Artificial intelligence will greatly benefit all of humanity. But do we really want to leave this revolutionary technology to just a handful of US tech companies?
Silicon Valley is experiencing more than a little moral disappointment. Google withdrew its “do no evil” pledge before firing the star ethicist. Self-proclaimed “free speech absolutist” Elon Musk bought Twitter to censor political speech, retaliate against journalists, and facilitate access to the platform for propaganda activists in Russia and China. Facebook lied about enabling Russian interference in the 2016 US presidential election, and paid public relations firms to blame Google and George Soros instead.
These and myriad other ethical blunders should prompt us to consider whether we want technology companies to have the additional ability to learn about our personal information and influence our everyday decisions. Tech companies already have access to our daily whereabouts and search queries. Digital devices monitor many aspects of our lives. We have cameras in our homes and heart rate sensors on our wrists that send what we detect to Silicon Valley.
Tech giants are now developing ever more powerful AI systems that do more than simply monitor users. They actually interact with you and others on your behalf. If searching Google in the 2010s was like being watched by surveillance cameras, using AI in the late 2020s will be like having a butler. You will be happy to include them in every conversation you have, in everything you write, in every item you shop for, in every want, in every fear, everything. It will never be forgotten. And even though you depend on it, it will secretly work to further the interests of any of these commercial companies.
There’s a reason Google, Microsoft, Facebook, and other tech giants are leading the AI revolution. Building a competitive Large Language Model (LLM) like the one that powers ChatGPT is incredibly expensive. In addition to accessing large amounts of data, a single model training run requires over $100 million in computational costs. It also requires technical expertise. Technical expertise is becoming more open and available, but is still concentrated in a few companies. Efforts to break the AI oligopoly by funding start-ups have seen Big Tech profit from the cloud computing services and AI models that power those start-ups, eventually buying the start-ups themselves. It is self-defeating because it often leads to
But enterprises aren’t the only organizations big enough to absorb the costs of training large models. Governments can do that too. The time has come to take AI development out of the monopoly hands of private companies and start introducing it into the public sector. The United States needs a government-funded, government-led AI program to build on the technical expertise held by federal agencies to develop widely reusable models for the public good. increase.
So far, the debate on AI regulation in Washington has focused on the governance of private sector activity, and the U.S. Congress is in no rush to push this forward. Not only should Congress rush to push AI regulation, but it needs to go one step further and develop its own program for AI. Legislators need to reframe the AI debate from one about public regulation to one about public development.
AI development programs may be subject to public opinion and may be subject to political scrutiny. The aim could be to address critical issues such as privacy protection, low wages for tech workers, the dreaded carbon footprint of AI, and misuse of unlicensed data. Compared to keeping AI in the hands of ethically questionable tech companies, the public alternative is both ethically and economically better. And the switch should happen quickly. By the time AI becomes a critical infrastructure essential to broader economic activity and everyday life, it will be too late to start.
Other countries are already there. China has highly prioritized public investment in AI research and development, betting on select giants that are ostensibly private but widely understood to be an extension of the state. The government has ordered the likes of Alibaba and Huawei to develop products that support the larger ecosystem of state surveillance and authoritarianism.
The European Union is also actively promoting AI development. The European Commission is already investing €1 billion a year in AI and plans to increase that to €20 billion a year by 2030. The funds will be donated to a continent-wide network of public research institutes, universities and private companies working together on development. Different parts of AI. Europeans are focusing on knowledge transfer, development of the technology sector, use of AI in public administration, mitigation of security risks and protection of fundamental rights. The EU also remains at the forefront of actively regulating both data and AI.
Neither the Chinese model nor the European model is always right for the United States. State control of private enterprise remains anathema in American political culture and will be difficult to gain mainstream traction. Tech companies and their supporters in both US political parties oppose strong public governance of AI. But the U.S. government can draw inspiration from China’s and Europe’s long-term plans and leadership on regulation and public investment. With the global economic value associated with AI set to reach tens of trillions of dollars, the risks of international competition cannot be ignored. Just as energy research and medical research have their own federal agencies at the Department of Energy and the National Institutes of Health, respectively, his AI research and development arena within government.
Besides the moral arguments against letting private companies develop AI, there are also strong economic arguments in favor of the public option. Publicly funded LLMs will serve as an open platform for innovation, helping small businesses, nonprofits, and individual entrepreneurs build AI-assisted applications.
There are also practical discussions. Building AI is within the reach of ordinary people because governments don’t have to own and operate the entire AI supply chain. Chip and computer manufacturing, cloud data centers, and a range of value-added applications that integrate AI with consumer electronics and entertainment software do not need to be publicly controlled or funded.
One reason for skepticism about public funding of AI is that AI will be of lower quality and less innovation, given reduced incentives due to increased ethical scrutiny, political constraints, and lack of market competition. that it may be delayed. But even so, it’s worth making the most important technology of the 21st century widely accessible. And public AI is by no means at a disadvantage. The open source community proves that the most innovative are not necessarily private companies.
Those concerned about quality trade-offs might suggest a public buyer model, where Washington licenses or purchases private language models from big tech rather than developing them on their own. But that alone is not enough to ensure that tools are aligned with public priorities and address public needs. Detailed insight and control over the inner workings and training procedures of these models cannot be provided to the public and still require strict and complex regulations.
While there is political will to act on AI development through public rather than private funding, this is not yet aligned with the will to establish a fully public AI development agency. A task force set up by Congress in January recommended $2.6 billion in federal investment in computing and data resources to advance the U.S. AI research ecosystem. However, the investment will primarily drive the interests of big tech companies, leaving opportunities for public ownership and oversight.
Nonprofits and academic institutions have already created open access LLMs. These should be celebrated, but they are no substitute for public choice. Nonprofit projects, even charitable ones, still rely on private interests. As when OpenAI effectively abandoned its not-for-profit origins, these private interests could change without public opinion, and its founding intentions and operations were subject to market pressures, capricious donors, and leaders. I don’t know if it will survive the alternation of
The U.S. government is by no means a perfect beacon of transparency, keeps our data safe and responsible, and does not truly reflect the interests of the people. But the risk of leaving AI development entirely in the hands of an apparently unreliable Silicon Valley company is too high. AI, like any technology, affects the general public and should be developed by the general public.
