Ep 386: Was 2025 a Great or Terrible Year for AI? (w/ Ed Zitron)

The podcast analyzes the AI landscape of 2025, questioning whether it was a successful or disastrous year for the technology. Cal Newport interviews Ed Zitron, an AI commentator, to dissect major AI stories month by month, examining the realities behind the headlines. They discuss DeepSeek, a Chinese AI model trained for significantly less money, and debate the economic viability and hype surrounding AI agents. The conversation covers the shift from scaling to tuning AI models, NVIDIA's role, and the AI 2027 fan fiction. Zitron expresses skepticism towards AI safety advocates and their focus on distant threats rather than current harms, while both dissect the economic realities and marketing ploys within the AI industry.

Outlines

Part 1: DeepSeek and the Global AI Shift

Part 2: The Agent Hype and Coding Reality

Part 3: Scaling Struggles and Technical Inefficiencies

Part 4: Existential Risks and Industry "Grifts"

Part 5: GPT-5, Sora, and Financial Realities

Part 6: The 2025 Verdict and AI Culture

Part 7: Algorithmic Curation and Social Media

Part 8: Listener Feedback and Closing

Sign in to continue reading, translating and more.

Open full episode in Podwise

Deep Questions with Cal Newport

Part 1: DeepSeek and the Global AI Shift

00:02AI in 2025: A Year of Great Advancements or Terrible Setbacks?

AI in 2025: A Year of Great Advancements or Terrible Setbacks?

02:43DeepSeek's Rise: A Chinese AI Model Challenging US Tech Dominance

DeepSeek's Rise: A Chinese AI Model Challenging US Tech Dominance

07:19The DeepSeek Aftermath: Memory Holes, Small Models, and Frontier Model Dependence

The DeepSeek Aftermath: Memory Holes, Small Models, and Frontier Model Dependence

Part 2: The Agent Hype and Coding Reality

12:27The Agent Hype: OpenAI's CPO and Benioff's Trillion-Dollar Promises

The Agent Hype: OpenAI's CPO and Benioff's Trillion-Dollar Promises

17:35Coding Agents: Vibe Coding Frauds and the Tab-Complete Reality

Coding Agents: Vibe Coding Frauds and the Tab-Complete Reality

Part 3: Scaling Struggles and Technical Inefficiencies

28:16GPT-4.5 and Project Orion: OpenAI's Scaling Struggles and the Shift to Tuning

GPT-4.5 and Project Orion: OpenAI's Scaling Struggles and the Shift to Tuning

32:33Test Time Compute: NVIDIA's GPU Dominance and the AI Industry's Inefficiency

Test Time Compute: NVIDIA's GPU Dominance and the AI Industry's Inefficiency

Part 4: Existential Risks and Industry "Grifts"

39:04AI 2027: Fan Fiction, Extinction Scares, and the Effective Altruism Grift

AI 2027: Fan Fiction, Extinction Scares, and the Effective Altruism Grift

45:55Hinton's Grift: Attention, Glory, and the Neglect of Real-World AI Harms

Hinton's Grift: Attention, Glory, and the Neglect of Real-World AI Harms

55:03Amadei's Doomsday: PhD-Level AI and the Myth of Efficient Anthropic

Amadei's Doomsday: PhD-Level AI and the Myth of Efficient Anthropic

Part 5: GPT-5, Sora, and Financial Realities

1:01:11GPT-5's Underwhelming Release: Altman's Hype, Zitron's Reporting, and the Router Model

GPT-5's Underwhelming Release: Altman's Hype, Zitron's Reporting, and the Router Model

1:07:12The GPT-5 Aftermath: Financial Realities, Bubble Talk, and the F-150 Strategy

The GPT-5 Aftermath: Financial Realities, Bubble Talk, and the F-150 Strategy

1:15:10OpenAI's Data Center Deals: $300 Billion to Oracle and the Market's Toddler Mentality

OpenAI's Data Center Deals: $300 Billion to Oracle and the Market's Toddler Mentality

1:20:21Anthropic's AWS Bills: $2.66 Billion Spent and the Myth of Efficiency

Anthropic's AWS Bills: $2.66 Billion Spent and the Myth of Efficiency

1:27:00Sora's High Costs and the Inefficiency of Large Language Models

Sora's High Costs and the Inefficiency of Large Language Models

1:34:40The Impossibility of Cost Control: Google Search vs. Large Language Models

The Impossibility of Cost Control: Google Search vs. Large Language Models

Part 6: The 2025 Verdict and AI Culture

1:40:02OpenAI's Code Red: De-Emphasizing Agents and the Lack of a Business Model

OpenAI's Code Red: De-Emphasizing Agents and the Lack of a Business Model

1:47:17The Cult of AI: Loneliness, Grifters, and the Ultra-Grift Economy

The Cult of AI: Loneliness, Grifters, and the Ultra-Grift Economy

1:57:29The Verdict: 2025, a Terrible Year for AI

The Verdict: 2025, a Terrible Year for AI

Part 7: Algorithmic Curation and Social Media

1:59:40Social Networks Won't Show You Content From Friends and Family

Social Networks Won't Show You Content From Friends and Family

2:02:20Algorithmic Curation: Bad News for Social Media Companies

Algorithmic Curation: Bad News for Social Media Companies

2:03:57Television is a Good Metaphor for How Media Will Work

Television is a Good Metaphor for How Media Will Work

2:04:24People Like Diversion: A History of Media and Distraction

People Like Diversion: A History of Media and Distraction

2:08:30The Battle Against Being Lost in Distraction

The Battle Against Being Lost in Distraction

2:09:53Social Media is More Dangerous Than TV

Social Media is More Dangerous Than TV

2:11:20AI Defends AI: A More Refined Perspective on the Shift

AI Defends AI: A More Refined Perspective on the Shift

2:13:02It's Mainly Slop Now: The Goal of Social Media Algorithms

It's Mainly Slop Now: The Goal of Social Media Algorithms

Part 8: Listener Feedback and Closing

2:14:17Zuckerberg Was Never the Brightest Bulb in the Pack

Zuckerberg Was Never the Brightest Bulb in the Pack

2:17:12Zuckerberg's Success: A Savvy and Savage Corporate Infighter

Zuckerberg's Success: A Savvy and Savage Corporate Infighter

2:19:21More Nerd Shit, Please

More Nerd Shit, Please

2:20:05Is Newport Being Paid to Read These Adverts?

Is Newport Being Paid to Read These Adverts?

2:20:21On the Internet, the People Who Yelled at the Television Now Yell at Each Other

On the Internet, the People Who Yelled at the Television Now Yell at Each Other

2:21:21Sign Up For My Newsletter

Sign Up For My Newsletter