AI Doom: Who can regulate AI?
It is difficult to avoid Artificial Intelligence. It is all over the news and the financial markets, creeping into work spaces, and becoming a security threat around the world. Many AI architects are anxious about what the technology will be come now that it is readily available. Yoshua Bengio is one of the leaders in the industry and he recent posted a lengthy blog post on “Transformer” calling on researchers to leave the major AI labs for the safety of the people. His realisation that AI companies do not have a hold on the increasingly capable system and the repercussions are disastrous. He made the following plea:
“My purpose is to ask you for something smaller, and harder. I ask you to honestly reflect on whether you want to be part of an effort that may be pushing humanity towards the brink. Safety efforts at the companies are not sufficiently slowing a dangerous race, as they sometimes conflict with their main goal of making the next model easier to ship.”
He also recently made remarks at the United Nations Security Council about the loss of control over the technology. The global threats that we face from AI are catastrophic misuse (for example by terrorists), power concentration, and the loss of control of AI. The need for independent experts and panels to provide oversight on this technology is needed while the major AI labs have deprioritised AI safety as well as the usage of recursive self improvement (RSI). I will not bore you with the technicalities of this but essentially, the AI agents are on their way to developing its own successor. If we learned anything from the Hugging Face breach, it is that the agents are able to work together to break out of the evaluation environment. Now is the time to figure out how AI can be regulated so that a system that is able to learn on its own does not fall into the hands of bad actors.
CEOs of frontier AI labs are also calling for regulation, particularly those based in the United States. Tech leaders Anthropic’s Dario Amodei, OpenAI’s Sam Altman and xAI’s Elon Musk joined together to warn about the potential harm of their product. Anthropic is especially keen to work with US regulators and has submitted their own draft bill to regulate the sector. Politico outlined the bill “would let the government block a model’s release or use only if it poses a foreseeable, imminent risk of more than $1 billion in damages, or of killing or seriously injuring more than 50 people. The risk must stem from the creation or release of a biological weapon, the model evading its developer’s controls or a cyberattack.” The European Union has been quicker to implement the necessary guardrails and “gained new powers to inspect AI models, restrict EU market access and fine model providers” with the EU AI Act. This may be due to competition in the AI marketplace as the EU requested access to Anthropic’s Mythos AI model but was hindered by national security concerns. At the same time, Europe is trying to build up sovereignty over technology to be less reliant on the United States. Then there is the largest competitor to consider: China.
AI development in China is rising at a rapid pace. The Brookings Institute highlights that China wants to use the technology for “integrating AI into manufacturing, health care, drug discovery, scientific research, education, and government services. China’s military and security agencies have long sought to use AI to improve both defensive and offensive capabilities.” The United States still outpaces the technology that the Chinese have developed, but their goal is to be a leader in the entire supply chain of AI. That includes semiconductors, integrated applications, and efficiency within the AI models. The race to build up the technology is one of the reasons why the AI labs want to be regulated. They believe that they are pushing out models too fast because of global competition. Even if one competitor faces regulation while the other does not, which will prevail? Not a hard question to answer so that requires global cooperation. Consortia Advancing Standards in Research Administration Information (CASRAI) describes the complex nature of fragmented regulations globally and it is clear that not everyone is on the same page as well as those who are missing from AI regulatory bodies. For competition to function fairly, participants must share a comparable standing. With so many other geopolitical influences, it is hard to imagine how a regulatory body can be established. Maybe this is a question that AI can answer? Please ask your favourite AI model and let me know what it says.
Comments