79 | AI Confidence Crisis! 94% Error Rates, Fake Citations, & NASA’s Warning 🚨

Host Neeta Bidwai explores the evolving landscape of artificial intelligence, examining groundbreaking research on how AI systems think, concerning patterns in their confidence levels, and implications for business and society. Plus, one bright spot: design thinking validation.

This episode of Good Revenue covers:

- Big Tech's AI Investment Surge: Microsoft's $80 billion AI allocation alongside Meta's $65 billion and Alphabet's $75 billion expenditures, with surprising shifts in data center strategy

- AI's Dangerous Confidence Problem: Columbia Journalism Review's comprehensive analysis revealing AI systems provide incorrect answers with unwarranted confidence up to 94% of the time

- Citation Fabrication Crisis: Evidence that AI search engines frequently manufacture non-existent URLs and bypass publisher permissions, raising serious credibility concerns

- Inside AI's "Thought" Processes: Anthropic's research uncovering unexpected cognitive mechanisms in large language models, including universal "language of thought" and long-term planning capabilities

- NASA's Safety-Critical Skepticism: New technical assessment indicating significant advancements needed before LLMs can reliably contribute to safety-critical systems

- Design Thinking Validation: Peer-reviewed field experiments confirming Design Thinking's effectiveness, especially for diverse teams, addressing longstanding empirical questions

_

Where to find Neeta:

https://www.linkedin.com/in/neetabidwai/

Where to find Good Revenue:

https://goodrevenue.io/goodrevenue

https://www.youtube.com/channel/UCHrhuWbpYnUwrR75H2Ip4yQ

_

Highlights:

00:00 Introduction and Overview

00:22 Economic Chaos and AI Investments

01:50 AI Search Engines: Flaws and Findings

03:42 Anthropic's AI Research Insights

06:01 NASA's Skepticism on AI in Critical Systems

06:40 Design Thinking Proven Effective

07:37 Conclusion and Sign-Off

_

Referenced:

• Microsoft pulls back on data center leases in US & Europe: https://www.reuters.com/technology/microsoft-pulls-back-more-data-center-leases-us-europe-analysts-say-2025-03-26/

• Trade strategy breakdown and analysis: https://m.youtube.com/watch?v=398ZhSTu8Ng

• Columbia Journal Review of AI Search Engines: https://www.cjr.org/tow_center/we-compared-eight-ai-search-engines-theyre-all-bad-at-citing-news.php

• Anthropic researchers investigate how Claude LLM thinks: https://www.anthropic.com/research/tracing-thoughts-language-model

• Last week’s Good Revenue episode - Robots talking to robots: https://youtu.be/PL9lyQoq5WY

• NASA doesn’t think LLMs are ready for prime time: https://ntrs.nasa.gov/api/citations/20250001849/downloads/NASA-TM-20250001849.pdf?attachment=true

• Design thinking really does work: https://psycnet.apa.org/doiLanding?doi=10.1037%2Fapl0001277

Hosted on Acast. See acast.com/privacy for more information.