2020 saw more and more applications of artificial intelligence (AI) and machine learning (ML). In part, this was because the coronavirus pandemic presented a need for AI that technology was ready to meet.
As everything from supply chains and creative fields to government agencies discover applications for AI, developers find themselves with more opportunities to implement ML as well as a need for more sophisticated and optimized systems.
2021 looks like it’s on track to be a bountiful year for ML!
AI and Creative Decisions
Writing for Forbes in March 2021, Skylum CEO Alex Tsepko argues that, “We need to accept that our subjective decision-making process is just a slower, more manual version of what AI can much more effectively do on its own using more data points.” Tsepko, whose team created the AI image editor Luminar AI, envisions more content that’s data-driven rather than driven by creators’ subjective tastes.
One example of this currently nascent field is Jukebox, which comes courtesy of OpenAI. Jukebox creates new songs when fed a genre, artist, and lyrics, which are sometimes also co-written by AI. ML algorithms like Jukebox, like other similar products, could democratize content creation by making it accessible to a more extensive user base.
Jukebox paves the way for more affordable background music for everything from films and commercials to supermarkets and office spaces.
Greater Use of AI by Governments
As reported by GCN, the coronavirus pandemic accelerated movement that was already underway. The need to minimize face-to-face interaction meant that many tasks that could be automated were, such as the distribution of unemployment checks and food stamps. Many bureaucratic first steps can also be easily handed over to AI, like scanning forms for discrepancies or missing signatures.
2020 saw government agencies get comfortable with using ML to process the tremendous amount of data that they have at their disposal. In 2021, that trend is expected to continue into the arena of predictive analytics. The Department of Agriculture, for example, has already started using predictive analytics to help farm, forest, and ranch managers make environmentally sound decisions.
AI and Supply Chains
Supply and Demand Chain Executive reported that AI and ML’s use was one of the biggest trends in the supply chain industry in 2020. “Across the industry, companies are using machine learning to integrate systems, forecast demand, provide real-time visibility on shipments and improve efficiencies in last-mile deliveries.”
One example was maintaining visibility of supplier inventory in real-time and automatically making suggestions to customers. So, if you ordered vanilla ice cream and were informed while your order was being processed that the supplier was out and you’d need a replacement flavor, there’s a good chance you were interacting with AI.
In 2020, companies were already using AI to make complex decisions about stock warehouses, how much inventory to carry and how to distribute it, and predict demand. This trend is expected to continue given that AI’s use was identified as providing companies a crucial advantage over their competitors.
Rethinking and Restructuring Data
2020 saw many companies applying machine learning to pre-existing databases. These applications’ ad hoc nature meant that there was a built-in bottleneck because these databases weren’t built with ML in mind. Writing for Techwire, Deloitte Consulting LLP expects 2021 to see companies rethinking data management entirely.
“As part of a growing trend, they are deploying new technologies and approaches including advanced data capture and structuring capabilities, analytics to identify connections among random data, and next-generation cloud-based data stores to support complex modeling,” writes Deloitte.
The result will be data that’s already optimized for ML from the start and adaptable to future developments in ML as well.
AI with Built-In Skepticism
MIT is developing an AI model that considers training data that presents inaccurate information to better avoid real-world scenarios where information may suddenly and unexpectedly shift, like in the case of collision avoidance using self-driving cars.
One of the ways the model has been trained is using the game, Pong. As the ball moves toward the algorithm’s paddle, the ball is suddenly shifted downward a few pixels. At first, this led to an algorithm that had previously been winning the game to lose continuously.
But thanks to reinforcement learning, where specific actions in response to certain inputs are encouraged based on their outcome, the algorithm could win the game even with these pixel shifts!
The result is that cars in smart cities can avoid sudden events like kids running onto the street, bicyclists swerving out of the bike lane, or cars swinging open their doors into the path of oncoming traffic.
2020 has set the stage for an explosion of applications for AI in 2021. From governments to businesses and even to creatives, AI is an opportunity to optimize pre-existing processes. But it’s not only simple tasks that ML is ready to take on.
With technology rapidly advancing in its capabilities and ML training methods producing increasingly more complex results, we’ll soon discover the true potential of how AI can change the world.