An AI-powered system for training and assessing IT engineers

今年2月，OpenAI的首席执行官Sam Altman承认ChatGPT存在“偏见方面的缺点”。“这意味着人工智能模型有时可能会因其训练而产生偏见或偏见的反应。ChatGPT的公式就是这样的一个例子，它将一个好员工描述为一个白色男性，或者当我们要求为一个聪明的软件工程师写一封推荐信时，它的默认假设是一个男性工程师，并使用\[He/Him\]代词。在Uptime Labs，我们专注于开发一个模拟平台，以培训IT工程师和经理有效地应对系统故障。该计划旨在解决IT中断恢复时间过长的问题，从而影响企业的利润。工程师必须在工作中学习，而这是以雇主、客户和心理健康为代价的。我们的平台是世界上第一个逼真地模拟各种场景的平台，类似于安全关键行业，以帮助员工发展肌肉记忆并改善响应时间。投资模拟培训可以提高处理复杂问题的效率和有效性，主动识别弱点，并最大限度地减少停机的经济影响。为了创建可扩展的解决方案，使我们的产品可以在整个行业轻松访问，我们为我们的模拟器提供了会话和生成AI。考虑到多样性，我们设计了数字“队友”与从业者互动，并在现实环境中合作解决某些预定义的事件。为了确保我们提供最佳体验并避免偏见，我们的现场专家监督会议并审查我们的AI模型生成的答案。从第一天起，我们就实施了人工智能护栏，以创造一个心理安全的环境，让所有从业者，无论性别或种族，都可以茁壮成长并有效地做出贡献。我们的人工智能系统根据从业者的表现进行评估，提供反馈，并为客户设计学习路径。最终，我们的目标是完全自动化我们的培训程序和评估，使其更容易在整个行业中采用。为了从对客户的监督培训转向无监督培训，我们必须在人工智能模型中加入大量的专家知识。这个项目将使我们能够在六个月内雇佣更多的资源和专家，专注于根据我们从模拟中获得的无偏见数据构建人工智能模型。我们相信，我们的创新将大大有助于增加创新，提高员工参与度，改善协作，增强解决问题的能力，减少人员流动和缺勤。

In February, OpenAI's CEO, Sam Altman, acknowledged that ChatGPT has "shortcomings around bias." This implies that the AI model may sometimes produce biased or prejudiced responses due to its training. An example of such a response is ChatGPT's formula, which describes a good employee as a white male, or when we ask to write a recommendation letter for an intelligent software engineer, it's default assumption is a male engineer and uses \[He/Him\] pronouns.At Uptime Labs, we focus on developing a simulation platform to train IT engineers and managers to respond effectively to system failures. The initiative aims to address the prolonged recovery time of IT outages, affecting businesses' bottom lines. Engineers have to learn on the job at the cost of their employer, their customers, and their mental health. Our platform is the first in the world that realistically simulates a diverse set of scenarios, similar to safety-critical industries, to help personnel develop muscle memory and improve response times. Investing in simulation training can improve the efficiency and effectiveness in dealing with complex problems, identify weaknesses proactively, and minimize the economic impact of outages.To create a scalable solution where our product could be easily accessible across the industry, we have powered our simulator with conversational and generative AI. Having diversity in mind, we have designed digital 'teammates' to interact with practitioners and collaborate to resolve certain predefined incidents in a realistic environment. To ensure we deliver the best experience and avoid bias, our field experts supervise the sessions and review the generated answers by our AI models. From day one, we've implemented AI guardrails to create a psychology-safe environment where all practitioners, regardless of gender or ethnicity, can thrive and contribute effectively. Based on the performance of the practitioners, our AI system runs assessments, provides feedback and designs a learning path for the customer.Ultimately, we aim to fully automate our training procedures and assessments to make them easier for adoption across the industry. To move from supervised to unsupervised training of our customers, we must bake in significant amounts of expert knowledge in our AI models. This project will enable us to hire more resources and experts over six months to laser focus on building AI models on unbiased data we acquire from the simulations. We believe our innovation will significantly contribute to Increased Innovation, Higher Employee Engagement, Improved Collaboration, Enhanced Problem-Solving, and Reduced Turnover and Absenteeism.