关于 AI 时代的一些思考
9 月 23 日,Sam 发了一篇名为 《智能时代》(The Intelligence Age[1])的文章,从远古聊到未来,写的挺长,我将中文对照翻译放在了文章最后一部分(建议大家细细阅读)。关于 AI,我之前也写过一些思考(AI 浪潮下的一些浅思),后来又有了新收获就想补充一下。
个人思考
从 2022 年 11 月 ChatGPT 发布到现在快两年时间了,如果让你停下来想想,你认为 AI 对你最大的改变或影响是什么?可以评论区留言,我想先来谈谈自己的感受:
ChatGPT 出现的,让一些有心人发现了“巨大商机”,镰刀、媒体为之疯狂...
镰刀:人人都是 AI 专家,写书的、卖课的好不热闹。
媒体:各种“震惊体”无时无刻不在为大家制造着焦虑。
凡是打着旗号教你使用 AI 如何赚钱的,很有可能是在骗你的钱。从个人了解来看,目前在 AI 领域能真正赚钱的并不多:卖显卡(英伟达)、写教程卖课的、卖服务的(API 套壳)、写文搞流量的(赚广告费)等,但这些基本都不适合普通人。
几乎没有适合普通人可以赚钱的路子,唯一能让自己不亏的就是利用 AI 给日常工作或生活提效。
就目前而言,AI 或许很强,但能够深入理解业务,做到真正提效的并不多。AI 可以解决 80% 的问题,也不代表你只需解决剩余 20% 即可。问题往往是交错穿插出现的,当 AI 无法决策时,则需我们进行引导,这也是现阶段我不看好 Agent 智能体的原因(想一次性给出完美答案)。
作为普通人,我们接触最多的 AI 无非以下几类:
文字生成:ChatGPT[2]、Claude[3]、Gemini[4]、Microsoft Copilot[5]、Perplexity[6]、HuggingChat[7] 、Coze[8]、Le Chat - Mistral AI[9]、Pi[10]、YOU[11]、智谱清言[12]、通义千问[13]、豆包[14]、腾讯元宝[15]、Kimi[16]、秘塔[17] 、文心一言[18] 、讯飞星火[19] 等。
图像生成:Stable Diffusion[20](两款主流应用:Stable Diffusion web UI[21] 和 ComfyUI[22])、Midjourney[23]、DALL-E[24](已内置在 ChatGPT 中)、Adobe Firefly[25]、Flux AI[26] 等。
视频生成:Runway[27]、HeyGen[28]、Synthesia[29]、KLING AI[30] 等。
音频生成:Whisper[31]、Suno[32]、ElevenLabs[33]、NotebookLM[34](可以根据文字生成为播客,但能力远非于此) 等。
编程助手:GitHub Copilot[35]、Cursor[36]、V0.dev[37]、Warp[38](下一代命令终端,支持 AI 对话)等。
混合生成:DeepAI[39] 支持文字、图像、视频、音频生成。
聚合类 AI:Poe[40] 是比较特别的存在,它整合了目前市面上所有主流 AI,提供文字、图像生成。
快速推理:Groq[41] 提供 Llama 3.1 8B、Llama 3.1 70B、Llama 3 8B、Llama 3 70B、Mixtral 8x7B、Gemma 7B、Gemma 2 9B、Whisper Large V3 等开源模型快速推理云服务支持。
...
