{"id":9470,"date":"2022-04-15T08:52:50","date_gmt":"2022-04-15T12:52:50","guid":{"rendered":"https:\/\/blogs.mathworks.com\/deep-learning\/?p=9470"},"modified":"2022-04-15T08:52:50","modified_gmt":"2022-04-15T12:52:50","slug":"current-trends-in-ai","status":"publish","type":"post","link":"https:\/\/blogs.mathworks.com\/deep-learning\/2022\/04\/15\/current-trends-in-ai\/","title":{"rendered":"Current Trends in AI"},"content":{"rendered":"This post is from guest blogger Paul Pilotte, Product Marketing, AI and Data Science and David Willingham, Product Marketing, Deep Learning<\/em>\r\n
<\/h6>\r\nThe world has changed significantly since we tracked the AI Trends for 2021 a year ago<\/a>. While some industries have cut back on AI investment in the near term, others have increased their investment above what they had forecasted, offsetting the loss. Many are using this time to invest in upskilling through remote learning, with AI-themed courses among the most sought after by the engineering and scientific community. The goal is to make sure they are primed and ready to take on more AI projects in 2022.\r\n
<\/h6>\r\nHere's a closer look at 4 current trends within AI.\r\n
<\/h6>\r\n\r\n\r\n\r\n
\r\n

1<\/strong><\/p>\r\n<\/td>\r\n

Collaboration between frameworks<\/strong><\/span><\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n\u00a0New trend materializing for 2022<\/i><\/strong>\r\n
<\/h6>\r\nTensorFlow, PyTorch, and MATLAB are three different deep learning frameworks which you can use to develop AI models. When developing a larger system that includes AI, engineering teams can be faced with the question: which framework to use? Choosing one becomes complicated, especially if no single framework can satisfy all the requirements of the system being developed: an engineering team may want to use a data preprocessing app in MATLAB as well as reuse a pretrained AI model created in TensorFlow.\u00a0 The solution to meeting these requirements is to have the frameworks interoperate with each other.\u00a0<\/span>\r\n\r\n
<\/h6>\r\n\r\n\r\n\r\n
Prior to 2021, interoperating was quite complicated, and often involved some level of time-intensive manual recoding. However, 2021 saw the maturation of interoperability advancements such as a direct <\/span>TensorFlow Converter for MATLAB<\/span><\/a>, which has reduced the need for manual recoding, opening up more interoperability workflows<\/a><\/span>. In 2022, we predict more collaboration between TensorFlow, PyTorch and MATLAB, through the understanding, adoption and expanding support of interoperability functionality.<\/span><\/td>\r\n\"\"<\/strong><\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\n
<\/h6>\r\n\r\n\r\n\r\n\r\n
\r\n

2<\/strong><\/p>\r\n<\/td>\r\n

AI aligns Engineering and IT<\/b><\/span><\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\nA 2021 prediction, continuing in 2022 and beyond<\/i><\/strong>\r\n
<\/h6>\r\nThe need to get value from AI models by getting them into production was a dominant theme in 2021.\u00a0 This proved not only the potential of AI models, but also the realization that the value of an AI Model is in production as part of an engineered system.\u00a0 Nearly every analyst in this space echoed this imperative.\u00a0 2021 also marked the year of a growing trend to incorporate DevOps to AI models so that IT and OT teams can take ownership of the deployment, monitoring, and lifecycle of AI models in operational systems. We expect this trend to continue to accelerate in 2022, fueled by many corporate digital transformation initiatives.\u00a0 If you're not already involved in close collaboration between R&D teams developing AI models and production teams responsible for operational systems, you likely will be in 2022.<\/span>\r\n
<\/h6>\r\nSee an example of this trend: Korea Institute of Energy Research Develops AI-Based Predictive Maintenance Models for Offshore Wind Power<\/a>\r\n
<\/h6>\r\n\r\n\r\n\r\n
\r\n

3<\/strong><\/p>\r\n<\/td>\r\n

No-code \/ Low-code<\/b><\/span><\/td>\r\n<\/tr>\r\n<\/tbody>\r\n<\/table>\r\nA trend that materialized in 2021; we expect this trend to continue in 2022 and beyond<\/i><\/strong>\r\n
<\/h6>\r\nAs the benefits of applying AI is becoming increasingly realized, many of the common techniques and workflows are being commoditized by vendors through no-code, low-code and auto-code workflows for AI modeling.\u00a0 The benefits of this are vast, first it broadens the community of AI users from a few data science and coding experts to a larger audience of people who have the domain knowledge to know where to apply AI, but not necessarily the coding skill set. Secondly it increases efficiency and scale through automated workflows.<\/span>\r\n
<\/h6>\r\nLike the data-centric AI movement, no-code\/low-code\/auto-code is not a new trend. For example, for over 20 years engineers have been using MATLAB and Simulink to auto-generate or auto-code, production quality code for embedded systems.\u00a0<\/span>\r\n
<\/h6>\r\nHere are highlights of the use of no-code\/low-code, auto-code by engineers using MATLAB and Simulink:<\/span>\r\n