In a bold and unconventional move, Mistral AI, a Paris-based open-source model startup, recently released their latest language AI model called MoE 8x7B. Rather than following the traditional route of papers, blogs, code, or press releases, Mistral took a completely different approach. They simply shared a link to a large torrent file for users to download their new model. This unexpected and distinctive release strategy immediately caught the attention of the AI community due to its stark contrast with the highly polished and rehearsed video releases commonly seen from tech giants like Google.
Genuine Comparison and Criticism
Mistral’s release of MoE 8x7B sparked discussions and comparisons within the AI community. Unlike Google’s Gemini release, which faced heavy criticism for its edited and staged demo videos, Mistral opted for transparency and authenticity. Reddit users described MoE 8x7B as a “scaled-down GPT-4” and noted that it consists of 8 experts with 7B parameters each. In contrast, GPT-4 is speculated to have 8 experts with 111B parameters individually and 55B shared attention parameters (166B parameters per model). Mistral’s straightforward approach of sharing the torrent file, without any bells and whistles, allowed the community to focus solely on the model’s capabilities and potential.
Mistral’s guerrilla-style release strategy received praise and support from various AI enthusiasts. Notably, entrepreneur George Hotz expressed his admiration for the move, acknowledging its significance. Similarly, Eric Jang, the Vice President of AI at 1X Technologies and a former research scientist at Google, stated that Mistral’s brand was quickly becoming one of his favorites in the AI space. This positive reception highlights the impact of Mistral’s approach in generating buzz and capturing the attention of industry professionals.
Mistral AI has been paving its own path in the AI landscape, showcasing a consistent pattern of innovation and remarkable achievements. The company recently secured a staggering $2 billion valuation in a funding round led by Andreessen Horowitz, firmly establishing its position as a major player in the industry. The success of Mistral’s record-setting $118 million seed round, reportedly the largest in European history, further solidifies their position as an influential force in the AI startup ecosystem. Moreover, their first large language AI model, Mistral 7B, was launched in September, further demonstrating their commitment to pushing boundaries and building cutting-edge AI solutions.
Challenging the Regulatory Landscape
In addition to their groundbreaking technological advancements, Mistral AI has actively participated in the ongoing debate surrounding the EU AI Act. As reported, the company has been lobbying the European Parliament to advocate for less regulation on open-source AI. This engagement showcases Mistral’s dedication to shaping the regulatory landscape and promoting the continued growth and development of open-source AI technologies.
Mistral AI’s recent release of their MoE 8x7B model through a torrent link has been met with intrigue and admiration within the AI community. By taking a bold and unconventional approach, Mistral has differentiated itself from its counterparts, generating substantial interest and buzz. This move exemplifies their commitment to transparency, authenticity, and innovation, solidifying their position as a leading open-source model startup. With their remarkable track record and ongoing influence, Mistral AI continues to shape the future of AI technologies.
Leave a Reply