4 of the Most Recent Advancements in AI You Need to Know
Artificial intelligence (AI), including machine learning, natural language processing (NLP), and generative AI dominated the headlines in 2023 and became much more accessible for organizations and individuals thanks in large part to programs like ChatGPT, which is used by more than 90 percent of Fortune 500 companies. The technology has proven effective in a wide range of industries, including healthcare and manufacturing, and is expected to become more integrated into business and daily life in the future.
Statista, for example, projects the AI software market to reach more than $125 billion in revenue by 2025, at which point the technology could be used in 95 percent of consumer interactions, according to Servion Global Solutions. AI is constantly evolving and, while humans have yet to realize its full potential, there have been some major advancements in the past year that highlight significant progress in that regard.
Here are five of the most significant recent advancements in AI.
AI Robots Learning by Observing Humans
Machine learning is the process by which AI “learns” via human-directed data processing. This is the same process used by some mobile devices in navigation and weather tracking. For example, a bot might notice that the person who owns the device visits the same location, i.e. work, at the same time most days and provides an estimated driving time based on traffic and weather conditions. Now, researchers are working on new and more advanced ways to teach robots to learn through observation. This could have massive implications for the future of AI and human-robot coexistence.
Researchers at Carnegie Mellon University’s (CMU) School of Computer Science have been exploring the idea of robots learning from human observation for several years. Their latest observational learning model, known as Vision-Robotics Bridge (VRB), has been relatively successful in training robots to complete household tasks, such as opening drawers and picking up objects. Robots with VRB capabilities viewed videos of humans performing these tasks and were able to replicate them within 25 minutes of practice.
Observational learning isn’t an entirely new concept in AI, but past models required humans to be present in the same environment for the robot to execute the desired task. The VRB model, however, allows the robot to learn without a human present and, according to Shikhar Bahl, a CMU Ph.D robotics student, “could enable robots to learn from the vast amount of Internet and YouTube videos available.”
AI in Caregiving
AI is already dramatically improving healthcare, as the technology is being used to optimize patient treatment plans by analyzing their genetics and health history. It has also enhanced medical imaging algorithms to improve disease diagnoses accuracy through CT scans, MRIs, and X-rays. With recent advancements in AI technology and public acceptance, we might not be too far away from AI-powered robots being commonplace in assisted living facilities.
Japan is at the forefront of AI-backed nursing care. The country has a significant shortage of caregivers, and researchers have been working on developing robots aimed at complementing the work of human staff for more than a decade. While they have had varying degrees of success, a 2023 MIT Technology Review study suggests there’s still plenty of work to be done before these robots can provide a level of care similar to humans.
However, conversational AI bots in Japan have been effective in helping people with dementia and alleviating feelings of loneliness and isolation among seniors. The University of British Columbia in Canada recently initiated a study to explore the effectiveness of these robots, which can adapt to their environment, recognize voices and faces, and give hugs.
ChatGPT Evolving
ChatGPT is a powerful tool used for a variety of applications, including translating and summarizing text and creating new content. As an AI program, it is constantly evolving to become even more efficient and effective in all areas. OpenAI, which developed the tool, released an updated ChatGPT-4 model in 2023. In addition to having a 10-times greater database, ChatGPT-4 is better than the prior ChatGPT-3.5 model at understanding context, 40 percent more accurate in its responses, and more creative, with the ability to generate art, music, and essays with simple prompts.
OpenAI is already working on producing ChatGTP-5, which is expected to be more reliable, with enhanced personalization and multi-modal outputs.
Explainable AI
Until recently, AI algorithms could produce accurate predictions, but without any insight or explanation into those results. Now, developers have created tools and frameworks, known as explainable AI (XAI) to demonstrate transparency in these predictions, allowing users to better understand and more accurately interpret machine learning models. Explanations can be displayed in easy-to-understand illustrations, narratives, or interactive user interfaces.
Google offers free-to-try XAI tools that can help users create better and more transparent models to mitigate data challenges. XAI is especially important in the business world. A McKinsey study found that, among companies with more than 20 percent of EBIT attributed to AI use, those with the greatest bottom-line returns were more likely to adhere to best practices enabling explainability.
