AI开发者能从查尔斯·布考斯基身上学到什么?
What AI developers could learn from Charles Bukowski?

原始链接: https://galjot.si/what-ai-developers-could-learn-from-charles-bukowski

借用查尔斯·布考斯基(Charles Bukowski)的智慧,作者批判了现代科技行业对无限产出的痴迷。布考斯基深知创造力需要耐心;当面对写作障碍时,他选择退避而非强行产出次品,他宁愿舍弃平庸之作也不愿将其发表。 作者将这种深思熟虑与当今的软件开发人员进行了对比。尽管人工智能工具现在能秒速完成过去需要数周才能做完的工作,但开发者并没有利用这种效率来换取闲暇时间。相反,他们陷入了“996”奋斗文化的怪圈,痴迷于提交成千上万个合并请求,而这些请求往往无法解决真正的实际问题。 文章提出了一个挑战:为什么我们是在更努力地工作,而不是更聪明地工作?引用尼尔·德格拉塞·泰森(Neil deGrasse Tyson)的观点——人工智能本应让我们有更多时间在海滩上放松——作者建议科技从业者应停止将“数量”置于“价值”之上。通过放慢脚步,拒绝“粗制滥造”,并专注于有意义的人际连接,我们或许能产出更高质量的作品,并由此夺回技术本应赋予我们的自由。

所提供的 Hacker News 讨论反映了人们对人工智能发展与文学之间(特别是查尔斯·布考斯基相关议题)交叉点的怀疑态度。 评论者强调了一种日益增长的反人工智能情绪,其根源在于人们认为该技术本质上是“掠夺性”的——即仅将创意作品视为股东盈利的数据点,而非将其作为艺术来欣赏。一位用户质疑了原文的前提,认为如果人工智能已经对布考斯基的作品进行了训练,那么除了模型已经摄入的内容外,开发者几乎没什么可“学习”的了。另一位参与者则通过分享一则关于诗歌教授对布考斯基尖刻性格发出严厉警告的轶事,转变了讨论基调。 总体而言,该话题将人工智能与创意产出的关系界定为一种交易性和非人性的关系,并因其将复杂的人类表达降格为机器学习的燃料而受到批评。
相关文章

原文
Published:
Charles Bukowski holding a beer and smoking
(C) Ulvis Alberts (I think! – correct me if I'm wrong)

Allow me to jump straight in:

now I have become spoiled,
I walked into here tonight expecting
more luck
but the night has been slow.
and rightfully so—
occurrence must precede action,
the tank must refill.

/…/

when the lulls arrive there is
nothing to do but
wait,
do other things.

/…/

there are other things to do.
like now I am going downstairs
to see my wife
and my 6 cats

/…/

it will be all right.
I’m sure it
will.

they might even remember me.

– an excerpt from the "block" by Charles Bukowski.

Charles is describing a writers block. Then he's being wise and self-conscious about it and instead of forcing himself to write more slop (or ill words in his world) he stands up, moves away from his computer notebook and goes downstairs to his wife and 6 cats. He knows this block is temporal and things will – presumably soon – be all right. Actually, he's sure of that.

So why do we, developers supercharged with AI tools, have a sudden urge to spend more time building software than ever? Silicon valley, since its inception, has been obsessed with working crazy hours, or at least that's what successful folks in their autobiographies told us. Charles knew that more output does not equal quality output.

Even before the AI, he was okay to throw away work that didn't cut the mustard 👇🏻

this is great, I just wrote two
poems I didn't like.
there is a trash can on this
computer.
I just moved the poems
over
and dropped them into
the trash can.
/…/
it's always better
to reject yourself before
the editors do.
/…/

– an excerpt from "the trash can" by Charles Bukowski.

In a recent discussion between Jaron Lanier (that geeky dreadlocks scientist who also makes music) and Neil deGrasse Tyson (that geeky astrophysicist who wrote a book on that subject for normal people that I still didn't understand), Neil said that he has written around 50,000 lines of code – and spent awful lot of time doing so – which AI models today produce in seconds. Then he said:

If I had access [to AI] back then, I'd probably just spend more time on the beach.

Preach on! But… how come we don't see this happening? Why aren't tech folks sipping tequilas on the beach?

Why are we instead witnessing a 996 phenomenon. Why is there a sudden urge to ship 4,500 PRs per day? How many of those PRs have improved anything for our users (to those lucky enough to have users) or solve "a" problem actual people – not personas – verbally said they have?

Would we be more creative if we wandered off to listen to people about ideas we're trying to build?

But maybe the fact that Bukowski turned his inability to write into a poem about his inability to write, and wrote "the trash can" about poems he deleted…
Maybe that's enough to give us all a FOMO, so I'm off vibe-coding my next zero-time Pulitzer Prize winning startup.

PS: If you enjoyed this read, here's part 2: What AI developers could learn from Ernest Hemingway?


If you enjoyed the "block" or "the trash can", consider buying Betting on the Muse, a collection of poetry and short stories by Charles Bukowski.

联系我们 contact @ memedata.com