要想学习了解更多 AI 相关的东西,一手信息很重要(别人嚼过的如果是错的就很尴尬)。推荐几个社区:
X[42](原名 Twitter):这里几乎汇聚着整个 AI 领域的专家学者,每天都会有高频的互动产生,很适合用来了解前沿资讯。
HuggingFace 日报[43]:AI 最新的一些进展或论文都可以在这里看到,它是一个很重要的专业资讯集合。
GitHub Trending[44]:GitHub 每天都会有大量的新开源项目产生,但是人们很难发现它们。Trending 榜可以有效筛选出当下热门项目,其中不乏 AI 类项目。
arxiv[45]:一个免费开放获取的学术文章档案,包含近 240 万篇未经过同行评审的研究文献,涵盖多个科学领域(物理、数学、计算机科学、生物学、金融、统计学、电气工程与系统科学以及经济学等领域)。许多 AI 相关论文都可以在这里找到,因论文多以 PDF 文件形式出现不利于阅读,ar5iv[46] 是一个将论文转换为响应式 HTML 5 网页的工具,用法简单,只需将链接中
arxiv
的x
改为5
即可(https://arxiv.org/abs/1910.06709 → https://ar5iv.org/abs/1910.06709)。提示词未来或许会不重要,但现阶段依然有着重大价值,多学点没坏处(每天都在用 AI,可能连基本的 CoT 是什么都不知道)。推荐几个 Prompt 工程学习:
Prompt Engineering Guide[47]
OpenAI Cookbook[48]
Anthropic Prompt engineering[49]
...
关于编程
前段时间 Cursor 大火,有人曾我问对 Cursor 之类的 AI 代码编辑器有什么看法以及如何学习编程,我简单整理了一下当时的回答。
Cursor 我没有深度体验。我用 GitHub Copilot 比较多,但大部分场景也限于代码补齐场景(目前 ChatGPT 使用较多,但也需要自己拆解细化问题,不断对其进行纠正才能达到比较理想的效果)。如何设计功能、组织代码、选择适合的依赖等等,可能都是 AI editor (AI 自主编程)需要考虑的问题。否则,人的经验很难被取代。
总感觉“AI 自主编程”对复杂项目来说有点鸡肋。正常编程场景:一个项目在早期时,代码基本都是平铺式。随着需求的迭代,项目复杂度急剧上升,老代码需要抽象或分离(架构开始产生变化)。这里应该让 AI 自我调整还是人为去调整架构?
所以灵魂两问:当项目复杂度变化导致 AI 无法覆盖所有场景时,让人去理解 AI 生成的代码会不会崩溃?技术迭代,新框架或技术应该让人还是 AI 去学习?
我想 AI 的出现或许只是加快了人类学习的脚步,并不是为了让人停止学习。换句话说,AI 编程的本质其实是让小白可以用一种更加友好的方式开始接触编程。
关于如何学习编程,我觉得没啥可分享的技巧。只要自己有基本的学习能力,会分析抽象问题,编程(简单粗暴看文档,跟着写)就是一个水到渠成的事情。代码或工具也只是将解决问题的逻辑进行具象化,并没有什么神奇之处。另外,兴趣往往更容易推着你去了解学习一些未知新事物。
所以学编程还是应以解决问题为出发点。举个简单例子:我想租房,要找一个最具性价比的房子。这时你就可以写个简单的爬虫来做数据分析,或者你想对文件进行批量化读写操作,也可以写个程序脚本来解决。
对不知从什么项目入门的人,我都推荐搭建一个属于自己的 blog,因为它是一个可以伴随自己整个编程生涯的东西,可以不断对其进行功能迭代。例如:通过 GitHub 来搭建一个开源 blog,然后不断为其添加新功能。这里面你可以学到很多东西(如 github action、git、markdown、github api,域名绑定等)。
编程的本质:以问题为导向去了解学习所需技术或工具,通过观察分析抽象来建立标准化流程以解决重复性问题。
最后补充个看文档技巧:看文档并不一定非要记住它,脑子里有印象,遇到了可以快速检索到 API 使用方式就足够了。很多人看文档之所以会头大,那是因为在试图一次性掌握所有 API 细节。这既不现实,也没必要,而且版本迭代也会随时更新 API(时不时就需要翻阅一下)。技术文档为啥重要?其实它就像是产品使用说明书,让你可以了解一个框架或库的能力,如果都不知道它有哪些能力,写代码也就无从谈起了。
The Intelligence Age
智能时代 (2024.09.23)
In the next couple of decades, we will be able to do things that would have seemed like magic to our grandparents.
