AI Companies Develop Potent Technology and Masterful Marketing
Artificial intelligence firms are creating powerful products, but they are also making extravagant claims that blur the line between innovation and hype. Last week, Anthropic unveiled Claude Mythos, an AI model focused on cybersecurity, which has generated both excitement and alarm over its purported capabilities. However, there is a significant caveat: the public cannot access it. Similarly, OpenAI recently announced it has developed a highly capable cybersecurity AI, adding to the industry's buzz.
Anthropic's Claims and Cybersecurity Implications
Anthropic described Mythos as a "reckoning" for the cybersecurity industry, asserting that the tool is exceptionally adept at uncovering software vulnerabilities. According to the company, Mythos has exposed thousands of weaknesses in widely used applications for which no patches exist. In response, Anthropic has formed an alliance with cybersecurity experts, dubbed Project Glasswing, to enhance defenses and restrict the model's distribution, likening it to an exclusive arthouse film shown only in Los Angeles and New York.
In a Guardian op-ed, Shakeel Hashim notes that Anthropic claims Mythos has identified vulnerabilities in every major browser and operating system. This suggests the AI could potentially aid hackers in disrupting critical global software. If such technology were widely available and as effective as advertised, the consequences could be severe. Cyber-attacks now impact physical infrastructure, with airports, hospitals, and transport networks recently suffering disruptions. While large-scale attacks previously required expert knowledge, Mythos might empower amateurs and amplify professionals' destructive potential.
Experts Question the Hype
Cybersecurity specialists are already challenging Anthropic's assertions. Aisha Down reports that while Mythos represents a real advancement, some of Anthropic's claims, such as discovering thousands of "zero-day vulnerabilities," may not be as impactful in practical cybersecurity terms. Jameison O'Reilly, an offensive cybersecurity expert, acknowledged the development's seriousness but cautioned against overstating its significance.
This situation highlights a broader trend: AI companies excel not only in technology but also in marketing. Drawing a parallel to Apple's ability to engineer desire, Anthropic appears to wield similar influence. Claude AI is indeed powerful, with businesses eagerly adopting its coding features. Major corporations like Apple, Nvidia, Google, JPMorganChase, Amazon Web Services, and Broadcom have joined Project Glasswing. Yet, the strategy of declaring a product too dangerous for release can fuel public fascination, as seen when Anthropic dominated discussions at the HumanX AI conference in San Francisco.
Historical Context and Public Perception
Hype has clouded public understanding of generative AI since its inception, with journalists and analysts striving to cut through the noise. In 2019, Slate reported that OpenAI deemed its GPT-2 text generator too risky to release, a story reminiscent of current events. Notably, Anthropic's CEO, Dario Amodei, previously served as OpenAI's vice-president of research before departing in 2020. OpenAI also delayed the wide release of its video generator Sora for months, though it did not revolutionize Hollywood as some feared; the company discontinued Sora late last month.
While early concerns about text generation seem minor compared to potential cybersecurity threats, the evolution past 2019's fears offers hope. Society may eventually navigate between today's alarm over digital security and exaggerated predictions, finding a balanced reality. This underscores the need for critical evaluation as AI continues to advance amidst marketing fervor.
