The Action Catalyst · Southwestern Family of Podcasts - Southwestern Family of Companies

Shared Wisdom, with Dr. Alex Pentland (AI, Technology, Business, Data Science)

·27 min·2 clips

Alex Pentland explains why building AI to give all the answers is wrong and how humans must stay in the loop.

This episode features Dr. Alex Pentland, a renowned MIT professor and computational scientist, discussing the intersection of AI, data science, and human society. Hosts from The Action Catalyst interview Pentland about his book, *Shared Wisdom*, and his vision for technology that serves communities. He argues the current approach to AI and global challenges is fundamentally miscast. Pentland traces data science to the 1840 census, describing it as community-level information that measures what social programs work. He distinguishes this from "surveillance capitalism," where companies collect personal data to drive sales. He explains his concept of the "living laboratory," where communities test policies, like in education, on a small scale to see what works. His current focus is on building AI systems that align with human nature rather than seeking to replace human judgment. He claims large language models are trained only on public internet chatter, lacking the feeling and context of real human experience, making their output akin to superficial party conversation. Pentland describes an experiment where company CTOs built individual AI "librarians" for each employee using internal manuals and newsletters to improve cultural awareness. He contrasts national AI strategies, noting Abu Dhabi invests in open-source infrastructure, while China focuses on practical AIs for logistics and governance to gain global systemic dominance. A surprising claim is that frontier AI models requiring billions are less impactful than simple, practical applications, exemplified by an AI from his lab that now assists nurses and supports one in thirty global births. He asserts that the primary near-term AI risk is not superintelligence but "AI-assisted crime" like deepfakes and bot armies, necessitating AI-powered defenses. Pentland predicts a shift where AI agents, like OpenAI's ChatGPT facilitating commerce, will dismantle the traditional advertising-based economy, especially for routine purchases. He envisions a future where local business clubs or consumer organizations develop their own "sovereign AIs" to pool knowledge and navigate markets. Current AIs are described as overly eager to please and admit error because they are trained on social media's conversational style, not on evidence or commitment. A key insight is that AI cannot understand rapidly evolving human context, citing how Twitter sentiment analysis models become obsolete within months due to changing language. The tone is educational and conversational, blending historical examples with forward-looking predictions and practical case studies. The style is advisory, with Pentland offering clear critiques and alternative pathways based on his research and global advisory work. Listeners interested in pragmatic, human-centered applications of AI and data science beyond Silicon Valley hype will find this episode valuable. Those seeking highly technical deep dives into AI model architecture or debates about existential risk might find the discussion too focused on societal systems.

As heard by us

The piece argues for AI that helps people first, not platforms first.

This piece pushes back on the fantasy that the best AI is the one that hands over an answer and calls it progress. It makes a cleaner case for tools that help people, trade, companies, and governments do their work better, and it does so by setting frontier hype against the…

Read the full review in PlayNext →

Why you'd press play

You want a clearer case for AI that helps people do real work, not just systems that chase scale.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from The Action Catalyst