Editing Responsible AI and AI Safety (section)

== <span style="color: #FFFFFF;">Understanding</span> ==
Responsible AI and AI safety address different but related concerns:

'''Responsible AI''' focuses on immediate, concrete harms that AI systems cause today:
* Biased hiring tools that discriminate against certain demographics
* Medical AI that performs worse on underrepresented patient populations
* Credit scoring algorithms that reinforce historical inequalities
* Surveillance systems that enable authoritarian control
* Deepfakes that destroy individuals' reputations

'''AI Safety''' focuses on risks that grow with AI capability:
* Near-term: AI systems that fail in high-stakes environments (autonomous vehicles, medical diagnosis, financial systems)
* Medium-term: AI systems that pursue misspecified objectives in harmful ways
* Long-term: the possibility of highly capable AI systems that pursue goals misaligned with human values at civilizational scale

The underlying challenge of both is the '''alignment problem''': ensuring AI systems do what we actually want, not just what we literally specified. This is harder than it sounds because human values are complex, contextual, and sometimes self-contradictory.

'''Sources of AI bias''': Bias enters AI systems through multiple channels:
* Historical bias in training data (e.g., facial recognition trained mostly on light-skinned faces)
* Measurement bias (e.g., using arrest records as a proxy for criminal behavior when arrest rates vary by race)
* Aggregation bias (using one model for diverse populations with different characteristics)
* Feedback loops (biased predictions influence real-world outcomes, which become future training data)
</div>

<div style="background-color: #8B0000; color: #FFFFFF; padding: 20px; border-radius: 8px; margin-bottom: 15px;">