Google has introduced its next-generation artificial intelligence (AI) model, Gemini 4 Argon, but it has received criticism for its relatively poor coding capabilities.
According to industry sources, Google recently showcased the new AI model, Argon.
Google stated that Argon excels particularly in the field of cybersecurity. It has also demonstrated strong capabilities in organizing complex and lengthy tasks in software development, finance, and law. In a blog post, Google noted, "Argon is designed to maintain deep reasoning across complex, long-term workflows, fundamentally changing the way we work and develop."
However, Google has decided not to make Argon available to the general public immediately, citing concerns that high-performance AI could be misused for hacking and other malicious activities. Currently, Argon is only accessible to partners in Google's security program, the Fairwind Program.
Alongside the launch of Argon, performance controversies have emerged. According to Bloomberg, internal sources at Google indicated that while Argon may have scored well in performance evaluations, it has not met expectations in real-world applications. Reports suggest that its performance in certain coding tasks, such as website and application development, has been subpar.
In response, Google refuted claims that Argon is underperforming in coding, stating, "It is inaccurate to say that Argon is lacking in coding capabilities."
This model has garnered attention amid a growing call within the global AI community to slow down the pace of AI development, often referred to as the 'AI development speed control theory.' Dario Amodei, CEO of Anthropic, argued last month that development should be slowed if AI capabilities reach dangerous levels. Sam Altman, CEO of OpenAI, has also expressed the need for speed regulation in the frontier AI development process.
* This article has been translated by AI.
Copyright ⓒ Aju Press All rights reserved.