内容来源：https://www.theverge.com/ai-artificial-intelligence/974265/fender-ceo-bud-cole-ai-music

内容总结：

芬达吉他CEO争议言论引火烧身 称翻唱和乐队成员如同“模拟AI”

今年五月，芬达吉他首席执行官爱德华·“巴德”·科尔为庆祝Telecaster吉他问世75周年接受科技媒体T3专访，其关于人工智能与音乐的言论起初未引起关注，但近日却在网络上持续发酵，为本已因版权纠纷而陷入公关危机的公司再添一把火。此前，芬达公司向众多手工制琴师发出停止侵权函，声称拥有Stratocaster琴体外观的版权，此举几乎激怒了整个吉他演奏群体。

在引发争议的采访中，科尔将翻唱他人歌曲类比为AI的训练数据，甚至认为乐队成员也是一种“模拟AI”。此番言论令公司形象雪上加霜，多位颇具影响力的吉他类YouTube博主已公开宣称今后不再购买芬达产品。T3主编马特·加拉格尔的报道大多转述科尔的表述，芬达公司目前尚未回应置评请求。

科尔在采访中的核心观点是，音乐领域的AI并非新鲜事物。“我认为自录音音乐诞生以来，AI就一直存在于音乐之中。”在他看来，学习吉他最大的障碍是投入时间，而第二大障碍——写歌——正是AI可以介入的环节。“实际上，我认为翻唱音乐在很长一段时间里就是一种模拟AI，”科尔表示，那些尚未具备创作能力的人可以通过演奏偶像的作品来起步。他回忆自己曾大量聆听REM、U2、史密斯乐团和治愈乐团的作品，“听腻了就想自己弹，于是学会了吉他。”

科尔还认为，初学创作的人可以借助另一种“模拟AI”——他们的乐队伙伴。可能你只有一个副歌或一个即兴段子，鼓手或贝斯手的加入让一首歌完整成形。AI同样可以扮演这个角色。他说：“我认为我们正处于一个临界点，AI能让人们超越老套的翻唱，真正像和乐队一起协作那样去创作。”

科尔的类比试图说明，人类学习几首翻唱歌曲与AI消化海量受版权保护的音乐数据本质相同。他似乎在暗示，通过学弹他人歌曲、内化影响并融会贯通创作新作品，与AI的工作原理别无二致。然而这一观点被普遍认为是严重误判——要么科尔不懂AI，要么他不尊重艺术家。

首先，规模差异巨大。没有任何人类能学完训练一个生成式AI模型所需的全部歌曲——以音乐生成工具Suno为例，其训练数据据称达数百万首。此外，这种说法抹杀了艺术家在创作过程中无数微小决策中蕴含的人性特质——无论是有意还是无意，无论是情感驱动、应对意外灵感还是弥补自身限制，人类做出的艺术抉择都是独一无二的。

这与模型根据提示和数据库生成输出有着本质区别。正如知名吉他YouTuber“武士吉他手”史蒂夫·奥诺特拉所指出的，演奏者的身体条件以及每个人都难免的微小失误，使其无法完美复制他人作品，这种偶然性无法被大语言模型复现。

乐队伙伴的类比同样站不住脚。人类从各自独特的“训练数据”、人生经历和身体技能中汲取养分，这与聊天机器人截然不同。AI不具备你那位大学主修爵士作曲的挑剔贝斯手那样的品味和直觉。如果你对鼓手说他与AI模型毫无区别，他完全有理由感到被冒犯。

采访后段，科尔还表示：“我相信AI将帮助创造一个新世界的吉他手，帮助他们与其他音乐人连接，提高效率，跨越从歌曲创作学生到大师之间的鸿沟。”

但这一说法同样被指荒谬。越来越多的证据表明，依赖AI工具反而会导致技能退化。用AI来建议押韵或痛苦的情感隐喻，与真正练习写歌、培养能力不可同日而语。AI从未经历过失恋之痛，也从未为一段完美的副歌过渡而绞尽脑汁。重复练习才是关键——俗话说，你得写出一百首、一千首甚至一万首烂歌，才能写出第一首好歌。唯有如此，你才能超越陈词滥调，学会在偶然间认出自己的灵感闪光。

