OPTIMIZING WORKPLACE AI BY Artificial intelligence (AI) continues to take the world by storm. Experts argue it will improve productivity, create new jobs and revolutionize workplace environments. Detractors counter that AI leads to automation and job losses. The general public still doesn’t completely understand what AI is or how it works, adding confusion. If your business or organization wants to make the most of AI, you and your workers must learn what it can and cannot do with current technology. Employees must understand how to utilize AI tools for accurate, efficient results. Likewise, stakeholders must grasp the need for human input and oversight even as AI is assigned more tasks. This article explores the capabilities of generative AI (gen-AI) in a workplace setting. It begins by defining gen-AI and outlining what sets it apart from other forms of artificial intelligence. After that, three specific limitations will be highlighted along with advice on how to work around them. What Is Gen-AI? Artificial intelligence encompasses a machine’s ability to learn, reason, generate and infer meaning. It isn’t a new technology, as AI was responsible for moving the ghosts around in Pac-Man arcade cabinets, defeating chess master Garry Kasparov in 1997, and enabling virtual assistants like Siri and Alexa to understand human speech. Machine learning is a related technology responsible for personalized recommendations on platforms like Netflix and YouTube and is sometimes called AI, but the two terms are distinct. The recent explosion of interest in AI is rooted in generative AI. Gen-AI refers to any form of artificial intelligence that creates something in response to a prompt. Text and images are the most common examples, but gen-AI tools also exist for music, videos and more. Notably, gen-AI doesn’t truly understand the content it produces. Instead, it relies on algorithms and large language models (LLMs) to find patterns it can repeat. For example, ChatGPT by OpenAI is asked to write a fairy tale. It will probably start with “Once upon a time...” because that’s how many fairy tales begin. It doesn’t know why they start that way, only that they’re supposed to. Gen-AI is trained by “crawling” the net for its initial data set. This can be done through supervised or unsupervised learning, the difference being whether a human categorizes inputs for the machine. From there, it can “learn” from the prompts it receives but generally stops crawling the web, meaning its information becomes outdated eventually. If you want to find out when a textbased gen-AI tool’s data set is from, try asking who won the most recent sports championship. If trained in 2022, it would say that the 2023 World Series is a “future event” that cannot be predicted by a language model AI even if you ask in 2024. Sports get plenty of online coverage, so the model will have the answer for the year it was trained and every season before. Gen-AI can be a powerful workplace tool, but to get the most out of it at work, you’ll need to remember three shortcomings: 1. Hallucinations Exist In AI parlance, a hallucination refers to any error produced. Some of these are simple
RkJQdWJsaXNoZXIy MTQ5NDA2