在未来几十年里,我们将能够做出一些在祖辈眼中看似魔法的事情。
This phenomenon is not new, but it will be newly accelerated. People have become dramatically more capable over time; we can already accomplish things now that our predecessors would have believed to be impossible.
这一现象并不新鲜,但将被新加速。随着时间的推移,人们的能力显著提升;如今我们能做的事情,过去的先辈们认为是不可能的。
We are more capable not because of genetic change, but because we benefit from the infrastructure of society being way smarter and more capable than any one of us; in an important sense, society itself is a form of advanced intelligence. Our grandparents – and the generations that came before them – built and achieved great things. They contributed to the scaffolding of human progress that we all benefit from. AI will give people tools to solve hard problems and help us add new struts to that scaffolding that we couldn’t have figured out on our own. The story of progress will continue, and our children will be able to do things we can’t.
我们的能力提升不是因为基因的改变,而是得益于社会基础设施变得比我们任何一个人都更聪明、更有能力;从某种重要意义上说,社会本身就是一种高级智能。我们的祖辈以及之前的世代,创造了伟大的成就。他们为人类进步的框架做出了贡献,而我们都从中受益。人工智能将为人们提供解决难题的工具,帮助我们添加新的支撑,这些支撑我们无法独自想出。进步的故事将继续,我们的孩子将能够做我们无法做到的事情。
It won’t happen all at once, but we’ll soon be able to work with AI that helps us accomplish much more than we ever could without AI; eventually we can each have a personal AI team, full of virtual experts in different areas, working together to create almost anything we can imagine. Our children will have virtual tutors who can provide personalized instruction in any subject, in any language, and at whatever pace they need. We can imagine similar ideas for better healthcare, the ability to create any kind of software someone can imagine, and much more.
这不会一下子发生,但我们很快就能与人工智能合作,帮助我们完成更多事情。最终,我们每个人都可以拥有一个个人AI团队,由不同领域的虚拟专家协同工作,创造几乎任何我们能想象的事物。我们的孩子将有虚拟导师,可以在任何科目、任何语言中,按照他们需要的节奏提供个性化指导。我们可以想象类似的理念,用于更好的医疗保健、创造任何人所能想象的软件等。
With these new abilities, we can have shared prosperity to a degree that seems unimaginable today; in the future, everyone’s lives can be better than anyone’s life is now. Prosperity alone doesn’t necessarily make people happy – there are plenty of miserable rich people – but it would meaningfully improve the lives of people around the world.
凭借这些新能力,我们可以实现今天看来难以想象的共享繁荣;未来,每个人的生活都可以比现在的任何人的生活都要好。单靠繁荣并不一定让人快乐——有很多富人也很痛苦——但它确实会在全球范围内显著改善人们的生活。
Here is one narrow way to look at human history: after thousands of years of compounding scientific discovery and technological progress, we have figured out how to melt sand, add some impurities, arrange it with astonishing precision at extraordinarily tiny scale into computer chips, run energy through it, and end up with systems capable of creating increasingly capable artificial intelligence.
从一个狭隘的角度看人类历史:经过数千年的科学发现和技术进步,我们已经弄清楚如何熔化沙子,添加一些杂质,以惊人的精度在极小的尺度上排列成计算机芯片,通电后,最终形成能够创造越来越强大人工智能的系统。
This may turn out to be the most consequential fact about all of history so far. It is possible that we will have superintelligence in a few thousand days (!); it may take longer, but I’m confident we’ll get there.
这可能是迄今为止所有历史中最重要的事实。我们可能在几千天内就会拥有超级智能(!);虽然可能需要更长时间,但我相信我们会实现这一目标。
How did we get to the doorstep of the next leap in prosperity?
我们如何迈入下一次繁荣的大门?
In three words: deep learning worked.
