Zambia Copper Discovery Shows AI Accelerating AI Research

Improvements to AI capabilities will only accelerate. The copper discovery in Zambia is just one example.

It’s commonly accepted that there are three drivers of progress in AI: computing power, algorithms, and data. For better or for worse, AI itself can help with improvements along each of those dimensions. 

Take for instance the recent discovery of copper in Zambia. While Copperbelt Province was known for, well, copper, KoBold’s machine learning tool was able to identify a previously undiscovered mile-long stretch of copper. If transformed into an operational mine, the copper is in high demand. It could be used to build semiconductor chips, power data centers, and form physical components of these centers such as power cables and heat exchangers. All of this could contribute to computing power, enabling further AI development. 

Beyond computing power, we have already seen AI assist with the development of algorithms and data. Many models can write code and others have been used to conduct reinforcement learning, though there are concerns about this approach. Research is already underway to find ways to train AI systems on AI-generated data. 

These are only a few examples of why, as time goes on, the ways in which AI can enhance itself will multiply. Recent improvements in AI have been rapid and will only continue to accelerate.

This acceleration has important implications for policy making. Often policy makers wait until an industry is relatively mature so that they can observe it before regulating. In the case of AI, we don’t have the luxury of time that we did with cars, electricity, and the internet. As AI capabilities develop, there is a risk that powerful tools fall into the wrong hands or that they are misaligned to human values. 

The US needs concrete steps towards AI safety such as whistleblower protections for AI employees, cybersecurity standards for AI companies, or government simulations to prepare for AI-enabled cyberattacks. The EU and soon, the UK, will have taken strides in that direction. Given the urgency and scale of these risks, we need Congress to act now. 

‍The Center for AI Policy (CAIP) has a 2024 action plan and full proposed model legislation. We encourage you to visit both for specific policy measures to ensure safer AI.