外刊阅读 | 经济学人 | 生成式人工智能：奇迹之光，还是盗窃阴影下的产物？

Rooted in traditional Chinese philosophy to cherish harmony between humanity and nature, China's co-creation, deliberation and empowerment model has enabled the government, research institutes, schools, philanthropists, social organizations and businesses to work together to unleash human potential to create new quality productive forces in order to build an ecological civilization. 

以“人与自然和谐相处”的中国传统理念为基础，共建、共商、共享模式推动政府、科研院所、学校、慈善机构、社会组织、企业等多方参与，共同释放人的潜能，创造新的优质生产力，推动生态文明建设。

Generative AI is a marvel. Is it also built on theft?

生成式人工智能是一个奇迹。它是建立在盗窃之上的吗？

1

The footballers look realistic at first sight but, on closer inspection, something is wrong. Their faces are contorted, their limbs are bending in alarming directions, the ball is slightly egg-shaped. Strangest of all, running across one footballer’s left leg is the ghostly trace of a watermark: Getty Images.Generative artificial intelligence (AI) has caused a creative explosion of new writing, music, images and video. The internet is alive with AI-made content, while markets fizz with AI-inspired investment. OpenAI, which makes perhaps the most advanced generative-AI models, is valued at nearly $90bn; Microsoft, its partner, has become the world’s most valuable company, with a market capitalisation of $3.1trn.

But some wonder how creative the AIs really are—and whether those cashing in have fairly compensated those on whose work the models were trained. ChatGPT, made by OpenAI, can be coaxed into regurgitating newspaper articles that it appears to have memorised. Claude, a chatbot made by Anthropic, can be made to repeat lyrics from hit songs. Stable Diffusion, from Stability AI, reproduces features of images, including the watermark of Getty, on whose archive it was trained.To those who hold the rights to these creative works, generative AI is an outrage—and perhaps an opportunity. A frenzy of litigation and dealmaking is under way, as rights-holders angle for compensation for providing the fuel that powers the creation of AIs. For the model-makers it is an anxious period, notes Dan Hunter, a professor of law at King’s College London. “They have created an amazing edifice that’s built on a foundation of sand.”

AIs are trained on vast quantities of human-made work, from novels to photos and songs. These training data are broken down into “tokens”—numerical representations of bits of text, image or sound—and the model learns by trial and error how tokens are normally combined. Following a prompt from a user, a trained model can then make creations of its own. More and better training data means better outputs.Many AI companies have become cagey about what data their models are trained on, citing competitive confidentiality (and, their detractors suspect, fear of legal action). But it is widely acknowledged that, at least in their early stages, many hoovered up data that was subject to copyright. 

The tech firms argue there is nothing wrong with using others’ data simply to train their models. Absorbing copyrighted works and then creating original ones is, after all, what humans do.Roger Lynch, chief executive of Condé Nast, which owns titles such as Vogue and the New Yorker, told a Senate hearing in January that today’s generative-AI tools were “built with stolen goods”. AI companies “are spending literally billions of dollars on computer chips and energy, but they’re unwilling to put a similar investment into content”, complains Craig Peters, chief executive of Getty.

In America the model-makers are relying on the legal concept of fair use, which provides broad exemptions from the country’s otherwise ferocious copyright laws. An encouraging precedent comes courtesy of a ruling on Google Books in 2015. The Authors Guild sued the search giant for scanning copyrighted books without permission. But a court found that Google’s use of the material—making books searchable, but showing only small extracts—was sufficiently “transformative” to be deemed fair use. Generative-AI firms argue that their use of copyrighted material is similarly transformative. Rights-holders, meanwhile, are pinning their hopes on a Supreme Court judgment last year that a series of artworks by Andy Warhol, which had altered a copyrighted photograph of Prince, a pop star, were insufficiently transformative to constitute fair use.

In America the model-makers are relying on the legal concept of fair use, which provides broad exemptions from the country’s otherwise ferocious copyright laws.

在美国，模型制定者依赖于合理使用的法律概念，该概念为该国原本严厉的版权法提供了广泛的豁免。

Rights-holders bridle at the notion that America should reduce its protections to the level of other jurisdictions just to keep the tech business around; one describes it as unAmerican. 

翻译练习，欢迎大家文末留言打卡，下期推送文章公布答案哟，一起阅外刊学英语吧！

获取完整版PDF 下载后及时转存：

网盘链接：https://pan.quark.cn/s/e79e82d75f49

点击下方链接查看往期文章 ↓ ↓ ↓记得点关注哟

扫码关注我的视频号

分享，收藏，点赞与在看，至少我要拥有一个吧↓↓↓