As artificial intelligence continues to rapidly evolve, experts are urging caution and critical thinking when using generative AI tools. At this week's South by Southwest (SXSW) conference in Austin, Texas, Sherry Wu and Maarten Sap, assistant professors of computer science at Carnegie Mellon University (CMU), shared valuable insights on how to navigate the world of AI chatbots and large language models (LLMs) effectively."They're actually far from perfect and not actually suited for all the use cases that people want to use them for," Sap noted, emphasizing the importance of understanding the limitations of these technologies.Here are five key takeaways from their talk, offering guidance on how to stay sharp and avoid potential pitfalls when interacting with generative AI:1. Clarity is Key:When communicating with AI models, precision is paramount. Unlike humans, LLMs struggle with interpreting sarcasm, nuance, and social cues. To get the desired results, users must be explicit and structured in their prompts.Wu advised, "Make sure the model knows what you're asking it to produce. Focus on what exactly you want, and don't assume the LLM will extrapolate your actual question."2. Confidence Doesn't Equal Accuracy:One of the most significant challenges with generative AI is its tendency to "hallucinate," or fabricate information. According to Sap, LLMs can generate inaccurate responses up to 25% of the time, with even higher rates in specialized fields like law and medicine.The issue is further compounded by the fact that chatbots often present their answers with unwavering confidence, making it difficult to discern fact from fiction. To combat this, users should always verify the information provided by AI models through external sources.Wu recommends cross-checking the AI's responses by asking the same question multiple times or in different ways. "Sometimes you will see that the model doesn't really know what it is saying," she cautioned. It's particularly important to exercise caution when seeking answers to questions outside of your own expertise.3. Privacy Matters:LLMs are not designed to keep secrets, raising significant privacy concerns. Sharing sensitive or personal information with an AI model could lead to unintended disclosures. In one demonstration, ChatGPT inadvertently revealed a surprise party plan to the person who was supposed to be surprised.Wu stressed the importance of carefully reviewing any information before sharing it with an LLM. "Whenever you share anything produced by you to the model, always double-check if there's anything in that that you don't want to release to the LLM," she advised.4. Machines Aren't Human:The ability of chatbots to mimic human speech has contributed to their widespread adoption. However, it's crucial to remember that these are machines, not people. LLMs are trained on vast amounts of text data, enabling them to generate human-like responses, but they lack genuine understanding or consciousness.Sap cautioned against anthropomorphizing AI models, as this can lead to misplaced trust and the reinforcement of social stereotypes. "Humans are much more likely to over-attribute human-likeness or consciousness to AI systems," he said.5. Consider the Alternatives:Despite the hype surrounding LLMs, they may not always be the best tool for the job. While benchmarks may suggest that certain models can perform at a human expert level, these tests may not accurately reflect real-world performance.Sap warned against making hasty decisions based on the perceived capabilities of AI, noting that "there's this illusion of the robustness of AI capabilities going around that leads people to make rash decisions in their businesses."Before using a generative AI model, it's essential to carefully weigh the potential benefits and harms against those of not using it. By making conscious decisions about when to rely on AI and when to seek alternative solutions, users can harness the power of these technologies while mitigating their risks.As AI continues to advance, critical thinking and informed decision-making will be essential for navigating this evolving landscape responsibly.
