Can We Fix AI? (peterbloem.nl)

🤖 AI Summary
The recent discussion surrounding the ethical implications of AI highlights a significant dilemma: while generative AI models like ChatGPT and Claude improve remarkably, public sentiment toward them remains polarized. Critics express ethical concerns over issues such as copyright infringement and the potential for misuse, particularly in safety-critical applications like code review. The essay argues that AI should not only provide assistance in spotting vulnerabilities but do so without leveraging controversial practices like training on unauthorized data. This raises the question of whether AI can be reengineered to eliminate these objectionable aspects while maintaining effectiveness. Exploring this possibility, the author introduces the concept of building AI systems using ethically sourced data—specifically, a "Common Pile" of openly licensed or public domain resources. This initiative aims to create a foundation for AI training that respects creators' rights and mitigates feelings of betrayal among content producers. The essay signals the beginning of a broader inquiry into the ethical boundaries of AI development, urging the AI/ML community to reassess current practices and strive for solutions that prioritize ethical integrity alongside technological advancement.
Loading comments...
loading comments...