The Wall Street Journal recently covered Sam Altman’s intention to secure funding of up to $7 million to revolutionize the global semiconductor industry in support of artificial intelligence. This underscores the remarkable market dominance of relational AI, where a single company’s funding goal could surpass Japan’s entire gross domestic product without raising eyebrows.
The monopolization of AI is becoming increasingly crucial, encompassing everything from potential breakthroughs in healthcare to the risks of election manipulation, the aspirations for impactful climate studies, and the challenges of unraveling complex physics phenomena.
Nevertheless, the current trend leans toward a scenario where a handful of corporations wield significant influence in advancing cutting-edge AI technologies, particularly evident in the realm of training extensive language models for diverse AI applications. This process demands substantial data and computational resources, commonly referred to as “compute.” Without embracing the latest innovations, researchers and small to medium-sized enterprises risk falling into a perilous reliance on Big Tech giants once again.
To level the computational playing field, substantial public investments are being made on both sides of the Atlantic. The US government is launching the National AI Research Resource next month, overseen by the US National Science Foundation, to ensure that experts have access to resources comparable to those of Silicon Valley titans. This initiative involves collaboration with 25 civil society organizations, 10 additional federal agencies, and government-backed data and computing facilities to facilitate AI research and education.
Anticipating the surge in conceptual AI advancements, the EU had established a decentralized network of supercomputers back in 2018 with a similar objective. Despite the EuroHPC’s existence, its full potential appears underutilized. European Commission President Ursula von der Leyen emphasized the need to harness this energy, suggesting that democratized access to computational resources could foster the creation of “AI factories,” where smaller enterprises pool resources to develop state-of-the-art models.
Recognizing the critical role of internet access in education, careers, and data acquisition, there have been ongoing discussions about treating internet access as a public utility. However, concrete guidelines for this purpose have been scarce. The US and EU are now demonstrating a genuine commitment to democratizing digital infrastructure by providing universal access to computational resources.
These recent initiatives mark a long-overdue effort to shape the online landscape and counterbalance the overwhelming influence of major tech corporations across various societal domains, albeit under the guise of industrial policy.
By expanding access to essential computational resources, these governments are making strides in the right direction. However, these investments must be complemented by legislative and regulatory measures as they represent just the initial phase. Competitive entities must remain vigilant against the exponential growth of dominant AI firms, while security agencies must prevent malicious actors from exploiting critical computing resources.
Addressing the inherent biases and prejudices in AI software poses a challenge for anti-discrimination watchdogs. Similarly, investments in open AI initiatives support policies aimed at preventing market monopolies from transforming into knowledge monopolies. While the Digital Services Act outlines obligations for software companies, the AI Act lacks explicit provisions in certain areas, such as data access for academics.
As public investments in digital infrastructure continue to rise, there is a potential shift of state funds away from Big Tech, even for projects serving the public good. The US government’s $3.3 billion investment in AI in 2022, though substantial, falls short of Altman’s billion-dollar target or the tens of billions that the industry invests annually.
Creating an environment that discourages AI conglomerates is crucial for enhancing public comprehension of the technology. These objectives align in promoting academic research as the cornerstone of valuable advancements, emphasizing the importance of preserving this vital ecosystem.
