Multimodal learning: The future of artificial intelligence
Artificial intelligence (AI) not only helps businesses operate more efficiently, but also generates critical insights businesses use to improve. AI has proliferated through the enterprise, becoming a crucial component of successful businesses. This success fuels AI’s continued growth, with AI devices predicted to increase by 1.8 billion in the next five years, ABI Research found.
ABI Research’s blog post, Multimodal learning and the future of artificial intelligence, outlined the impact and future of AI in business. Currently, AI devices work independently of one another, with high volumes of data flowing through each device. As AI continues developing, these devices will be able to powerfully work in accordance with one another, unveiling the full potential of AI, according to the post.
The action of consolidating independent data from various AI devices into a single model is called multimodal learning. Multimodal systems are able to process multiple datasets, using learning-based methods to generate more intelligent insights, the blog post said. Rather than having to separately analyze data from different devices and draw conclusions, a multimodal system automatically does the work.
The blog post identified the following two main benefits of multimodal learning:
- Multiple sensors observing the same data can make more robust predictions because detecting changes in it may only be possible when both modalities are present.
- The fusion of multiple sensors can facilitate the capture of complementary information or trends that may not be captured by individual modalities.
In an effort to break out of AI silos, organizations are open to embracing multimodal learning. The cost for developing multimodal systems isn’t overwhelming, as the hardware sensor and perception software market landscape is very competitive, the blog post said.
Some of the most well-known multimodal platforms include IBM Watson and Microsoft Azure, but most organizations have only focused on the expansion of unimodal systems. A gap between supply and demand exists, the blog post said, giving platform companies a big opportunity to enter the field. Multimodal learning also presents opportunities for chip vendors, whose skills will be beneficial at the edge.
Multimodal use cases
Use cases for multimodal applications span across industries, according to the post. In the automotive industry, Advanced Driver Assistance Systems (ADA), In-Vehicle Human Machine Interface (HMI) assistants, and Driver Monitoring Systems (DMS) are all being introduced to multimodal systems.
Robotics vendors are integrating multimodal systems into robotics HMIs and movement automation, the post said. Consumer device companies are seeing multimodal system applications in security and payment authentication, recommendation and personalization engines, and personal assistants.
Medical companies and hospitals are in the early stages of adopting multimodal learning techniques, but promising applications exist with medical imaging. The media and entertainment industries are also beginning to adopt multimodal learning with content structuring, content recommendation systems, personal advertising, and automated compliance marketing, the post said.
Until more companies publicly adopt this way of operating, multimodal learning systems will remain unfamiliar to most people. However, the future of AI is heading in the multimodal direction.