大规模失业、地球沸腾以及十年内人类灭绝：人工智能的悲观愿景已被反复论述。学者们现在主张，欧盟政策制定者应确保帮助资助这项技术的纳税人能够获得回报，即“幸福红利”。
Mass unemployment, boiling the planet, and human extinction within a decade: the dystopian visions of AI are well rehearsed. Academics now argue that EU policymakers should ensure the taxpayers helping to fund the tech get something back, a “happiness dividend.”
由桑德拉·瓦赫特（Sandra Wachter）教授领导的牛津互联网研究所（OII）研究人员表示，无意中协助创造下一代科技亿万富翁的公民，需要的不仅仅是人工智能将促进经济增长和增加企业利润的承诺。
Researchers at the Oxford Internet Institute (OII), led by Professor Sandra Wachter, say citizens who are inadvertently helping to create the next generation of tech billionaires need more than promises that AI will boost economic growth and fatten corporate profits.
他们的论文题为《优化欧盟技术政策中的幸福感，以回报人工智能的公共补贴》（Optimising for happiness in EU technology policy to repay public subsidies for AI），发表在《皇家学会开放科学》（Royal Society Open Science）杂志上。论文指出，公众通过资金、税收优惠以及对土地、水、能源、基础设施和数据的访问，支持了人工智能的发展。
Their paper, "Optimising for happiness in EU technology policy to repay public subsidies for AI," published in the Royal Society Open Science, notes that the public backs AI development through funding, tax breaks, and access to land, water, energy, infrastructure, and data.
在数据方面，大多数人对此没有发言权。创意作品、社交媒体帖子和用户互动都被席卷一空，用于训练前沿模型，并引发了一连串的诉讼。
On data, most people have no say in the matter. Creative work, social media posts, and user interactions have all been swept up to train frontier models, spurring a stream of lawsuits.
OII团队建议将幸福感作为人工智能巨头的第三大目标，与经济增长和风险缓解并列。研究人员对我们表示：“使技术危害减小并不等于使其对社会有益。”
The OII team proposes happiness as a third goal for AI heavyweights, alongside economic growth and risk mitigation. “Making technology less harmful is not the same as making it socially beneficial,” the researchers told us.
什么能让人感到快乐？研究人员指出，这包括归属感、胜任感和自主感，而人工智能的一些应用可能会破坏所有这些要素。模仿人类情感的聊天机器人可能会侵蚀人们对真实人类联系的重视。对人工智能的过度依赖可能会使技能变得迟钝。人工智能的影响可能会让人们更难做出自己的选择。此外，数据中心威胁着土地和水资源，留给人们亲近自然的树木和绿地减少了。
What makes people happy? The researchers point to relatedness, competence, and autonomy, all of which some uses of AI could undermine. Chatbots that mimic human emotion may erode the value people place on real human connection. Heavy reliance on AI could dull skills. AI influence could make it harder for people to make their own choices. And datacenters threaten land and water, leaving fewer trees and green spaces for people to connect with nature.
研究人员表示：“根据拟议的方法，受益于公共支持的公司将被期望预测其技术将如何改善幸福感和福祉，并为兑现这些益处承担责任。”
“Under the proposed approach, companies benefiting from public support would be expected to predict how their technologies will improve happiness and well-being, and be held accountable for delivering those benefits,” the researchers said.
欧盟已为人工智能超级工厂拨款200亿欧元，“欧洲生成式人工智能”（GenAI4EU）和“数字欧洲”计划则通过“地平线”生成式人工智能和超级计算研究项目投入了7亿欧元。
The EU has allocated €20 billion for AI gigafactories, while GenAI4EU and the Digital Europe programme is using €700 million through Horizon research projects into generative AI and supercomputing initiatives.
主要作者桑德拉·瓦赫特教授说：“公众已经通过补贴、资源和数据为人工智能付出了代价，但在获得什么回报方面却几乎没有发言权。我们认为，人们理应获得能增加幸福感的技术。如果公众在帮助构建人工智能产业，它也应该分享其带来的好处。”
“The public already pays for AI through subsidies, resources, and data, yet has little say over what it gets in return,” said lead author Professor Sandra Wachter. “We believe people deserve technology that increases their happiness. If the public is helping to build the AI industry, it should also share in its benefits.”
越来越多的数据中心中更密集、更强大的服务器正在消耗更多能源，这可能会推高家庭电费。内存制造商正在优先生产利润率更高的芯片，因此普通的DRAM和NAND成本更高，使用它们的设备也是如此。被一些人视为地球上下一个最珍贵商品的淡水，正被用于冷却服务器机架。
Denser, more powerful servers in a growing number of datacenters are drawing more energy, potentially pushing up household electricity bills. Memory makers are prioritizing higher-margin chips, so ordinary DRAM and NAND cost more, and so do the devices that use them. Water, which some see as the planet’s next most precious commodity, is being used to cool server racks.
Anthropic今年警告称，除非对强大模型进行适当控制，否则人类面临着招致灭绝的风险。去年，该公司与OpenAI提出了大规模失业的可能性，因为雇主转向人工智能来削减成本。
Anthropic warned this year that humanity risks courting extinction unless powerful models are properly controlled. Last year it and OpenAI raised the prospect of mass unemployment as employers turn to AI to cut costs.
研究人员补充说，从模型制造商及其云服务提供商那里获取数据的难度，使得评估数据中心造成的环境破坏变得更加困难。
The researchers add that the difficulty of getting data out of model makers and their cloud providers makes it harder to gauge the environmental damage datacenters are doing.
在此背景下，寻找某种改善人类福祉的方法显得至关重要。模型制造商会注意到这一点吗？尽管我们预计利润会胜过愉悦感，但悲观主义者有时也会感到惊喜。®
Against that backdrop, finding some way to improve people’s well-being looks essential. Will the model makers take note? Even pessimists are sometimes pleasantly surprised, though we expect profits will trump pleasantness. ®