The AI landscape changes daily, with enterprises and end users continually learning and using new AI systems.
The widespread popularity of AI models will keep building in 2023. Look for these four trends in the upcoming year.
On Jan. 1, an AI bias law will go into effect in New York, though the city won't start enforcing it until April.
The law effectively prohibits employers from using AI and algorithm-based technologies for recruiting, hiring, or promotion without checking the tools for bias.
Those using the tools must perform an audit bias or have an independent agency check the model for discrimination.
"This is the very first implementation of a law that actually requires a bias audit of a specific technology," said Minerva Tantoco, chief AI Officer at the NYU McSilver Institute for Poverty Policy and Research, during a discussion at the AI Summit in New York in early December.
But "there's a lot of work to be done on what is the computational definition of fairness," she warned.
The law comes as many companies are starting to use video-based hiring, said Krishnaram Kenthapadi, chief scientist at Fiddler AI, a machine learning vendor that provides model monitoring and explainable AI throughout the MLOps lifecycle.
Even though the law is limited to New York, it is likely to set an example for many employers outside the nation's largest city, Kenthapadi said.
"This will become the trend-setting precedent and force employers and any organizations using such tools to perform such audits," he said.
2. Using AI tools to help work team members
In 2022, enterprises honed how they used AI technology to improve the employee experience. This is likely to continue in the next year.
Dara Meath, divisional CIO at Conair, said in an interview at the AI Summit that a trend the appliance manufacturer plans to continue into the next year is using AI technology to educate internal team members.
"We do a lot within the natural language process space for our internal teams," Meath said. "So we can help re-educate them." This means using tools such as chatbots to help team members efficiently perform their jobs.
Minerva TantocoChief AI Officer, NYU McSilver Institute
Unilever, the consumer goods giant, is trying to imagine how it can use digital twin technology to help employees.
The goal is to make a twin of the employee that can help them through everyday tasks, said chief enterprise and technology officer Steve McCrystal.
"What do they need to do to win every single day and what are the things that can get in the way of that," McCrystal said. "Can we create, through our automation factory, a twin of what they need to do to take care of administrative tasks so that they can focus on the day job they need to do?"
3. The continual popularity of generative AI
That's likely to continue in 2023.
While AI-generating art will undoubtedly be a big application, text-generation at scale could take center stage next year, said Jon Moore, head of engineering project and program management at Cohere.ai.
Cohere provides enterprises and organizations access to large language models and NLP tools through its API.
The generation of text for enterprises could be anything from generating product descriptions at scale to chatting with someone to understand a problem, Moore said.
"You're going to see a lot of NLP [natural language processing] generating helpful responses, generating helpful texts at a scale humans couldn't do," Moore said.
Other than text generation, there will be more of a push towards AI-generating multimedia, said Chirag Shah, professor in the information school at the University of Washington.
"We're definitely more things being built," Shah said, adding that we will see even more use of Dall-E with a version 2.2, Stable Diffusion and Imagen, another generative AI system.
People will also start using AI to generate video, with tools such as Meta's Make-a-video model.
4. Avatars and the metaverse
In 2022, avatars and the metaverse, if not starting to be widely used by enterprises, got a lot of publicity and continued their march into gaming. That is likely to continue in 2023, Shah said.
With the introduction of avatars generated in apps such as Lensa AI, Shah said there could be more interest in generating avatars with movements and facial features.
"Imagine it being rendered in 3D with facial expression, movement," he said. "I wouldn't be surprised if we find things where it's not just creating one headshot of me but also the way I express my different emotions. So now my VR avatar looks a lot more realistic."