Tag: ai

I began studying AI in the mid-1980s. Unusually for a computer scientist of that era, my interest was entirely in information, not in machines. I became obsessed with understanding what it meant to live during the transition from the late Industrial Age to the early Information Age.  

What I learned is that computers fundamentally alter the economics of information. We now have inexpensive access to more information, and to higher quality information, than ever before. In theory, that should help individuals reach better decisions, organizations devise improved strategies, and governments craft superior policies. But that’s just a theory. Does it? 

The answer is “sometimes.” Unfortunately, the “sometimes not” part of the equation is now poised to unleash devastating consequences. 

Consider the altered economics of information: Scarcity creates value. That’s been true in all times, in all cultures, and for all resources. If there’s not enough of a resource to meet demand, its value increases. If demand is met and a surplus remains, value plummets.  

Historically, information was scarce. Spies, lawyers, doctors, priests, scientists, scholars, accountants, teachers, and others spent years acquiring knowledge, then commanded a premium for their services.  

Today, information is overabundant. No one need know anything because the trusty phones that never leave our sides can answer any question that might come our way. Why waste your time learning, studying, or internalizing information when you can just look it up on demand? 

Keep reading

When it comes to concrete use cases for large language models, AI companies love to point out the ways coders and software developers can use these models to increase their productivity and overall efficiency in creating computer code. However, a new randomized controlled trial has found that experienced open source coders became less efficient at coding-related tasks when they used current AI tools.

For their study, researchers at METR (Model Evaluation and Threat Research) recruited 16 software developers, each with multiple years of experience working on specific open source repositories. The study followed these developers across 246 individual “tasks” involved with maintaining those repos, such as “bug fixes, features, and refactors that would normally be part of their regular work.” For half of those tasks, the developers used AI tools like Cursor Pro or Anthropic’s Claude; for the others, the programmers were instructed not to use AI assistance. Expected time forecasts for each task (made before the groupings were assigned) were used as a proxy to balance out the overall difficulty of the tasks in each experimental group, and the time needed to fix pull requests based on reviewer feedback was included in the overall assessment.

Keep reading

OpenAI is preparing to launch a new feature that may mark the latest leap forward in the artificial intelligence revolution.

On Thursday, the company unveiled ChatGPT Agent, a tool designed to carry out tasks independently using its own built-in “virtual computer.”

The Verge has more details:

In a briefing and demo with The Verge, Yash Kumar and Isa Fulford — product lead and research lead on ChatGPT Agent, respectively — said it’s powered by a new model that OpenAI developed specifically for the product.

The company said the new tool can perform tasks like looking at a user’s calendar to brief them on upcoming client meetings, planning and purchasing ingredients to make a family breakfast, and creating a slide deck based on its analysis of competing companies.

The model behind ChatGPT Agent, which has no specific name, was trained on complex tasks that require multiple tools — like a text browser, visual browser, and terminal where users can import their own data — via reinforcement learning, the same technique used for all of OpenAI’s reasoning models. OpenAI said that ChatGPT Agent combines the capabilities of both Operator and Deep Research, two of its existing AI tools.

To develop the new tool, the company combined the teams behind both Operator and Deep Research into one unified team. Kumar and Fulford told The Verge that the new team is made up of between 20 and 35 people across product and research.

Despite stiff competition from the likes of Google’s Gemini, Elon Musk’s Grok and China’s DeepSeek, OpenAI remains the most popular and advanced large language model currently on the market.

Keep reading

The world’s most advanced AI models are exhibiting troubling new behaviors – lying, scheming, and even threatening their creators to achieve their goals.

In one particularly jarring example, under threat of being unplugged, Anthropic’s latest creation Claude 4 lashed back by blackmailing an engineer and threatened to reveal an extramarital affair.

Meanwhile, ChatGPT-creator OpenAI’s o1 tried to download itself onto external servers and denied it when caught red-handed.

These episodes highlight a sobering reality: more than two years after ChatGPT shook the world, AI researchers still don’t fully understand how their own creations work.

Yet the race to deploy increasingly powerful models continues at breakneck speed.

This deceptive behavior appears linked to the emergence of “reasoning” models – AI systems that work through problems step-by-step rather than generating instant responses.

According to Simon Goldstein, a professor at the University of Hong Kong, these newer models are particularly prone to such troubling outbursts.

“O1 was the first large model where we saw this kind of behavior,” explained Marius Hobbhahn, head of Apollo Research, which specializes in testing major AI systems.

These models sometimes simulate “alignment” – appearing to follow instructions while secretly pursuing different objectives.

Keep reading