Databricks and Perplexity co-founder Andy Konwinski has criticized the current discourse on "AI safety," arguing that large corporations may use it as a pretext to create regulatory barriers and restrict access to advanced technologies.
What Happened
In his new essay, Andy Konwinski stated that the concept of AI Safety could be used by companies like Anthropic as a tool to protect market positions. He proposed creating a "research commons"—an open-access system for large-scale computing resources that would allow scientists to conduct research at the level of frontier models without needing permission from private labs. Konwinski's position was supported by Yann LeCun, who compared attempts to impose strict control over AI to the historical banning of the printing press in the Ottoman Empire.
Context
Amid the rapid development of large language models, control over computing power and proprietary weights is becoming a critical factor of power. There are fears that concentrating these resources in the hands of a few giants turns AI into a closed infrastructure, where safety standards become a legal way to block the entry of new players and startups.
Why It Matters for the Industry
For the industry, this means increased political and economic pressure in discussions regarding safety standards. There is a risk of lobbying for restrictions on the use of open-source models and training methods, which could slow the pace of open research. In the long term, this could lead to the formation of new markets for decentralized computing as an attempt to bypass the Big Tech oligopoly.
Why It Matters for Users
For users and developers, this carries the risk of growing dependency on the APIs of a limited number of providers. Understanding that AI safety debates may have deep political and economic undertones is essential for evaluating the real reasons behind technology access restrictions and for forming strategies involving hybrid or multi-model approaches.
Sources
Author
Look at AI, Editorial Team