用三个词来说:深度学习奏效。
In 15 words: deep learning worked, got predictably better with scale, and we dedicated increasing resources to it.
用十五个词来说:深度学习奏效,随着规模的增加变得可预见性地更好,我们投入了更多资源。
That’s really it; humanity discovered an algorithm that could really, truly learn any distribution of data (or really, the underlying “rules” that produce any distribution of data). To a shocking degree of precision, the more compute and data available, the better it gets at helping people solve hard problems. I find that no matter how much time I spend thinking about this, I can never really internalize how consequential it is.
这就是关键;人类发现了一种算法,可以真正学习任何数据分布(或说,生成任何数据分布的底层“规则”)。出乎意料的是,计算能力和数据越多,它在帮助人们解决难题方面就变得越好。我发现,无论我花多少时间思考这一点,我永远无法真正内化其重要性。
There are a lot of details we still have to figure out, but it’s a mistake to get distracted by any particular challenge. Deep learning works, and we will solve the remaining problems. We can say a lot of things about what may happen next, but the main one is that AI is going to get better with scale, and that will lead to meaningful improvements to the lives of people around the world.
我们仍有许多细节需要搞清楚,但转移注意力到任何特定挑战上都是错误的。深度学习有效,我们将解决剩下的问题。我们可以说很多关于未来可能发生的事情,但主要的一点是,人工智能将随着规模的增加变得更好,这将对全球人们的生活产生重大改善。
AI models will soon serve as autonomous personal assistants who carry out specific tasks on our behalf like coordinating medical care on your behalf. At some point further down the road, AI systems are going to get so good that they help us make better next-generation systems and make scientific progress across the board.
人工智能模型很快将作为自主个人助手,代表我们执行特定任务,比如协调医疗照护。未来某个时刻,人工智能系统将变得如此优秀,以至于帮助我们构建更好的下一代系统,并在各个领域推动科学进步。
Technology brought us from the Stone Age to the Agricultural Age and then to the Industrial Age. From here, the path to the Intelligence Age is paved with compute, energy, and human will.
技术使我们从石器时代走到农业时代,再到工业时代。从这里开始,通往智能时代的道路由算力、能源和人类意志铺就。
If we want to put AI into the hands of as many people as possible, we need to drive down the cost of compute and make it abundant (which requires lots of energy and chips). If we don’t build enough infrastructure, AI will be a very limited resource that wars get fought over and that becomes mostly a tool for rich people.
如果我们想把人工智能交到尽可能多的人手中,我们需要降低计算成本,使其变得丰富(这需要大量的能源和芯片)。如果我们不构建足够的基础设施,人工智能将成为一种非常有限的资源,战争将为此而起,它主要成为富人的工具。
We need to act wisely but with conviction. The dawn of the Intelligence Age is a momentous development with very complex and extremely high-stakes challenges. It will not be an entirely positive story, but the upside is so tremendous that we owe it to ourselves, and the future, to figure out how to navigate the risks in front of us.
我们需要谨慎行事,但要坚定信念。智能时代的曙光是一个重大进展,面临着复杂而高风险的挑战。这并不完全是一个积极的故事,但潜在的好处如此巨大,我们有责任为自己和未来找到应对眼前风险的方法。
I believe the future is going to be so bright that no one can do it justice by trying to write about it now; a defining characteristic of the Intelligence Age will be massive prosperity.
我相信未来将会如此美好,以至于现在写出来也无法公正地表达;智能时代的一个决定性特征将是大规模繁荣。
Although it will happen incrementally, astounding triumphs – fixing the climate, establishing a space colony, and the discovery of all of physics – will eventually become commonplace. With nearly-limitless intelligence and abundant energy – the ability to generate great ideas, and the ability to make them happen – we can do quite a lot.
