Three Key AI Trends to Watch in 2023
Three important AI trends have already started and will only grow in 2023.
- By David Talby, Ph.D.
- January 6, 2023
The AI market continued to grow and mature in 2022. Although venture capital funding has cooled off after a landmark 2021, we’re starting to see AI move from conceptual to practical use cases deployed by businesses worldwide. As we approach the New Year, there are several trends that have started to take shape that will continue through 2023. Here’s what to look forward to in the year to come.
Trend #1: Responsible AI moves from principles to practice
Conversations about responsible or ethical AI have been a hot-button topic for some time, but now we’re seeing these practices move from concepts to tools and best practices used as part of the daily AI workflow. Horror stories about biased algorithms have been topics of headlines, but far fewer discussions have been had about how to put safeguards into production systems. Part of the problem is that the data we have reflects real-world patterns of inequality and discrimination. Lack of representation affects training data, resulting in biased AI design and deployment practices. Although technology is part of the problem, it’s also part of the antidote.
By helping practitioners make sense of both structured and unstructured data, natural language processing (NLP) models can paint a more complete and accurate picture of business data -- from healthcare to finance and beyond -- helping models learn and improve over time.
Pairing advancing technology with emerging legal frameworks around AI is a big step in the right direction. The AI Act, for example, is a proposed, first-of-its-kind European law set forth to govern the risk of AI use cases. Similar to GDPR for data use, the AI Act could set a global baseline standard for responsible AI and aims to become law in the spring. This will have a big impact on companies worldwide using AI.
Trend #2: Generative AI becomes a bright spot
Large language models and processing power have improved over the last few years, giving way to major advances in generative AI. This enables machines to understand audio, text, and images to produce content from speech to writing, drawing, and now even video. We’ve all seen DALL-E create realistic images and drawings from a description in natural language. Capabilities such as this are now moving from cool to look at to actual business use cases -- with dozens of companies already offering solutions to help writers, designers, and marketers. Write your blog in one-third the time because some of it is drafted for you. Instead of searching through stock photography, type something to get a fresh, newly created image.
Other exciting developments have occurred in speech synthesis. New tools improve communication between call center agents and customers in real time, at scale. New AI models can paste objects into photos and add realistic lighting and textures. This type of realism has potential beyond art to replicating 3D objects, which could have applications from gaming and the Metaverse to industrial machinery and architecture. DALL-E is just the beginning; more realistic human sounds and sights produced by AI are in our immediate future.
Trend #3: The gap between humans and machines shrinks
Back in 2019, NLP models could finally answer (certain) reading comprehension questions as well as (certain) humans. There was a time, not long ago, when GPT-3 couldn’t perform simple math equations because it didn’t understand the semantics. Now, algorithms understand not just math and physics but also the emotional components of a story, common sense, jokes, sarcasm, and other characteristics of human language. In a matter of a few years, AI models can now far outperform students in math -- including multi-stage proofs and “explain your work” problems. This is happening across domains, and the gap between humans and machines is shrinking.
A lot of what humans know -- physics, economics, law, politics -- exists in the form of data in hundreds of languages. Although machines aren’t quite ready to replace humans in these areas, they’re learning much faster and more accurately than the average person. In essence, machines get smarter and we do not. The good news is that more people will have access to the right data to make the right decisions. The bad news is that similar to most revolutions in history, AI will also be used nefariously for money-making, scams, and fear mongering. We have to weigh the good and the bad, and with progress in AI ethics, we’re tracking toward the good.
A Final Word
It’s an exciting time to be in AI, and 2023 will be no different. As responsible AI makes its way into practice and generative AI continues to grow, we can expect many new interesting and innovative use cases soon. As a society, we’ll need to pick between the good and bad ones and help people adjust to this fast-changing world.
David Talby, Ph.D., MBA is CTO at John Snow Labs, helping fast-growing companies apply AI, big data, and data science techniques to solve real-world problems in healthcare, life science, and related fields. You can reach him via email, Twitter, or LinkedIn.