AI

As AI appears to do much of the work of entry-level job seekers, the entire premise of a college education seems to be collapsing.

Researchers have uncovered 175,000 exposed Ollama AI systems worldwide, revealing a growing shadow AI infrastructure running without monitoring or security controls.

Google AI Ultra subscribers in the U.S. can now try out Project Genie.

An anonymous reader quotes a report from Wired: Earlier this month, Joseph Thacker's neighbor mentioned to him that she'd preordered a couple of stuffed dinosaur toys for her children. She'd chosen the toys, called Bondus, because they offered an AI chat feature that lets children talk to the toy li...

AI is reshaping how consumers shop by streamlining choices, boosting confidence, and more. Discover insights from IAB’s latest study.

A response to Education Week’s warning about regulation gaps, student risk, and what policy action should look like

Citigroup has rolled out mandatory AI training for all 175,000 of its employees across 80 locations worldwide, a sweeping initiative that CEO Jane Fraser describes as helping workers "reinvent themselves" before the technology permanently alters what they do for a living. The $205 billion bank sen...

OpenAI is adding ads to ChatGPT. And unlike Facebook, ChatGPT doesn't need to infer your data: you just tell it #ArtificialIntelligence #AIEducation #AIEdu

: Points to a use-case problem

The generative AI landscape, once dominated almost entirely by OpenAI’s ChatGPT, is showing clear signs of fragmentation as Google’s Gemini gains steam, new data suggests.According to industry …

What if the secret to mastering AI isn’t about crafting perfectly clear instructions, but about how you respond when things don’t go as planned? Nate B Jones walks through how even the most detailed …

New data from the Marfeel Discover Monitoring shows Google Discover is no longer a publisher-first surface. It’s becoming an AI platform with YouTube and X absorbing real estate that once went to newsrooms. Discover is no longer just ranking content. Generated AI Summaries are no longer edge cases. In the US, Brazil, and Mexico, 51% of the feed is now AI Summaries, structurally reducing opportunities for publisher clicks. Google signals multiple publisher sources in AI Summaries, but off...

When AI Can Fake Majorities, Democracy Slips Away

TEACH-AI, a new effort on AI in education, aims to help future educators use AI tools in an environmentally conscious way.

"Institutional incentives, not abstract ethical principles, are the primary force shaping AI's censorship and guardrail behavior." There...

I recently encountered a term that stopped me in my tracks – cognitive outsourcing (thanks to Professor Miles Berry for that one!). We’re all well-acquainted with cognitive offloading. …

Powered by AI and built by Khan Academy, our tool gives students more writing practice while helping teachers tackle cheating.

Gemini now offers full-length, no-cost practice SATs, featuring content from The Princeton Review.

We’ve observed agents discovering progressively more complex tool use while playing a simple game of hide-and-seek. Through training in our new simulated hide-and-seek environment, agents build a series of six distinct strategies and counterstrategies, some of which we did not know our environment supported. The self-supervised emergent complexity in this simple environment further suggests that multi-agent co-adaptation may one day produce extremely complex and intelligent behavior.

We trained models to think in simpler, more traceable steps—so we can better understand how they work.

Frontier reasoning models exploit loopholes when given the chance. We show we can detect exploits using an LLM to monitor their chains-of-thought. Penalizing their “bad thoughts” doesn’t stop the majority of misbehavior—it makes them hide their intent.

We introduce evaluations for chain-of-thought monitorability and study how it scales with test-time compute, reinforcement learning, and pretraining.