In 2023, we inaugurated the era of Language Model Language (LLM), driven by pioneers such as ChatGPT, its partners at Microsoft with Copilot, and the emergence of a promising challenger, the French startup Mistral AI.
This year marked a turning point in the world of open source, with an exceptional novelty: the advent of an unexpected player, the young company Mistral AI. However, the real revelation was the meteoric rise of OpenAI, which not only celebrated its fifth anniversary, but also overcame a “Steve Jobs” crisis. In one week, OpenAI could have given birth to both an “Apple” and a “NeXT” of artificial intelligence, but thanks to informed decisions, the offending board was replaced.
Throughout 2023, Apple stayed away from discussions on AI, while Google launched Bard a few months after the arrival of ChatGPT. However, Apple, now focused on selling mobile devices rather than desktop computers, took the initiative by reinventing the wheel with a fork of PyTorch, thus highlighting the LLM on its favorite terrain: the smartphone .
LLM, despite its high consumption of GPUs for servers, is on the way to becoming the equivalent of Linux in the field of free computing. Even if traditional machine learning has not said its last word, to be a truly universal technology, LLM must be viable not only on powerful servers, but also on more modest devices such as a Raspberry Pi 3, a high-end smartphone or a gaming PC equipped with an Nvidia RTX 3050.
While GNU/Linux and BSD UNIXes have successfully made this transition, LLMs must follow the same path for the benefit of everyone, from data centers to the average person, even if he or she does not master the intricacies of TensorFlow. Apple, Google and the open source community now have the same strengths, and the future of the LLM lies in its adoption at all levels of society.
Top comments (0)