中文翻译：

芬达公司首席执行官爱德华·“巴德”·科尔在五月接受T3采访时庆祝了Telecaster吉他75周年，其间他对人工智能和音乐的一些评论起初并未引起太多关注。但这些言论最近开始在网上流传，给这家公司本已熊熊燃烧的公关危机之火又浇了一桶油——此前芬达向制琴师发出停止侵权函，声称对Stratocaster琴型拥有版权，惹怒了几乎整个吉他演奏圈。

芬达首席执行官似乎认为你的乐队成员就是“模拟版人工智能”

巴德·科尔在那次争议性采访中还把学弹翻唱歌曲比作AI训练数据。

一些有影响力的吉他YouTube博主甚至表示，在这起争议之后，他们再也不买芬达的产品了。而科尔重新被翻出来的那段把翻唱歌曲和乐队成员比作某种“模拟版人工智能”的言论，进一步让该公司在网络最活跃粉丝群体中的形象雪上加霜。

T3主编马特·加拉格尔的专题报道大部分是对科尔言论的转述，芬达公司未立即回应置评或澄清请求。以下是采访中的相关要点：

科尔的核心观点是，音乐中的人工智能并不是什么新鲜事。“我认为只要有录音音乐以来，AI就一直存在于音乐之中，”他说。虽然学吉他的最大障碍是花费大量时间练习，但他认为第二个障碍——写歌——才是AI能发挥作用的地方。

“我确实认为翻唱音乐在很长一段时间里就相当于一种模拟版AI，”科尔说。那些还不具备自己写歌能力的人，可以改而演奏他们最喜欢的艺人创作的歌曲。“我听了大量REM、U2、The Smiths和The Cure的歌，到某个时候我厌倦了只是听。我想自己弹，于是我去学了吉他。”

科尔表示，那些刚开始尝试写歌的人，也可以依靠第二种模拟形式的AI：他们的乐队成员。你可能只有一段副歌或一个riff作为起点，但鼓手或贝斯手可以在此基础上加东西，一首歌就这样诞生了。AI也可以扮演这个角色。“我确实认为我们正处于一个临界点，AI能让人们摆脱一成不变的翻唱，真正像和乐队一起创作那样去工作，”科尔说。

科尔试图在人类“学习”几首翻唱歌曲与AI吞入海量受版权保护的音乐数据集之间建立类比。他似乎是在暗示，通过学习弹奏别人的歌曲、内化这些影响、再将其综合成新的东西，你本质上做的就是和AI一样的事情。这是一个非常令人遗憾的误导性观点，在我看来，这说明科尔要么不懂AI，要么不尊重艺术家。

首先，规模很重要。没有人能够学会训练一个普通生成式AI模型所用的所有歌曲——以Suno为例，其训练数据据推测有数百万首。此外，这种说法还抹杀了艺术家在写歌过程中做出数百万个细微决定时所蕴含的与生俱来的人性，无论这些决定是有意还是无意的。无论是出于情感反应、对意外惊喜的回应，还是对自身局限的弥补，人类所做的艺术决策都是独一无二的。

这与模型基于提示词和数据点网络生成本质上是不同的。正如Steve Onotera（更为人熟知的名字是Samurai Guitarist）所指出的，演奏者的身体条件，或者说每个人都容易犯的小错误，使得他们无法完美复制他人的作品。那种偶然的灵感是LLM无法复制的。

乐队成员也是如此。人类从自己独特的“训练数据”、人生经历和身体技能或局限中汲取养分，这与聊天机器人是不同的。AI没有你那位在大学学过爵士作曲的挑剔贝斯手那样的品味和直觉。如果你对鼓手说他和AI模型没什么区别，他们完全有理由感到被冒犯。

在采访的后半部分，科尔说：“我相信AI实际上将帮助创造一个全新的吉他手群体，让他们利用AI与更多音乐人建立联系，提高创作效率，并跨越那道鸿沟，从歌曲创作的学生成长为歌曲创作的大师。”

科尔声称AI将会以某种方式帮助人们“跨越鸿沟”成为创作大师，坦率地说，这种说法也很荒谬。越来越多的证据表明，依赖AI工具实际上会导致技能退化。用AI来建议押韵或关于痛苦的隐喻，并不等同于练习写歌和培养技能。AI从未被抛弃在婚礼圣坛前，也从未为完美的副歌前过渡绞尽脑汁。重复是关键。有句老话说，你需要写100首（或1000首）（或10000首）烂歌，才能写出一首好歌。唯有如此，你才能超越陈词滥调，学会在偶然碰出好东西时认出它来。

