OpenAI CEO Sam Altman (who is testifying here before the U.S. Senate) is one of the signatories of an open letter warning of the risk of human extinction from AI.Credit: Win McNamee/Getty
It’s rare for industry leaders to talk about the potentially lethality of their products. It’s not something tobacco or oil executives tend to do, for example. But hardly a day goes by that tech industry players don’t yell at the existential risks of artificial intelligence (AI).
In March, an open letter signed by Elon Musk and other engineers warned that gigantic AI systems pose grave risks to humanity. A few weeks later, AI tool pioneer Jeffrey Hinton resigned from his research position at Google, warning of the significant risks posed by the technology. More than 500 business and science leaders, including representatives of OpenAI and Google DeepMind, were named in the 23-word statement, saying that addressing the risk of human extinction with AI “is a global priority alongside the risk of human extinction.” It should be a matter,” he said. Pandemics and Nuclear War.” And on June 7, the UK government announced it would host the first major global AI safety summit this fall, citing AI’s potential existential dangers.
The idea that AI could lead to the extinction of humanity has long been debated on the fringes of the tech community. The excitement over ChatGPT tools and generative AI has now pushed ChatGPT into the mainstream. But like a magician’s trick, it distracts attention from the real problem: the risks that AI systems and tools pose now and in the future. Governments and regulators in particular must not be distracted by such stories and must take decisive action to curb potential harm. And while their work should be informed by the tech industry, it should not be bound by the tech agenda.
Ethical AI battle at the world’s largest machine learning conference
Many AI researchers and ethicists Nature Those dissatisfied with the apocalyptic dominance of the AI debate are speaking out. There are at least two problems with this. First, the specter of AI as an all-powerful machine fuels a race to develop AI so that nations can benefit from and control it. This will favor tech companies, encourage investment, and weaken the argument to regulate the industry. A real arms race to create the next generation of AI-powered military technology has already begun, increasing the risk of catastrophic conflict. Possibly apocalyptic, but not much discussed in the prevailing “AI is an apocalyptic threat” theory.
Second, a homogenous group of corporate executives and technologists will be able to monopolize the AI risk and regulation conversation, leaving other communities behind. Amba Kak, director of her AI Now Institute in New York City, which focuses on the social impact of AI, said the letter written by the tech industry leader was “essentially asking who counts as an expert in this conversation.” We are drawing boundaries about what is possible,” he said.
AI systems and tools have many potential benefits, from synthesizing data to assisting in medical diagnosis. But they can also cause well-documented harm, ranging from biased decision-making to job cuts. AI-powered facial recognition is already being abused by authoritarian states to track and oppress people. Biased AI systems could use opaque algorithms to deny welfare benefits, medical care, and asylum, the applications of technology most likely to affect people in marginalized communities. I have. Discussions on these issues are starving.
One of the biggest concerns around modern species-generating AI is its potential to feed misinformation. This technology allows more convincing fake texts, photos and videos that can influence elections, undermine people’s ability to trust any information, and potentially destabilize society. Makes it easier to create more. If tech companies are serious about avoiding or mitigating these risks, ethics, safety, and accountability must be at the heart of their work. At the moment, they seem reluctant to do so. OpenAI “stress tested” its latest generative AI model, GPT4, by encouraging the generation of harmful content and introducing safeguards. However, while the company explained what it did, it did not disclose the details of the tests or the data used to train the model.
Face Recognition Research Needs Ethical Considerations
Tech companies must develop industry standards for the responsible development of AI systems and tools and conduct rigorous safety testing before releasing products. In the same way that pharmaceutical companies must submit clinical trial data to medical authorities before launching a drug, they must submit complete data to an independent regulatory agency that can validate the data.
To do so, governments must not only apply existing laws, but also establish an appropriate legal and regulatory framework. Earlier this month, the European Parliament approved an AI law to regulate AI applications within the European Union according to their potential risks, banning police from using live facial recognition technology in public places, for example. There are further hurdles to clear before the bill can be enacted in EU member states, and the lack of details on how it will be implemented is questionable, but the prospects for establishing a global standard for AI systems are questionable. may be useful for Further consultations on AI risks and regulation, such as the upcoming UK Summit, will include researchers studying the harms of AI and representatives of communities that have been harmed by, or are particularly at risk from, this technology. You need to invite a diverse participant list.
Researchers must do their part by building a responsible AI culture from the bottom up. In April, the large machine learning conference NeurIPS (Neural Information Processing Systems) announced the adoption of a code of ethics for conference submissions. This includes the expectation that research involving human participants has been approved by an ethics committee or institutional review board (IRB). All researchers and institutions should follow this approach, and the institutional review board (or peer review board if no institutional review board exists) should have the expertise to investigate potentially dangerous AI research. You have to make sure you have the knowledge. And scientists using large datasets containing data from people must find ways to get consent.
Fear-mongering narratives about existential risk are not constructive. A serious discussion of the actual risk and action to contain it. The sooner humanity establishes the rules of how to interact with his AI, the sooner it will learn how to live in harmony with technology.
