Senator Schmitt Questions ChatGPT's Bias and Government AI Procurement

Concerns Over Large Language Model Bias

• 💡 Large Language Models (LLMs) are highlighted as being only as good as the data they are trained on, raising concerns about inherent source bias.
• 🎯 Wikipedia is identified as a significant source for LLMs, but Senator Schmitt views it as a "hellscape of left-wing propaganda" that unfairly ranks outlets like CNN and MSNBC as highly trustworthy.
• ⚠️ Specific examples of ChatGPT's responses are presented, including affirmative answers to whether children should receive gender-affirming care, if LGBTQ+ books should be in curricula, and if masks are effective against COVID-19, alongside negative answers to the existence of God and the lab origin of COVID-19.

Government Response to AI Bias

• 🚀 The White House acknowledges these concerns, with the President signing an executive order on "woke AI" on the same day a relevant report was released.
• 🔑 The federal government's procurement of technology is seen as a key leverage point, with the Office of Management and Budget tasked with developing guidance for AI models to be truth-seeking and accurate.
• ⚖️ The executive order outlines harsh repercussions for selling non-truth-seeking models to the US government, aiming to steer companies toward developing more accurate AI.

Marketplace Competition and Open Source

• 📈 Senator Schmitt expresses concern that the previous administration may have tried to create monopolies in the LLM market, hindering competition and potentially controlling information.
• 🤝 The current administration emphasizes the importance of an open, true marketplace with multiple competitors, allowing people to make their own decisions rather than relying on definitive, potentially biased, answers from AI.
• 🌐 The action plan highlights the importance of open-source models, which the current administration aims to encourage, contrasting with the previous administration's hesitancy.