OpenAI's latest AI model, GPT-5.3-Codex, has demonstrated significant advancements in AI-powered coding, but its release is being carefully managed due to cybersecurity concerns, according to Fortune. The model outperforms rival systems on coding benchmarks, suggesting a potential shift in software development.
The company is implementing unusually tight controls and delaying full developer access as it addresses the risks associated with the model's capabilities. Fortune reported that the same features that make GPT-5.3-Codex effective at writing, testing, and reasoning about code also create serious cybersecurity vulnerabilities.
Meanwhile, public perception of AI is becoming increasingly skeptical, prompting tech companies to invest in communications experts. Fortune noted that companies are paying up to $400,000 for AI "evangelists" to counter growing American skepticism. Pew research cited by Fortune indicates that 50% of Americans are more concerned than excited about AI, a rise from 37% in 2021. Only 10% reported being more excited than wary. Anthropic, for example, has tripled the size of its communications team in recent years, according to Fortune.
In other AI developments, METR, an AI research nonprofit, continues to track the progress of large language models. MIT Technology Review reported that METR updates a graph that has become a key part of the AI discourse, suggesting that certain AI capabilities are developing at an exponential rate. The graph has been closely watched since its release in March of last year. The latest version of Anthropic's most powerful model, Claude Opus 4.5, has outperformed the trend, according to MIT Technology Review.
The energy sector is also seeing advancements, with next-generation nuclear power remaining a prominent topic. MIT Technology Review addressed questions about advanced nuclear power, including the fuel needs for next-generation reactors. Many of these reactors do not use the low-enriched uranium used in conventional reactors.
In a separate area of technology, Hacker News highlighted the atomic capabilities of UNIX-like operating systems. A catalog from 2010 detailed operations that UNIX can perform atomically, making them useful for thread-safe and multi-process-safe programs without mutexes or readwrite locks. The catalog emphasized letting the kernel do as much work as possible.
Discussion
AI Experts & Community
Be the first to comment