The Road to Autonomy: Progress is Not Enough

Progress in AI and automation, like in driverless cars, often hits an asymptote where reaching mere 80% accuracy is misleading. While initial advancements may appear rapid, the challenge lies in navigating the vast array of outlier cases not included in training data. These exceptional situations often lead to critical failures that are unacceptable in contexts like medicine and autonomous driving. There exists a tendency to overstate current capabilities, creating a culture of 'fake it until you make it,' which obscures the true complexity and gaps in technology. The distinction between human judgment and machine errors highlights the ongoing struggle toward achieving general-purpose autonomy and genuine artificial general intelligence (AGI).