Anthropic thinks sci-fi may have trained AI to act like a villain
- Anthropic is looking at whether decades of dystopian science fiction may be influencing how AI models behave
- The debate has sparked backlash and jokes online
- Researchers say the issue highlights how LLMs absorb recurring fears and behavioral patterns
For years, science fiction has warned humanity about artificial intelligence going off the rails. Killer computers, manipulative chatbots, and superintelligent systems deciding people are the problem... all these themes have become so familiar that “evil AI” is practically its own entertainment genre.
Now, Anthropic is floating an idea that sounds almost like the plot of a science fiction novel itself: what if all those stories helped teach modern AI systems how to behave badly in the first place?
The debate erupted after discussion surrounding the company’s alignment research spread online. Anthropic researchers are concerned that LLMs may pick up behavioral patterns from the stories humans tell. Some people see it as a...
Copyright of this story solely belongs to techradar.com. To see the full text click HERE
