It is no big secret that the development of artificial intelligence arouses great interest not only of large corporations, but also of world powers. We are talking here not only about the USA, which is considered a kind of leader, but also about China. It has been clear for some time that the Middle Kingdom also has great ambitions in this field and does not intend to give up on them.
Now we can talk about a kind of breakthrough in this matter. The latest reports coming from that country confirm that the Xi Jingping government’s efforts towards technological independence are progressing.
Unequal competition between Chinese companies and Western competitors
NVIDIA H100 is something that Chinese AI companies will not see, or at least that is officially the case.
Companies from the Middle Kingdom have long been trying not to keep up with Western competitors, but to catch up with them in the race related to the development of AI. The situation is, of course, not that simple, because we are talking about competition on not particularly equal terms due to American export restrictions.
For this reason, the country ruled by Xi Jingping cannot import NVIDIA A100 and H100 systems (or even RTX 4090 series graphics cards). This forced local companies developing AI to create an alternative, thus becoming independent from Western suppliers.
The result was the development and production of its own chips that would power the development of Chinese AI. This art was of course not easy, but it was successful. Thanks to this, one of the local companies was able to announce what many observers call a real “breakthrough”..
China Telecom has created two LLM models using state-owned arrangements
China Telecom (Source: Free Malaysia Today)
China Telecom, one of the Chinese communications companies, confirmed the creation of two LLM models (The. Large Language Model), which were trained only on chips created in China. This is an event that could significantly impact the entire AI market and break it from the American primacy.
TeleChat2-115B and another unnamed model were trained on tens of thousands of chips. According to China Telecom, this nameless project will have as many as a trillion parameters, i.e. variables that are responsible for AI learning.
A statement shared by the company’s AI Institute said that this is an achievement that demonstrates China’s ability to independently train AI. Additionally, the aspect of technological independence was highlighted, which may herald what has been termed a “new era of innovation”.
Presentation of the Huawei Ascend 910 chip
Of course, this sounds very spectacular, but the mystery in this case is the origin of the chips used to power the AI. China Telecom did not specify the exact models, but the most likely use is Huawei Ascend systems.
This giant has been creating this type of equipment for some time and it can be considered the leader in this segment. Back in August this year, Huawei confirmed that it had managed to develop a new model of the system called Ascend 910C, which was even to be tested by the first customers, and its performance is to be comparable to the previously mentioned NVIDIA H100, originally released in 2022.
China’s path to independence
The Middle Kingdom, by developing its own language models and creating systems that will power them, may actually achieve technological independence. Importantly, it may happen exceptionally quickly, because the pace of development – considering the rather unfavorable geopolitical conditions – is truly impressive.
Of course, in this situation, much will depend on the interest in such “state chips” and whether companies responsible for AI development will want to use them. It is possible, however, that as the gray zone is combated (and the gray zone is quite large in the Middle Kingdom), these companies will remain without a solution.
So can we say that economic constraints have become the mother of innovation? Absolutely, although much will depend on whether the chips and AI itself offer a satisfactory level and capabilities. Xi Jingping’s government seems determined to make this happen, but – as I think we all know – intentions alone are not everything.