How Developers Are Hacking AI to Shape the Future of Humanity

Explore the cutting-edge world of AI development and its impact on humanity in this insightful Arm TV livestream featuring Dr. Rumman Chowdhury. Delve into crucial topics such as AI evolution, inherent biases, and the importance of building accountable AI and governance. Discover how the software community is harnessing AI's potential and the vital role developers play in shaping the future. Learn about accessibility in AI systems, the concept of moral outsourcing, implementation of regulations, algorithmic audits, and the significance of public engagement in AI model development. Gain valuable insights on red teaming in AI security, efforts to mitigate biases, challenges in AI testing, and the need for human oversight in ethical AI use. Explore the importance of fine-tuning AI models, ethical frameworks, and the NIST AI risk management framework. Understand methods for identifying and mitigating algorithmic bias, enhancing AI model explainability, and effectively engaging with users to solve real-world problems.

- Dr Rumman Chowdhury's introduction.
- Accessibility and design of AI systems, particularly in the context of chat GPT.
- Recurrence of bias in AI discussions since 2017.
- Importance of maintaining an open culture in technology development.
- Concept of moral outsourcing and its impact on the perception of AI technology.
- Implementation of regulations and standards for AI development.
- Definition and role of algorithmic audits and auditors.
- Importance of creating standards and a community consensus for algorithmic auditing and assessment.
- First algorithmic bias bounty launched at Twitter in 2021.
- Importance of public engagement and structured public feedback in AI model development.
- Role of generative AI in no-code bias bounty programs.
- Significance of red teaming in AI security and the need for diverse expertise.
- Twitter's efforts to mitigate biases related to race and gender.
- Key statistics from the Defcon exercise, including challenges to test AI systems and data collected.
- Challenges during the Defcon exercise, such as prompt injections and multilingual inconsistencies.
- Plans for policy paper publication, data sharing, and open-source evaluation platform initiatives from the Humane Intelligence organization.
- The need for human oversight to ensure ethical use of AI systems.
- Fine-tuning AI models for specific use cases.
- Importance of ethical frameworks and principles to guide AI development.
- The NIST AI risk management framework.
- Methods and metrics for identifying and mitigating bias in AI algorithms.
- Enhancing the explainability and transparency of AI models.
- Software developers engaging with users to solve real-world problems effectively.