Is regulation-induced innovation an oxymoron?

What DeepSeek tells about it?

This is the next in a series of excerpts from my paper “Governance at a Crossroads.”

ChatGPT-4o view of the issue

Some of the pushback against regulation assumes that it necessarily stifles innovation. What if the right type of regulation could be used to trigger innovation? To answer the question, we have looked at historical precedents for using policy and regulation to drive innovation. Engineering and Computer Science inventions are needed to solve open issues in this field, such as reliability and hallucinations, model bias, safety and control, high computational demands and associated environmental impact, data privacy and security, transparency and explainability, and diminishing returns on scaling. Regulation can act as one of the tools to align industry incentives to invest in this direction.

The views shared fifty years ago with the Senate Commerce Committee by researchers from the Center for Policy Alternatives at the Massachusetts Institute of Technology are instructive. The context for the testimony was the criticism of environmental regulations at that point in their infancy. They argued that safety regulations, while often criticized for imposing costs on industries, can serve as powerful drivers of innovation. These regulations can act as catalysts by compelling firms to rethink and adapt their products, processes, and technologies. By introducing new constraints, such as required safety enhancements and regulations, engineers and firms expand the dimensions they must consider when solving problems, encouraging creativity and resourcefulness, and leading to innovative solutions that might not have emerged otherwise. Moreover, regulations create new markets for technologies that address safety concerns. This demand stimulates the development of devices, advanced manufacturing processes, and safer products. The researchers highlighted that such regulation-induced innovation is particularly evident in mature or concentrated industries and that, contrary to the view that regulations stifle growth, they can drive technological progress and economic benefits, proving to be instrumental in allowing industries to adapt to societal demands. The parallels to this moment are uncanny.

A decade later, by 1987, the debate was still raging, as covered by William J. Broad, a science journalist, at The New York Times in the article “Does the Fear of Litigation Dampen the Drive to Innovate?” He explored the concern that safety-related litigation and regulatory constraints may hinder technological innovation across various fields. Experts at the time argued that the fear of liability lawsuits had created a “chilling effect,” deterring venture capital and slowing advancements in areas such as artificial intelligence, food processing, and nuclear engineering. Critics pointed out that regulators and the legal system focused excessively on risks, discouraging experimentation and innovation. However, consumer advocates countered that liability laws ensure product safety and accountability. They emphasized that companies had been compelled to improve safety rather than abandon innovation. They argued that the rise in product liability cases had heightened corporate responsibility. In that context, proposals to address the issue ranged from enhancing regulatory incentives for safety innovations to shifting decision-making from courts to regulatory agencies. He pointed out that balancing safety and innovation remained challenging, and that safety should be integral to engineering practices. The debate over private rights of action continues today as a disputed matter in several legislative proposals.

The framework for thinking about regulation as restrictive or expansive, proposed by Nicholas Ashford, is valuable as we analyze a path forward for digital technologies policy. Restrictive regulations impose constraints, increase costs, and limit choices, often delaying activities due to compliance requirements. They are seen as burdensome, focusing on controlling harmful practices. In contrast, expansive regulations stimulate innovation by introducing new performance dimensions, compelling firms to develop creative solutions and adapt technologies. The relationship between regulation and innovation continued to be a subject of scholarly discussion following his testimony. Independently, in the 1990s, Michael Porter’s work introduced the Porter Hypothesis, suggesting that stringent environmental regulations could stimulate innovation by encouraging incumbent firms to develop new technologies and processes that offset compliance costs. Porter’s hypothesis overlooks new entrants, focusing on how incumbents respond to stricter regulations, creating innovation offsets. Early adopters gain learning curve advantages, forcing latecomers into costly compliance. Building on the original work from MIT, proponents of a strong form of the regulation-induced innovation hypothesis suggested that stringent regulation can drive the emergence of new products and processes, potentially displacing dominant technologies and incumbent firms lacking the ability or willingness to adapt.

Plus ça change, plus c’est la même chose - the more things change, the more they stay the same. Today, the Organization for Economic Co-operation and Development (OECD) emphasizes the importance of effective regulatory frameworks in fostering innovation through its “Better Regulation and Innovation” initiative. Recognizing that poorly designed regulations can hinder technological progress and economic growth, the OECD advocates for regulatory policies that are flexible and forward-looking. The organization promotes adaptive governance, encouraging regulatory approaches that can evolve alongside technologies and market dynamics. They seek to position regulation as a catalyst for innovation, ensuring oversight while maintaining the flexibility necessary for technological progress. The OECD Observatory of Public Sector Innovation (OPSI) empowers governments to leverage innovative approaches to achieve policy priorities. It identifies trends, fosters shared learning, and enhances innovation capacity, including tools like regulatory sandboxes to test and implement solutions, such as AI technologies, in controlled environments.

U.S. export control policies on advanced technologies, such as GPUs, imposed on China illustrate how restrictive regulations can drive engineering innovation. These limitations have compelled Chinese AI developers to optimize resource utilization, create more efficient models, and explore alternative hardware and software solutions. This dynamic reflects a broader pattern where restrictive regulations can foster creativity and adaptation despite constraining resource access. DeepSeek is a Chinese AI firm founded by the hedge fund High-Flyer. They are known for open-source large language models, including the latest DeepSeek-V3, an ultra-large model designed to rival leading global models such as Meta’s LLaMA. Despite resource constraints due to U.S. export controls, their latest release has demonstrated exceptional performance by employing innovative engineering strategies. These include advanced model compression, efficient data utilization, and hardware-software co-design, enabling optimal performance on limited hardware. This case underscores how resource constraints can catalyze technological ingenuity, driving innovation and enhancing self-reliance in AI development. There is skepticism about some of DeepSeek’s training cost claims, but even the skeptics recognize their resourcefulness. A member of a frontier AI lab in the U.S. told us: “There are some questions, […] could they have done it without the existence of other foundational models? A little bit unclear, but definitely pretty disruptive.” It is still too early to tell the impact of export controls. These types of restrictions have a lagging effect, meaning that today’s restrictions to access the latest technology will impact future products being developed now. When you couple this with the increased need for processing power for inferencing, resulting from DeepSeek’s recent popularity, the chip export controls can impact their ability to maintain the current improvement trend. Alternatively, this can further motivate indigenous GPU designs.

Contemporary studies continue to examine the nuanced impacts of regulation on innovation. Regulators are trying to balance innovation and consumer safety by using tools like sandboxes to foster new technologies while ensuring protection. Soft law instruments enable rapid adaptation to change. Risk-based, customer-focused approaches improve regulator-business relationships while digital technologies streamline processes, maintaining consumer protection without imposing unnecessary burdens on businesses.