Vivaldi Browser Says No to Language Models | Best llm Training Dataset Github | Neuralnetfastai | Large Language Models Tutorial Python Github | Turtles AI

VIVALDI CHOOSES NOT TO INTEGRATE LANGUAGE MODELS INTO ITS BROWSERS 

Vivaldi takes a stand against language models in browsers

In recent times, many browsers have embraced the use of technologies based on advanced language models, but Vivaldi has decided to take a different path. Julien Picalausa, Vivaldi’s developer since 2017, explained the reasons for this choice, highlighting the risks associated with the use of these models.

How language models work

Picalausa explained that language models, or LLMs (Large Language Models), are trained on large amounts of text to predict which word comes after a given sequence of words, using probabilistic language modeling. However, according to Vivaldi, this technology has two main problems: the risk of copyright and privacy violations, and the creation of "plausible lies."

Vivaldi’s concerns: copyright and privacy

To train an effective language model, it is necessary to use texts from many different, often recent, sources. This leads to a significant risk of copyright infringement, since the models may end up reproducing existing texts almost identically. In addition, the texts used for training may contain personal data that the models may rework inaccurately, attributing incorrect information to real people.

Plausible lies: a hidden danger

Modern LLMs use filters to improve the quality of the output, but this is not always sufficient to ensure the accuracy of the information. Models can be influenced by low-quality or erroneous content on the Web, leading to the generation of text that seems plausible but may contain errors or misinformation.

Vivaldi’s ethical position

For these reasons, Vivaldi has chosen not to integrate LLMs into its browser. The company is also concerned about the significant energy consumption required to run these models. However, Vivaldi will still allow users to access and use LLMs online while exploring privacy-friendly and useful machine learning solutions.

An evolving landscape

Despite Vivaldi’s decision, the benefits of modern artificial intelligence are undeniable. However, it is crucial to always verify the information provided by LLMs, as they can make mistakes. The future of LLMs could see significant improvements through new techniques and approaches, such as multi-token prediction and knowledge graphs.

In summary, Vivaldi’s choice represents an ethical and prudent stance in the changing landscape of AI-based technologies.