"Large language models (LLMs) already contain vast amounts of information. They know about the physical world, human behaviors, our history, the nature of human jobs, and the behaviors of our pets. This stored information allows LLMs and other AIs to write books, make us laugh, fix computer code, get perfect scores on medical licensing exams, and file our taxes. LLMs, when given a physical body, are starting to autonomously navigate cities and hospitals,"
"The main challenge is how fast AI is improving. People have spent thousands of years developing systems for vetting and reputation. You trust your Uber driver because you can see their rating and ride history. Your family doctor (hopefully) has performed hundreds of successful procedures over years of training. You might trust a teacher because your school district hired them, presumably after extensive vetting. None of this exists yet for robots. A robot in your home or office could be a marvel or a liability."
Autonomous robots are gaining large-language-model intelligence that encodes extensive knowledge about the physical world, human behavior, jobs, history, and pets. These systems can generate creative and technical work, pass professional exams, and perform practical tasks when embodied. Embodied LLMs can navigate cities and hospitals, open doors, operate vehicles, hold conversations, and learn about nearby people. People form emotional attachments and protective behaviors when robot personalities change. Traditional trust mechanisms—ratings, professional training, and institutional hiring—are not yet established for robots. Rapid AI progress increases the risk that unvetted machines could become liabilities in homes, workplaces, and public spaces.
Read at Fortune
Unable to calculate read time
Collection
[
|
...
]