英文来源：

Fender CEO Edward “Bud” Cole gave an interview to T3 in May celebrating the 75th anniversary of the Telecaster with comments on AI and music that initially flew under the radar. But it has started making the rounds recently, pouring more fuel on an already raging fire of bad PR following the company pissing off basically the entire guitar-playing community by sending cease-and-desist letters to builders, claiming copyright of the Stratocaster body shape.
Fender’s CEO seems to think your bandmates are just analog AI
Bud Cole also compared learning cover songs to AI training data in a controversial interview.
Some influential guitar YouTubers have even said they’re done buying Fender gear in the wake of the controversy. And Cole’s resurfaced comments comparing cover songs and bandmates to a sort of “analog AI have only sunk the company’s standing further among the loudest of its fans online.
T3 editor-in-chief Mat Gallagher’s feature mostly paraphrases Cole’s statements, and Fender did not immediately respond to a request for comment or clarification. But here are the relevant bits from the interview:
Cole’s philosophy is that AI in music is nothing new. “I think AI has existed in music as long as there’s been recorded music,” he says. While the biggest barrier to playing guitar is the time it takes to learn the instrument, it’s the second barrier, writing songs, where he believes AI plays a part.
“I actually believe cover music has been sort of analog AI for a long time,” says Cole. Those that don’t yet have the skills to write their own songs can play the songs written by their favourite artists instead. “I listened a lot to REM, U2, The Smiths and The Cure, and at some point I got sick of just listening to them. I wanted to play it, so I learned to play guitar.”
Those taking their first steps into writing, according to Cole, can also lean on a second analogue form of AI: their band mates. You might just have a chorus or a riff to start with but then the drummer or bassist can add to it, and a song is born. AI can also play this role. “I actually think that we are in the brink of freeing up people to move beyond the same old covers and to really get into working like they do with their bands,” says Cole.
Cole is trying to draw a comparison between a human “training” on a handful of cover songs and an AI ingesting enormous datasets of copyrighted music. He appears to be suggesting that, by learning to play other people’s songs, internalizing those influences, and then synthesizing them into something new, you are essentially doing the same thing as an AI. This is a woefully misguided take that says to me that Cole either doesn’t understand AI or doesn’t respect artists.
For starters, scale matters. No person could possibly learn all of the songs used to train your average generative AI model, which, in the case of Suno, is suspected to be in the millions. Additionally, it dismisses the inherent humanity of the millions of tiny decisions, conscious or otherwise, that an artist makes during the songwriting process. Whether they’re driven by emotional response, reacting to a happy accident, or compensating for limitations, the artistic decisions made by a human are unique to them.
This is fundamentally different from a model spitting out something based on a prompt and a network of data points. As Steve Onotera, better known as Samurai Guitarist, points out, a player’s physicality, or the tiny errors that every human is prone to, prevent them from replicating someone else’s work perfectly. That kind of serendipity can’t be replicated by an LLM.
The same is true of bandmates. Humans who pull from their own unique sets of “training data,” life experiences, and physical skills or limitations are not the same as a chatbot. An AI doesn’t have taste or instincts in the way that your picky bassist who studied jazz composition in college does. If you told your drummer they were no different from an AI model, they’d rightfully be insulted.
Later in the interview, Cole says: “I believe that AI is actually going to help create a whole new world of guitar players that use it. To help connect with other musicians, to be more productive. And across the chasm into becoming a student of songwriting to a master of songwriting.”
Cole’s assertion that AI will somehow help people “across the chasm” to becoming master songwriters is also, frankly, ridiculous. Evidence is mounting that relying on AI tools is actually leading to deskilling. Using an AI to suggest rhymes or metaphors for pain isn’t the same as practicing songwriting and developing skills. The AI has never been left at the altar or sweated over the perfect pre-chorus transition. Repetition is the key. The adage is that you need to write 100 (or 1,000) (or 10,000) bad songs before you write one good one. That’s how you grow beyond tired tropes and learn to recognize when you’ve stumbled into something good.