Anthropic's New AI Model Sparks Global Oversight Calls from Industry Leaders

Artificial intelligence firm Anthropic has unveiled its latest model, Claude 3.5, a powerful system described as belonging to the 'Mythos' class. This release has generated significant excitement within the AI community, but also deep concern among industry leaders about the potential risks associated with advanced AI capabilities.

“

The platform Stripe used it to migrate 50 million lines of code – a task that would have taken a team of engineers over two months, the algorithm completed in one day.

The new model demonstrates remarkable abilities, including the migration of 50 million lines of code in just one day, a task that would typically take a team of engineers over two months. Claude 3.5 has also set new benchmarks in financial assessments and can independently play the game Pokémon FireRed using only a visual interface. Experts warn that 'Mythos' class systems possess advanced capacities in areas like biochemistry, hazardous material synthesis, and software reverse engineering, raising fears of misuse by malicious actors.

A primary concern is the potential for "recursive self-improvement," a scenario where an AI model becomes autonomous enough to write its own code and train its successor. This could lead to an exponential acceleration in AI development, potentially beyond human control. Leaders from OpenAI, including CEO Sam Altman and chief scientist Jakub Pachocki, have published a manifesto calling for a "third phase" of AI development focused on safety and oversight.

“

The biggest specter remains so-called recursive self-improvement, that is, the moment when a Mythos-class model becomes autonomous enough to write code and train its successor.

In response to these escalating concerns, prominent figures in the AI field, including those from OpenAI and Anthropic, are advocating for the establishment of an international governing body. They propose an organization modeled after the International Atomic Energy Agency (IAEA) that would have the authority to slow down or halt research on 'frontier' AI technologies globally. This call for regulation comes just as companies like OpenAI are preparing for potential stock market debuts, signaling a shift towards prioritizing safety alongside commercialization.

“

OpenAI leaders, along with Anthropic, which competes with them, are directly appealing for the establishment of an international governing body modeled after the Atomic Energy Agency.