Baidu CEO Robin Li on Tuesday said Chinese institutions and technology companies have released over 70 large language models (LLMs) with over one billion parameters, further escalating the ongoing US-China trade war that has seen both the nations vie for supremacy in technology prowess, including chipmaking, quantum computing, and AI.

Baidu last week secured regulatory approval in China to launch its AI chatbot which is seen as China's response to OpenAI's ChatGPT. Li claimed that the AI chatbot, Ernie 3.5, is capable of twice the processing power compared to its predecessor, with 50% improved efficiency.

Li’s comments come on the heels of several Chinese companies releasing ChatGPT-like generative AI assistants in the domestic market after getting due approvals from regulators, according to a news report from ChinaDaily. These companies included the likes of TikTok-owner ByteDance, SenseTime, Zhipu AI, MiniMax, and Baichuan Intelligent Technology.

On the other hand, several US-headquartered technology companies and institutions have also released several large language models over the past few months including Meta’s Llama 2, OpenAI’s GPT 3.5 and GPT-4, Google’s PaLM 2 and LaMDA, IBM’s watsonx models, Anthorpic’s Claude 2, Databricks’ Dolly, and Amazon’s Titan foundation models.

While there is no definitive number for the total number of models released by US institutions and companies till 2023, data from a Stanford University study showed that the majority of the world’s large language and multimodal models (54% in 2022) are produced by American institutions.

The same study also showed that collaboration between Chinese and American organizations for developing large language models had dropped severely post-2020.

The US put restrictions on the export of chipmaking technology to China in an effort to stop China’s growing geopolitical influence, including the country’s technology prowess.

In contrast to the period between 2010 and 2020, which saw the number of AI research collaborations between the United States and China increase approximately four times, the period between 2020 and 2021 saw the most negligible (2.1%) year-on-year growth since 2010, the study showed.

In addition, the study claims that the US still leads the world in terms of total private investment in artificial intelligence. In 2022, the $47.4 billion invested in the US was roughly 3.5 times the $13.4 billion invested in China.

Further, the US also continues to lead in terms of the total number of newly funded AI companies, according to the study. For 2022, the number of newly funded AI companies in the US was 1.9 times more than the European Union and the UK combined and 3.4 times more than China.

Analysts believe that China is progressing rapidly in the LLM domain, and it’s hard to predict who wins the race for supremacy.

“Their LLM models are growing at a fast pace as big Chinese tech giants such as Baidu and Alibaba are doubling down in the space,” said Pareek Jain, principal analyst at Pareekh Consulting. “As Chinese firms try to benchmark their models against US-developed models such as GPT-4, they might have an edge in the domestic market due to the availability of large data sets in local languages.”

The Stanford University study also showed that China was catching up with the US in terms of AI research papers and citations. While China still lagged in the number of AI research papers and citations in the non-profit, government, and industry sectors, it has overtaken the US in the education sector.  

The ongoing trade war between the US and China has also seen US lawmakers urging the Biden administration to take more action to impede China’s progress in gaining dominance in areas such as artificial intelligence and quantum computing.

These suggestions included imposing trade restrictions on Chinese memory chip maker Changxin Memory Technologies as a counteroffensive to China banning the use of Micron’s chips in order to tactfully pressurize the US to convince the Netherlands and Japan to join the chip-making technology ban on China.