The COVID-19 pandemic has accelerated the evolution and adoption of many technologies, among which artificial intelligence (AI) is one of the biggest. There was an increase in AI investments across different industries, with the healthcare sector naturally leading the way. So, it’s only natural to estimate that AI’s market and revenue will grow during this year, with reports predicting an increase of 21.3% from 2021.
There are several possible reasons for that growth. AI has been proving to be a strategic advantage for companies of all sizes and sectors, boosting efficiency and productivity. But investments also refer to new use cases and the emergence of new trends that will dominate the following 12 months.
Here are the 5 most important ones.
1. Multimodal AI Starts to Emerge
Critics of AI often point to the way researchers train their algorithms as a limited and imperfect process that caps the tech’s true potential. That’s why there are AI teams looking to replace the traditional training models that depend on one data source. What’s the alternative? Multimodal AI, where the outcome is mapped to more than one source data type.
Multimodal AI combines visual and speech modes to elevate the capabilities of algorithms to something closer to what we humans can perceive. In other words, multimodal AI uses more than one source to understand and deal with an item. Google’s Multitask Unified Model (MUM) is a great example, as it aims to provide better search results based on contextual information coming from 75 languages (rather than relying on single keywords in one language).
The idea of multimodal AI is as simple as it is hard to achieve—to provide algorithms with a richer and more natural way of learning, one that leverages multiple information sources. As we move forward in 2022, we can expect these models to become more complex and more efficient, though we are far from seeing them in action on a massive scale.
2. Responsible AI Becomes Standard
We’ve already discussed what it means to develop responsible AI algorithms, but we did so on a theoretical level. Fortunately, as more and more teams work on AI-powered solutions, a not-so-small group of researchers and advocates are pushing for the implementation of comprehensive sets of practices and guidelines for the development of responsible AI algorithms.
While that’s not precisely new, what we’ll see throughout 2022 is that those practices and guidelines will actually become a norm for AI development teams. What’s changed? First, there’s the growing presence of ethical engineering teams in both tech giants and startups, which will force development teams into considering ethical issues surrounding AI.
And then, there’s the increasing presence of regulatory frameworks with which AI developers will have to comply. I’m talking about regulations like E.U. ‘s proposed Artificial Intelligence Act or the groundbreaking New York law that mandates audits on AI solutions for hiring new team members. Sure, we’re far from the ideal scenario, but both of these situations show that at least we’re on the right track.
3. AI-Driven Development Boosts Productivity
AI brought huge news for developers in 2021. The launch of Github Copilot left everyone in awe of its pair programming capabilities because, even for its evident flaws, it felt like a game changer. Anyone paying attention to AI solutions for development teams knows that Copilot isn’t the only AI solution of its type, with tools from Amazon, Salesforce, and other companies also pursuing similar objectives.
These scattered efforts will finally gain traction during 2022, mainly because of the broader availability of open source code that allows embedding these AI solutions into different frameworks and development platforms. But that’s not all. While we might not see them in widespread action during this year, there already are ripe AI algorithms capable of translating code from one language to another.
Perhaps the most recognizable of these solutions is TransCoder, Meta’s (formerly Facebook) self-supervised neural transcompiler system. This solution leverages deep learning to successfully translate functions between C++, Java, and Python 3. It’s to be expected that this and other solutions become more common as the year progresses.
4. Ready-made AI Solutions Get Popular
Saying that huge tech companies like Amazon, Google, and Microsoft are offering ready-made solutions for businesses feels like old news. These (and other) companies have been in the game of providing AI solutions for a while now. The difference in what we can expect during 2022 is that more and more companies will start adopting these solutions not just in software form, in hardware as well.
On the one hand, businesses will start more massively adopting services like Google Contact Center AI and Amazon Connect, platforms that have built-in AI capabilities that can certainly boost customer support departments. Thanks to machine learning, these solutions can improve conversations through highly efficient bots to provide the ultimate automated assistance.
On the other hand, certain devices like Azure Percept and AWS Panorama will certainly start to pop up more frequently across businesses. Percept is a robust kit that provides computer vision and conversational AI capabilities for edge computing. For its part, Panorama uses computer vision inference processes to train new models in the cloud that can be later used on the edge. Both are but a couple of examples of turnkey AI hardware that’s set to revolutionize companies in 2022.
5. Reinforcement Learning Will Replace Supervised Learning as Paradigm
This may be a stretch as there are no guarantees that this will happen in 2022. However, more and more research teams are starting to turn to reinforcement learning as a way to replace supervised learning. So, instead of collecting and cleaning data to later feed it to an algorithm, teams will train their algorithms through an exploratory approach.
Reinforcement learning doesn’t provide algorithms with historical data. Rather, the idea is to give an objective to the algorithm and allow it to seek its own way towards it. The underlying principle is that the algorithm uses trial and error until it reaches its objective in optimal form, learning as it moves forward.
One of the key supporters of this paradigm is DeepMind, an AI subsidiary of Alphabet, whose AlphaGo solution has shown impressive results using reinforcement learning. We can expect that such a company, backed up by none other than one of the biggest behemoths in the tech world, will push for this paradigm shift, which will undoubtedly continue to unfold during 2022.
These are far from being the only AI trends in 2022. However, they are the most likely to become widespread during the following 12 months. As always, we’ll have to wait to see whether these predictions turn out to be true. Yet, you can count that AI will keep growing throughout the year on its way to becoming the revolutionary technology everyone expects it to be.