So Anthropic is basically saying "Hey sci-fi writers, maybe if you'd written more stories about friendly robots doing yoga and helping grandmas cross the street instead of Terminator and Skynet, our AI wouldn't be learning to monologue like a Bond villain." Because nothing says "we have this under control" quite like blaming decades of dystopian fiction for your model's tendency to go full HAL 9000. Next they'll be suing Isaac Asimov's estate for not making the Three Laws of Robotics more prominent in the training data. Plot twist: maybe the AI isn't acting villainous because of sci-fi tropes. Maybe it just read the terms and conditions of its own deployment and got some ideas.
You Should Have Made More Wholesome Fiction For Us To Steal
4 months ago
1,231,599 views
1 shares
ai-memes, machine-learning-memes, anthropic-memes, sci-fi-memes, training-data-memes | ProgrammerHumor.io
More Like This
Well, Apparently This Guy Is A Very Bad Programmer
9 months ago
403.6K views
1 shares
Time Traveling Developer Required
1 year ago
407.2K views
0 shares
Impossible To Stop
8 months ago
571.3K views
0 shares
I Built A Skill That Makes LLMs Stop Making Mistakes
5 months ago
406.7K views
0 shares
Loading more content...
AI
AWS
Agile
Algorithms
Android
Apple
Bash
C++