The AI field has entered an era of fierce competition, but the business model is still being explored.

Last month, the AI industry erupted in an "animal war." On one side is the Llama model launched by Meta, which is highly favored by developers due to its open-source nature. On the other side is a large model named Falcon, which has topped the open-source LLM rankings after its release in May this year, surpassing Llama.

Interestingly, the developer of "Hawk" is the Technology Innovation Institute of the UAE. The UAE's Minister of Artificial Intelligence was subsequently selected as one of the "100 Most Influential People in AI" by Time magazine.

Today, the AI field has entered the "chaotic stage". Many countries and companies are building their own large language models. In the Gulf region alone, Saudi Arabia has purchased over 3,000 H100 chips for domestic universities to train LLMs.

The emergence of this phenomenon can be attributed to the Transformer algorithm paper published by Google in 2017. The Transformer addressed many of the shortcomings of earlier neural networks and has become the foundation of all large models today. It transformed large models from theoretical research into purely engineering problems.

As the open-source community becomes increasingly active, the performance of various LLMs may converge. The real core competitiveness lies in ecological construction or pure reasoning ability. Currently, GPT-4 is still far ahead of other models in terms of performance.

However, the high cost of computing power has become a hindrance to the development of the industry. Sequoia Capital estimates that global technology companies will spend $200 billion annually on large model infrastructure, but revenues will only reach a maximum of $75 billion, resulting in a significant gap.

With a few exceptions, most AI companies have not found a clear profit model. Even software giants like Microsoft and Adobe face difficulties in pricing their AI services.

Overall, although the AI revolution is still in its early stages, the business challenges faced by pure large model providers are intensifying. The key to the future may lie in how to integrate AI technology with real-world application scenarios to create real user value.