Before Thanksgiving in the US, let’s do some big ideas to contemplate on the couch after all that turkey.
I covered that story about doctors and their use of AI – One Useful Thing goes deeper on why. A recent paper has highlighted the challenges doctors face when using artificial intelligence, specifically the model GPT-4, for diagnosing diseases. Despite the model’s impressive accuracy, doctors using AI performed no better than their peers who did not, and both groups were outperformed by the AI alone. This phenomenon is attributed to “algorithmic aversion,” where doctors often disregard AI recommendations that conflict with their judgment. The study indicates that most users struggle with effectively utilizing AI, often treating it like a search engine rather than a comprehensive diagnostic tool. Experts suggest that learning to use AI for specific tasks can enhance its utility, emphasizing the importance of spending about ten hours engaging with the technology to develop a better understanding of its capabilities. The key takeaway is that while AI can be a powerful ally, effective prompting and open-mindedness in its application are crucial for maximizing its potential in medical practice.
In the debate about human vs AI, In a recent discussion about the future of work, computer scientist Geoffrey Hinton warned that artificial intelligence could soon outperform humans across various occupations, potentially leading to widespread unemployment and rising inequality. He emphasized that while AI can replicate many cognitive tasks, it cannot replace the unique human traits that are valued in the workplace, such as vulnerability and social connections. The article outlines seven significant advantages humans hold over AI, starting with the fact that approximately eighty-four million jobs in the United States require physical presence and cannot be effectively performed remotely. These include roles in food preparation, healthcare, and education. Furthermore, while automation may eliminate some positions, it often creates new ones, as seen in food preparation facilities where robots complement human workers. The article concludes that the ability to form genuine relationships, the independence of human judgment, and the scarcity of human attention will ensure the continued importance of human labor in the economy, despite the advancements in AI technology.
With a title like “AI Has Enshittified America’s Advanced Stealth Fighter”, I had to know more. A recent report from the Office of the Director, Operational Test and Evaluation highlights significant issues with the F-35 stealth fighter jet, particularly its maintenance software known as the Autonomic Logistics Information System. The report reveals that the system, initially pitched as a user-friendly maintenance hub, has instead created numerous challenges, including high false alarm rates that lead to unnecessary maintenance tasks. Despite promising to streamline operations, the software has been criticized for its poor usability and has led to distrust among maintenance crews. Lockheed Martin is attempting to replace this system with a new solution called the Operational Data Integrated Network, but it has yet to be tested. This comes amid a backdrop of ongoing problems with the F-35, which has faced scrutiny for its performance and reliability, raising concerns as the Pentagon prepares to invest trillions in its continued production and support.
Why do we care?
Ask yourself these questions.
How might organizations and individuals be coupled with AI technologies to overcome algorithmic aversion?
How can Hinton’s insights provide some of those logical steps?
And what lessons of overcomplexity can we take from the F-35 to drive to better results?