When AI Gets Smarter, Do We Get Wiser?

This posting is inspired from my recent viewing of Fareed Zakaria GPS Broadcast on CNN.  

Anthropic’s AI chatbot Claude and Open AI’s ChatGPT both got major updates last week, and their improved abilities are quite stunning. But how much do we know about why these models think and behave in the way they do? Fareed sits down with New Yorker staff writer Gideon Lewis-Kraus, who spent months visiting Anthropic’s headquarters seeking an answer.  

When AI Gets Smarter, Do We Get Wiser?

Last week, two of the most influential AI systems on the planet-Anthropic’s Claude and OpenAI’s ChatGPT, received major upgrades. Users immediately noticed the difference. These models reason more fluidly, write more persuasively, and seem uncannily better at understanding context, nuance, and even human hesitation.

The improvements are, as one might say, stunning.

But as Fareed Zakaria explored on Fareed Zakaria GPS today, the real question is not what these systems can now do, but why they do it the way they do.

To get closer to that answer, Fareed sat down with Gideon Lewis-Kraus, a staff writer for The New Yorker, who spent months embedded with researchers at Anthropic. His reporting lifts the curtain on one of the most unsettling truths of modern AI: even the people building these systems don’t fully understand their inner workings.

The Black Box Problem

Large language models don’t “think” in any human sense. They are trained on staggering amounts of data and learn to predict what comes next word by word, token by token. Yet out of that statistical machinery emerges something that feels eerily like judgment, creativity, and intent.

Anthropic, more than most AI labs, is obsessed with interpretability, trying to peer inside the black box and understand why a model gives one answer instead of another. Lewis-Kraus describes researchers tracing patterns through neural networks, identifying circuits that seem to correspond to concepts like honesty, deception, or uncertainty.

Still, these insights are partial. We can observe behaviors. We can influence outcomes. But full comprehension remains elusive.

Power Without Transparency

That gap between capability and understanding is what makes this moment so consequential. As Claude and ChatGPT grow more capable, they are being trusted with decisions that affect education, medicine, law, and public opinion. Yet unlike traditional software, their reasoning cannot be neatly audited line by line.

Fareed’s conversation underscores a quiet unease shared even by AI insiders: we are deploying systems whose intelligence is advancing faster than our ability to explain it.

That doesn’t make these tools dangerous by default, but it does demand humility.

A Personal Reflection

Having lived through eras of technological upheaval, from the dawn of television to the rise of the internet, I’ve learned this much: every generation believes its tools are under control right up until they aren’t. What feels magical at first eventually becomes ordinary, then indispensable, and finally invisible. AI may follow that same arc. The difference this time is speed.

So perhaps the most important takeaway from Fareed’s interview is not fear, but responsibility. If we are going to build machines that sound like they understand us, we owe it to ourselves to keep asking how and to admit, honestly, when we don’t yet know the answer.

Because in the age of intelligent machines, wisdom still has to be human. There is no substitute for human wisdom. 

AI Overview:

Smart AI does not inherently make us wiser. In fact, research suggests a "performance-wisdom gap" where users become more effective but less capable of accurately judging their own abilities. 

The Wisdom vs. Intelligence Gap

• Performance Without Insight: Studies show that while AI tools can boost performance on complex tasks like the LSAT, users often overestimate their results and experience a decrease in metacognitive accuracy—the ability to think about their own thinking.
• Cognitive Offloading: Relying on AI for problem-solving can lead to "metacognitive laziness," where the brain skips the "productive struggle" required to build deep understanding and resilience.
• The "GPS Effect": Just as GPS usage can weaken spatial memory, over-reliance on AI for reasoning can dull critical thinking and analytical skills. 

Why AI Lacks Wisdom

Current AI systems are seen as "smarter" because they can process vast datasets and identify patterns faster than humans. However, they lack the foundational elements of wisdom: 

• Lived Experience: Wisdom is often defined as "metabolized experience"—learning from one's own suffering, failures, and physical navigation of the world, which AI cannot do.
• Ethical Judgment: AI lacks intrinsic values, moral intuition, and the ability to understand the human consequences of its recommendations.
• Contextual Awareness: While AI excels at narrow tasks, it often struggles to understand the broader social, cultural, or emotional context that informs human judgment. 

The Path to Becoming Wiser

Becoming wiser in the age of AI depends on how we choose to interact with the technology:

• Co-pilot vs. Autopilot: The best outcomes occur when humans use AI as a "thought partner" to handle heavy analysis while retaining final judgment and accountability.
• Maintaining "Inner Resonance": Wisdom requires developing an awareness of the boundary between borrowed machine cognition and embodied human insight. 
• Finally, my Photos of the Day: 
•  
  
  Valentine Dinner at Newtons, the Heritage Downtown, Walnut Creek, CA, 
• AI copies of the Photo in Water Color and Oil style Portrait removing the Background
•