User Review( votes)
By teaching machines to understand our true desires, one scientist hopes to avoid the potentially disastrous consequences of having them do what we command.
Artificial Intelligence Problem- The danger of having artificially intelligent machines do our bidding is that we might not be careful enough about what we wish for. The lines of code that animate these machines will inevitably lack nuance, forget to spell out caveats, and end up giving AI systems goals and incentives that don’t align with our true preferences.
A now-classic thought experiment illustrating this problem was posed by the Oxford philosopher Nick Bostrom in 2003. Bostrom imagined a superintelligent robot, programmed with the seemingly innocuous goal of manufacturing paper clips. The robot eventually turns the whole world into a giant paper clip factory.
Such a scenario can be dismissed as academic, a worry that might arise in some far-off future. But misaligned AI has become an issue far sooner than expected.
The most alarming example is one that affects billions of people. YouTube, aiming to maximize viewing time, deploys AI-based content recommendation algorithms. Two years ago, computer scientists and users began noticing that YouTube’s algorithm seemed to achieve its goal by recommending increasingly extreme and conspiratorial content. One researcher reported that after she viewed footage of Donald Trump campaign rallies, YouTube next offered her videos featuring “white supremacist rants, Holocaust denials and other disturbing content.” The algorithm’s upping-the-ante approach went beyond politics, she said: “Videos about vegetarianism led to videos about veganism. Videos about jogging led to videos about running ultramarathons.” As a result, research suggests, YouTube’s algorithm has been helping to polarize and radicalize people and spread misinformation, just to keep us watching. “If I were planning things out, I probably would not have made that the first test case of how we’re going to roll out this technology at a massive scale,” said Dylan Hadfield-Menell, an AI researcher at the University of California, Berkeley.