Alex Kontos, founder of Waterfox, a Firefox modification focused on privacy, productivity and eliminating frills, statedthat in the foreseeable future the Waterfox project will not integrate functionality that uses large language models, at least in the form in which it is developed in Firefox. We are talking about the use of large language models that influence the behavior of the browser and perform actions on behalf of the user. Local application of machine learning systems, for example, for translation from one language to another, is allowed.
Alex believes that Mozilla is making a fundamental mistake by intending to turn Firefox into an AI browser. He understands that Mozilla wants to strengthen its position in the market and maintain parity with competing browsers that are actively implementing AI. At the same time, by trying to target the average user and the mass market, which is already occupied by Chrome, Mozilla is making Firefox less attractive to the power users, developers and technicians who make up the core of the current audience.
According to Alex, one of the reasons for Firefox’s decline in the browser market share was the addition of features that the mainstream audience does not need. Mozilla promotes trust, transparency, and user freedom, but it introduces technologies that undermine all of these principles. Instead of giving the user the ability to fully control the browsing process, Mozilla intends to introduce an opaque AI layer that becomes an additional intermediary between the user and the browser.
The large language models underlying this layer are a “black box” that performs actions on behalf of the user, but does not allow analyzing the behavior and methods of working with data. A situation arises when the AI makes decisions about how and what to show the user, but the logic behind these decisions cannot be examined or understood.