OpenAI’s newest, most performant model, announced in December, has passed the ARC-AGI test, purportedly outperforming most ...

Red teaming has become the go-to technique for iteratively testing AI models to simulate diverse, lethal, unpredictable attacks.

AlexNet, created by Alex Krizhevsky, Sutskever and Geoffrey Hinton, used a deep convolutional neural network (CNN)—a powerful ...

There's not enough human-generated data to keep AI models improving at the same rate. 2025 will put a new solution to the ...

OpenAI's o3 AI model recently achieved 85% on the ARC-AGI benchmark, similar to human-level performance. Though impressive, experts caution that it does not necessarily mean true human-level ...

Coming to the ARC-AGI (Abstract Reasoning Corpus - Artificial General Intelligence) benchmark, it features a series of grid-based pattern recognition questions that require reasoning and spatial ...

New Likert-scale-based AI jailbreak technique boosts attack success rates by 60%, highlighting urgent safety challenges.

There can be no such thing as (artificial) intelligence benefiting all of humanity without (artificial) integrity. This goes ...

Here are six themes we expect to make news in 2025, from professional services to crypto and artificial intelligence across ...

The cost of new 'reasoning models' may make companies reluctant to use them, even as their capabilities close in on ...

OpenAI has unveiled a new model for its products, arriving for users near the end of January, 2025: It's called o3 (we seem ...

Today OpenAI announced o3, its next-gen reasoning model. We've worked with OpenAI to test it on ARC-AGI, and we believe it ...