Is the ‘Made by AI’ label pointless? - FT中文网
登录×
电子邮件/用户名
密码
记住我
请输入邮箱和密码进行绑定操作:
请输入手机号码,通过短信验证(目前仅支持中国大陆地区的手机号):
请您阅读我们的用户注册协议隐私权保护政策,点击下方按钮即视为您接受。
人工智能

Is the ‘Made by AI’ label pointless?

The experience of the video game industry suggests transparency about AI use won’t be straightforward
00:00

{"text":[[{"start":null,"text":"

"}],[{"start":8.59,"text":"Is a human writing this column? How can you be sure? As it becomes harder to distinguish people from AI, a wide array of different “solutions” are beginning to emerge, some more dystopian than others. Authors are posting videos of themselves writing their books in real time to prove they’re not using large language models. Sam Altman has come up with a device called The Orb which “verifies you are a unique human” by scanning your eyeballs. And under the EU’s AI Act, certain types of AI-generated content will have to be labelled as such from next year."}],[{"start":53.47,"text":"But the most interesting debate on the topic is taking place in the video games industry, where a conversation is unfolding that many more of us will probably need to have soon."}],[{"start":67.55,"text":"In January 2024, Steam — the most prominent digital storefront for PC games — began to require developers to disclose whether (and how) they used AI tools in the creation of their game. According to Ichiro Lambe, a games industry veteran who has been tracking the trends in AI disclosures, around 1,000 games had disclosed the use of generative AI by April 2024. That number has now mushroomed to more than 11,000, or roughly 9 per cent of the entire Steam library, he told me. According to the disclosures, developers are using AI for a wide array of tasks, from coding assistance and marketing materials to generating visuals, textures, background music and characters’ voices using “text-to-speech” tools."}],[{"start":122.38,"text":"But some in the industry are beginning to object to the whole idea of AI labelling."}],[{"start":128.78,"text":"One argument is that the labels are meaningless because AI use is becoming so commonplace in the development process. Tim Sweeney, the chief executive of video game company Epic Games, which made the wildly popular video game Fortnite, wrote recently on X that AI labelling “makes no sense for game stores, where AI will be involved in nearly all future production”."}],[{"start":158.12,"text":"Another objection is that smaller developers, which might use AI tools in order to compete with the resources of bigger players, could be bombarded with bad reviews from vocal “AI-haters”."}],[{"start":174.86,"text":"A third complaint is that it’s impossible to police whether people are being truthful about their AI use, which turns it into an “honesty tax” on those who disclose."}],[{"start":187.29000000000002,"text":"What to make of these complaints? It’s certainly true that some people don’t care about AI use in games, while others feel very strongly about it. Some of these objectors think AI use is an indicator of low effort or poor quality. Others object to the way LLMs were trained on human-made content without permission. And some think it’s leading to job losses for developers, artists and voice actors."}],[{"start":222.09000000000003,"text":"Polling suggests there is a similar split among the public when it comes to AI involvement in art and music. One survey by Pew found that if Americans learnt a painting they liked was made by AI, 49 per cent would like it less, while a very similar share (48 per cent) said their views wouldn’t change. Interestingly, young people had a more negative reaction to AI: 66 per cent of adults under 30 said they would like a painting less if they found out it was done by AI, compared with 36 per cent of Americans aged 65 and older."}],[{"start":265.09000000000003,"text":"In a world in which some people don’t care whether content was made by AI but others really do, transparency seems in principle like a good solution. Just like people who want to avoid certain ingredients in their food, labelling is useful for those who want to know, and can simply be ignored by those who don’t."}],[{"start":288.23,"text":"The other advantage of transparency is that it creates a market for human creative professionals: if there is a sizeable proportion of customers who want to continue to consume human-made content, they need to be able to identify it, or else both creators and consumers will lose out. Indeed, as AI use becomes more pervasive, it is possible to imagine “Made by Humans” labels for creative products, akin to the Fairtrade scheme."}],[{"start":319.68,"text":"But the experience of the video game industry suggests it won’t be that straightforward."}],[{"start":327.31,"text":"“The tricky thing is that generative AI usage is not necessarily a binary,” Lambe told me. At one extreme, you might have a developer that used Midjourney to create all the game’s images. But at the other, you might have someone who asked ChatGPT for advice about how to resolve a coding error. Do consumers really care equally about both? Where is the line? And what if that line moves over time?"}],[{"start":356.49,"text":"“The quality and the visibility of it, I think, very much matter,” he said. “AI-generated art is often very visibly AI-generated art and easy to pick out, [but] game developers and consumers seem more willing to accept AI-assisted programming.”"}],[{"start":375.93,"text":"Lambe thought the best solution might be to move towards a more detailed taxonomy of AI use in the development process, in order to enable people to figure out — and have a conversation about — what they’re comfortable with and what they want to avoid."}],[{"start":393.72,"text":"Transparency over AI use is a messy human solution that doesn’t lend itself to black-and-white labels. But for as long as a significant proportion of people care about whether creative products are made by humans or machines, it is surely better than giving up altogether. Either that or we could take our chances with The Orb."}],[{"start":425.51000000000005,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1766047673_4657.mp3"}

版权声明:本文版权归FT中文网所有,未经允许任何单位或个人不得转载,复制或以任何其他方式使用本文全部或部分,侵权必究。

特朗普任命美国格陵兰岛特使

杰夫•兰德里的角色凸显了美国总统控制北极岛屿的决心。

福特押注“未来卡车”的电动化如何导致195亿美元减记

F-150 Lightning的经济账算不拢,也折射出全行业对美国电动汽车普及速度的误判。

一周展望:金融市场会在最后关头上演“圣诞行情”吗?

由于担忧人工智能公司的巨额开支,今年的“圣诞行情”迟迟没有到来。

海外买家抢购低估值英国公司,交易金额创下新高

最新数据显示,今年海外竞购者收购英国公司的成交额达到1420亿美元,较2024年增长74%。

科学家发现人类生火比原先认知早35万年

英格兰萨福克郡的一处石器时代遗址显示出人类发展关键阶段的证据。

比尔•克林顿如何成了爱泼斯坦档案的焦点

美国司法部公布的一批文件将聚光灯转向这位前总统。
设置字号×
最小
较小
默认
较大
最大
分享×