The escalating costs associated with artificial intelligence are prompting companies to seek novel solutions, with some exploring a more rudimentary form of communication: instructing AI models to speak like cavemen. This approach aims to reduce the significant expense of running sophisticated AI services by minimizing their verbose outputs.
According to a report by 404 Media, a plugin named 'Caveman' is gaining traction among businesses, including at OpenAI itself. The plugin modifies the behavior of AI models like Claude, steering them away from conversational and human-like prose towards more direct and curt responses. The creator, Julius Brussee, developed the plugin to address the "unnecessary prose" such as pleasantries, hedging, and transitional language that he found consumed excessive computational resources.
Brussee claims that implementing the 'Caveman' plugin has led to a reduction in token usage by approximately 65 percent. This reduction in verbosity effectively 'dumbs down' the AI, making its operation more cost-effective. This shift represents a stark contrast to the years of marketing AI's human-like intelligence, now seemingly superseded by the need for financial viability.
The expense of maintaining advanced AI services has become a significant hurdle for many major firms, with few achieving profitability. The trend toward per-use subscription models, rather than flat rates, reflects these high operational costs. OpenAI's recent closure of its video generator, Sora, which reportedly incurred a million-dollar daily cost, further illustrates the financial challenges in the AI sector, even amidst lucrative partnerships like the one with Disney.
The 'Caveman' plugin, by stripping away non-essential linguistic elements, offers a pragmatic if somewhat ironic, solution to the economic pressures facing AI development and deployment. It highlights a growing tension between the pursuit of advanced AI capabilities and the fundamental business imperative of cost management.