#216 - Grok 4, Project Rainier, Kimi K2 | Last Week in AI

In the "Last Week in AI" podcast, hosts Andrey Kurenkov and Jeremie Harris dissect the latest AI news, starting with Grok 4's impressive launch and its implications for AI safety and alignment, including a discussion of its recent antisemitic outputs. They explore Perplexity and OpenAI's ventures into AI-powered web browsers, Repl.it and Cursor's new AI coding agent features, and the financial landscape with Lovable's fundraising and Amazon's massive AI supercluster for Anthropic. The hosts also cover Elon Musk's plans for xAI's power plant, Microsoft's AI chip delays, and personnel shifts at Safe Superintelligence. They further analyze open-source AI model releases, research on reasoning and transferability, and Meta's study on AI's impact on developer productivity, concluding with discussions on AI safety evaluations, regulatory concerns, and bioweapon risks.

Outlines

Sign in to continue reading, translating and more.

Continue

#216 - Grok 4, Project Rainier, Kimi K2

Last Week in AI

Introduction to AI News and Grok 4 Launch

Grok 4 Training Details and Antisemitic Trope Incident

Alignment Challenges and Truth-Seeking Priorities

Perplexity's AI Browser, OpenAI's Browser Plans, and Repl.it's New Features

Cursor's Pricing Changes, Lovable's Fundraise, and Amazon's AI Supercluster

xAI's Power Plant Purchase, Microsoft's AI Chip Delay, and Safe Superintelligence's Leadership Change

OpenAI's Stock Compensation and Open Source Model Releases

Math Reasoning and General LLM Capabilities, Meta's AI Study

Goal Misgeneralization, Correlated Errors in LLMs, and SVE Bench Evaluation

Stealth and Situational Awareness Evaluations, Chain of Thought and Model Honesty

Alignment Faking, Positive Review Prompts, and Data Transfer Concerns

#216 - Grok 4, Project Rainier, Kimi K2

Last Week in AI

00:00Introduction to AI News and Grok 4 Launch

Introduction to AI News and Grok 4 Launch

09:21Grok 4 Training Details and Antisemitic Trope Incident

Grok 4 Training Details and Antisemitic Trope Incident

17:17Alignment Challenges and Truth-Seeking Priorities

Alignment Challenges and Truth-Seeking Priorities

29:07Perplexity's AI Browser, OpenAI's Browser Plans, and Repl.it's New Features

Perplexity's AI Browser, OpenAI's Browser Plans, and Repl.it's New Features

37:11Cursor's Pricing Changes, Lovable's Fundraise, and Amazon's AI Supercluster

Cursor's Pricing Changes, Lovable's Fundraise, and Amazon's AI Supercluster

47:20xAI's Power Plant Purchase, Microsoft's AI Chip Delay, and Safe Superintelligence's Leadership Change

xAI's Power Plant Purchase, Microsoft's AI Chip Delay, and Safe Superintelligence's Leadership Change

57:05OpenAI's Stock Compensation and Open Source Model Releases

OpenAI's Stock Compensation and Open Source Model Releases

1:02:11Math Reasoning and General LLM Capabilities, Meta's AI Study

Math Reasoning and General LLM Capabilities, Meta's AI Study

1:11:19Goal Misgeneralization, Correlated Errors in LLMs, and SVE Bench Evaluation

Goal Misgeneralization, Correlated Errors in LLMs, and SVE Bench Evaluation

1:22:53Stealth and Situational Awareness Evaluations, Chain of Thought and Model Honesty

Stealth and Situational Awareness Evaluations, Chain of Thought and Model Honesty

1:35:40Alignment Faking, Positive Review Prompts, and Data Transfer Concerns

Alignment Faking, Positive Review Prompts, and Data Transfer Concerns