WSJ Tech News Briefing: AI Is No Substitute for the Human Brain

Release Date: May 13, 2025

In this insightful episode of the WSJ Tech News Briefing, the Wall Street Journal delves deep into the evolving landscape of artificial intelligence (AI), focusing on the capabilities and limitations of AI chatbots and the broader implications for data privacy and human intelligence.

AI Chatbots and Data Privacy

The episode opens with a discussion on the pervasive use of AI chatbots in everyday problem-solving. From selecting the right home gym equipment to crafting the perfect resume, millions rely on these tools daily. However, this reliance raises critical questions about data ownership and privacy.

Nicole Nguyen, WSJ Personal Tech Columnist, introduces the segment by referencing her series, Chatbot Confidential, where she explores these very concerns. She engages with listener queries, bringing in expert Christopher Mims to shed light on privacy regulations.

Key Discussion Points:

• FERPA and HIPAA in the Age of AI: Listening to a voicemail from Daniel Stewart, Christopher explains how AI tools intersect with federal privacy laws like FERPA (which protects educational records) and HIPAA (which safeguards medical records). He emphasizes the importance of using enterprise-grade AI versions that comply with regulations such as HIPAA, GDPR, and CCPA. For consumer-grade tools like ChatGPT or Anthropic's Claude, he advises scrubbing personally identifiable information to maintain privacy.

  "AI privacy laws extend to AI tools, particularly if you're using the public facing AI tools that are not enterprise versions... replacing student names, scrubbing as much personally identifiable information, sensitive information as possible is the right move." [02:28]

• Protection of Personal Media: Addressing Mitch's concern about family photos uploaded to AI chatbots, Christopher discusses options like opting out of AI training and using features such as temporary chat modes. However, he cautions that in the nascent stages of generative AI, some data might still be subject to human review or longer storage periods, drawing parallels to established platforms like Google Photos.

  "You can opt out of AI training with the caveat that in some instances it could be reviewed." [03:53]

This segment underscores the delicate balance between leveraging AI's convenience and safeguarding personal data, urging users to be proactive in understanding and utilizing privacy settings.

Skepticism on AI Achieving Human-like Intelligence

Transitioning from data privacy, the episode tackles the optimistic claims surrounding AI's potential to reach Artificial General Intelligence (AGI)—the elusive milestone where machines exhibit human-like understanding and reasoning.

Christopher Mims and Julie Chang engage in a robust discussion, questioning the current trajectory of AI development.

Key Insights:

• Current State of AI vs. AGI: Julie asserts that today's transformer-based AIs, such as ChatGPT, operate on extensive lists of heuristics rather than genuine understanding. She argues that these models, despite their sophistication, fall short of replicating the nuanced and adaptive nature of human intelligence.

  "We are definitely nowhere near AGI... Today's transformer based AIs... the way that they work is just they have this kind of almost infinitely long list of little rules of thumb that they apply." [06:37]

• The Manhattan Map Experiment: To illustrate AI's limitations, Julie references a study where an AI model was trained with turn-by-turn directions within Manhattan. While the AI could accurately provide directions, its internal "map" was a convoluted and inaccurate representation of the actual city layout. This discrepancy highlights the AI's inability to form coherent spatial and causal understanding akin to humans.

  "When you ask it for directions on the island of Manhattan... it looked totally crazy. Streets were connected that are very far distant and diagonal to one another." [08:21]

• Plateauing of AI Intelligence: The conversation delves into the stagnation in AI's general abilities. Julie points out that recent models, despite increased data and training, often perform worse on certain tasks, such as mathematics, while becoming more prone to hallucinations. This suggests that merely scaling up AI models without fundamental advancements in their architecture may not yield the desired improvements.

  "The general abilities of these AIs have definitely hit a ceiling... reinforcement learning... makes them more likely to hallucinate, makes them worse at other things." [11:20]

This segment critically assesses the hype around AGI, emphasizing that while AI continues to advance in specific domains, replicating the depth and adaptability of the human brain remains a distant goal.

Conclusion

The episode wraps up by highlighting the collaborative efforts behind the scenes, acknowledging producers Julie Chang and Emily Martosi, along with Melanie Roy's support. Listeners are encouraged to stay tuned for upcoming segments, including the TNB Tech Minute.

Takeaways:

• Data Privacy is Paramount: As AI chatbots become integral to daily tasks, understanding and managing data privacy settings is crucial to protect personal information.

• AI's Current Limitations: Despite impressive advancements, AI chatbots do not possess human-like understanding or reasoning. Their operations are based on complex rule sets rather than genuine intelligence, making them susceptible to errors that humans would effortlessly navigate.

• Future of AI: While AI continues to evolve, achieving AGI requires more than just scaling up existing models. It necessitates fundamental breakthroughs in how machines process and understand information.

For those keen on staying abreast of the latest in technology and AI, this episode offers a comprehensive exploration of both the potentials and pitfalls of current AI advancements.