Automating an inefficient process just makes it faster, not better," Faura said.
Enterprises' approach toward RPA is also changing. When AI was on the rise, many feared it would replace human workers. Ultimately, we learned that AI is best used when it augments the workforce. A similar approach works best with RPA, too.
"Companies are starting to realize the importance of including people in the implementation of these technologies," said Sagi Eliyahu, CEO and co-founder of the robotic automation platform Tonkean. "In 2020, we will see the breakthrough of human-in-the-loop automation platforms that allow companies to automate processes across both systems and people."
A need to understand
Grace Murray Hopper, a pioneering computer scientist in the middle of the last century, popularized the saying "It's easier to ask forgiveness than to get permission." Many entrepreneurs live by this "move fast and break things" approach, but a continuous backlash from the public, as well as governments, is changing that mentality. Silicon Valley's perceived failures -- from Facebook's inability or unwillingness to keep user data private to the never-ending stream of data breaches -- have dominated headlines all year.
This reflects heavily on AI projects, as they traditionally operate as black boxes by design. Bias, ethical concerns and complete miscalculations make explainable algorithms one of the top trends in AI for the year ahead.
Read our in-depth guide for details of how the role of the CIO has evolved along with the consumerization of IT and learn what is required of chief information officers today.
"There are significant unintended consequences when the data scientists and engineers who build and train the models don't fully understand the business and customer ramifications," said Greg Betz, senior vice president of data intelligence and automation company NTT Data Services. "I wouldn't be surprised to see Capitol Hill hearings where executives attempt to explain their company's decision-making based on AI output."
Matt Sanchez, CTO at CognitiveScale, an enterprise AI software development company, said the problem will grow as machine learning becomes more pervasive.
"Nearly eight out of 10 enterprise organizations currently engaged in AI and machine learning report that projects have stalled due to issues of data quality and model confidence," he said. "One of the biggest challenges in AI adoption is the lack of trust and transparency in automated decision systems, and unless these ethical issues are addressed, AI adoption will slow down, threatening its value in many enterprises."
2019's trends in AI witnessed many breakthroughs in the technology. For example, OpenIA released GPT-2, a natural language generation tool that was widely feared for its potential to create fake news or replace journalists.
But we've also learned more of what AI really is and what it is capable of. Despite the progress, the current advancements are only applicable to specific tasks -- the computers still fail miserably at some of the most basic tasks every human can perform, such as generalizing knowledge from one situation to another. As developers faced limitations when they tried to make the algorithms more intelligent, perhaps they could make their approach smarter.
One example of this is the predict and prevent methodology. While traditional IT operations rely on monitoring systems to detect and address problems, a different approach is to fix the problems before they even occur. In doing so, companies spend much less on their IT operations, as it is cheaper to prevent a problem than to fix it.
"We are using AI in an unprecedented way for IT operations," said Nitin Kumar, CEO of Appnomic, a company that uses machine learning to predict problems and automatically fix them. "Traditional monitoring tools detect problems, while our AI detects signals and can take corrective action in real time."
This kind of approach could help AI and RPA, which were both heavily hyped throughout 2019, get to the next level. By knowing what they can do best, companies can take a smart approach to optimize their processes, limit costs and improve their bottom lines.