The recent move of the Biden administration to confine China’s access to specific software packages, such as that employed in OpenAI, which powers ChatGPT, has expanded the scope of discussions as to the feasibility of curbing the spread of artificial intelligence. Concerns arise as open-source AI models become popular because, to some extent, these regulations can be rendered ineffective by this growing new trend.
US tightens AI export controls
The US government tries to prevent the spread of AI tools that may be used for security violations by increasing export controls to stop supplying AI models that belong to particular owners. These new measures, which cover China and a few other countries, such as Russia and Iran, will make it harder to cognize and obtain technologies that are meant for warfare or cyber-attacks. It can be denominated as customized models, which are not open-source and are created by US-based tech leaders such as OpenAI, Google DeepMind, and Anthropic.
While the US Commerce Department indulges in the process of drafting new regulations to implicate which control, on the other hand, the US intelligence community is working on new ways to collect and analyze data to further target and disrupt terrorist activities within and beyond the US borders. Such sanctions come after years of prior blocks on exporting better AI apparatus to China, which have significantly hit the largest US producers, such as Nvidia. Nvidia has hence begun to ship out less capable chips that need no special export license to appeal to the markets outside the US.
Open-source AI models challenge export curbs
However, even though these regulations are specifically aimed at excluding embedded AI models, there is a growing trend of embracing open-source models in the industry, which, in turn, may, in some instances, result in the weakening of these measures. Open-source models are open, which means the code and training data are those that anybody can get from anywhere globally. This is clearly demonstrated by Meta’s recent plan to release its LLaMA 3 model as open source and Google’s intended roll-out of the family of open models, and it leads to questions about whether there will be a desired achievement if the access to proprietary models will stay closed.
GlobalData senior analyst Josep Bori remarks that the movement toward open source makes the regulations less meaningful due to the fact that there is no patent protection; hence, any non-proprietary model can be accessed free of charge and used by everyone. Although it seems easy to do that so far, he argues that controlling AI technology is even harder than we know.
Growing limitations in the East due to the restriction of AI diffusion in this way is capable of splitting the sphere of technology and developing long-term challenges that might affect the system of global artificial intelligence. Christoph Cemper, CEO of AIPRM, is rather skeptical that the end-of-the-world case scenario would not hinder the development of secure AI systems. Up until now, doing AI research has mostly been a global endeavor, and international cooperation has been on the rise. Still, authoritarian countries could try to split the field, and this could have a negative impact on progress.
The situation could arise of the basement ones as another parallel AI ecosystem could interfere with global trade and thus may have a great impact on businesses depending on cross-border supply chains. Existing practice, which is dependent on cooperation between US firms and Huawei companies, is in danger now because new restrictions might break up this long-time practice. A collateral effect of AI that has now obviously affected access seems unwise when it is cooperatively done to address the grand challenges that exist today for humanity, says Cemper.
Global implications for the AI market
The AI market is surely still experiencing some restricting issues, which shows a high level of growth. A market analysis by GlobalData reckons that the world market of AI will hit a value of $909 billion by 2030, and the compound annual growth rate (CAGR) from 2022 will be 35%. Henceforth, the AI chips market is forecasted at about £116 billion in total value by 2030 due to the increasing demand from different sectors of the economy.
Evolving regulations could transform the global value chains and restructure industries, which could sound like the death knell for companies’ profits if they have to devise new business strategies. It is highly speculated that the recent constricting of export controls is directly proportional to a fierce struggle to win over technology firms from the US and China. It is a catalyst for innovation on both sides, with a potential reduction in collaboration.
The competency of the US administration’s solution will be predicted by how the mentioned measures manage to decelerate competitors’ frenzied procurement of AI technology without compromising innovation or adding to the two countries’ economic woes.