{"type":"rich","version":"1.0","provider_name":"Transistor","provider_url":"https://transistor.fm","author_name":"Make an IIIMPACT - The User Inexperience Podcast","title":"Top 5 Gotchas in AI","html":"","width":"100%","height":180,"duration":3421,"description":"Hey there listeners, Makoto here, and welcome to another insightful episode of Make an IiIMPACT! Today, we're diving deep into the world of AI integration and exploring the five crucial gotchas you need to watch out for. \nJoining me are my co-hosts Brynley Evans and Joe Kraft, who bring a wealth of experience and knowledge to the table. \nIn this episode, we'll be discussing a range of topics that are essential for anyone looking to harness the power of AI in their software.\nFirst, we'll discuss the concept of tokens as a measurement of AI currency and the importance of cost management when using tokens in AI models. We'll also touch on the cost structure of using AI models like Chat GPT and Microsoft Copilot and emphasize the need for budgeting and monitoring token usage to avoid unexpected costs.\nNext, we'll delve into the crucial role of data diversity in preventing bias in AI decision-making, and the necessity of monitoring and auditing data sources to ensure transparency. We'll also discuss the significance of inclusive design in avoiding biases and the concerns surrounding ethics, transparency, and bias in AI developments.\nWe'll then move on to the topic of adversarial attacks and how AI models can be tricked by intentionally designed input data, leading to misinformation and affecting business operations. We'll draw comparisons to jailbreaking and discuss the threats posed by manipulating data sources and introducing malicious data during AI model training.\nFurthermore, we'll explore the risks of model inversion attacks, where attackers can reverse engineer and infer sensitive information from the training data of AI models created by other companies. We'll also discuss the importance of privacy and data protection when using AI services, and the potential for bias and discriminatory patterns in AI models if left unchecked.\nLastly, we'll touch on the significance of quality training data, the challenges of caching answers for repeated questions, and the use...","thumbnail_url":"https://img.transistorcdn.com/IISy3LPjQLyTkoKT1vUQFVYWUlGQayk7I7J_LuexY2o/rs:fill:0:0:1/w:400/h:400/q:60/mb:500000/aHR0cHM6Ly9pbWct/dXBsb2FkLXByb2R1/Y3Rpb24udHJhbnNp/c3Rvci5mbS83NzQ4/M2M0MmZkZmJjNWFk/ZmU3YzM5MTFhMjE5/ZDZmOC5wbmc.webp","thumbnail_width":300,"thumbnail_height":300}