尽管这会逐步发生,但令人惊叹的胜利——修复气候、建立太空殖民地、发现所有物理法则——最终将变得司空见惯。凭借几乎无限的智能和丰富的能源——产生伟大思想的能力,以及使其成为现实的能力——我们可以做到很多事情。
As we have seen with other technologies, there will also be downsides, and we need to start working now to maximize AI’s benefits while minimizing its harms. As one example, we expect that this technology can cause a significant change in labor markets (good and bad) in the coming years, but most jobs will change more slowly than most people think, and I have no fear that we’ll run out of things to do (even if they don’t look like “real jobs” to us today). People have an innate desire to create and to be useful to each other, and AI will allow us to amplify our own abilities like never before. As a society, we will be back in an expanding world, and we can again focus on playing positive-sum games.
正如我们在其他技术中看到的,可能会有负面影响,我们需要现在就开始努力,最大化人工智能的好处,同时最小化其危害。举个例子,我们预计这项技术将在未来几年对劳动市场造成重大变化(好与坏),但大多数工作的变化速度将人们想象的要慢,而且我不担心我们会用完要做的事情(即使这些事情在今天看起来不像“真正的工作”)。人们有一种与生俱来的创造欲望和对彼此有用的渴望,人工智能将使我们前所未有地放大自身的能力。作为一个社会,我们将重新回到一个扩展的世界,再次专注于积极的和谐互动。
Many of the jobs we do today would have looked like trifling wastes of time to people a few hundred years ago, but nobody is looking back at the past, wishing they were a lamplighter. If a lamplighter could see the world today, he would think the prosperity all around him was unimaginable. And if we could fast-forward a hundred years from today, the prosperity all around us would feel just as unimaginable.
我们今天所做的许多工作在几百年前看起来可能是微不足道的浪费时间,但没有人会回首过去,渴望成为一名点灯人。如果一名点灯人能看到今天的世界,他会认为周围的繁荣是难以想象的。如果我们能够将时间快进到一百年后的今天,周围的繁荣也会显得同样不可思议。
References
The Intelligence Age: https://ia.samaltman.com
[2]ChatGPT: https://chatgpt.com
[3]Claude: https://claude.ai
[4]Gemini: https://gemini.google.com
[5]Microsoft Copilot: https://copilot.microsoft.com
[6]Perplexity: https://www.perplexity.ai
[7]HuggingChat: https://huggingface.co/chat
[8]Coze: https://www.coze.com
[9]Le Chat - Mistral AI: https://chat.mistral.ai
[10]Pi: https://pi.ai
[11]YOU: https://you.com
[12]智谱清言: https://chatglm.cn
[13]通义千问: https://tongyi.aliyun.com/qianwen
[14]豆包: https://www.doubao.com
[15]腾讯元宝: https://yuanbao.tencent.com
[16]Kimi: https://kimi.moonshot.cn
[17]秘塔: https://metaso.cn
[18]文心一言: https://yiyan.baidu.com
[19]讯飞星火: https://xinghuo.xfyun.cn
[20]Stable Diffusion: https://stability.ai
[21]Stable Diffusion web UI: https://github.com/AUTOMATIC1111/stable-diffusion-webui
[22]ComfyUI: https://github.com/comfyanonymous/ComfyUI
[23]Midjourney: https://www.midjourney.com
[24]DALL-E: https://openai.com/index/dall-e-2
[25]Adobe Firefly: https://www.adobe.com/products/firefly.html
[26]Flux AI: https://flux1.ai
[27]Runway: https://runwayml.com
[28]HeyGen: https://www.heygen.com
[29]Synthesia: https://www.synthesia.io
[30]KLING AI: https://klingai.com
[31]Whisper: https://openai.com/index/whisper
[32]Suno: https://suno.com
[33]ElevenLabs: https://elevenlabs.io
[34]NotebookLM: https://notebooklm.google
[35]GitHub Copilot: https://github.com/features/copilot
[36]Cursor: https://www.cursor.com
[37]V0.dev: https://v0.dev
[38]Warp: https://www.warp.dev
[39]DeepAI: https://deepai.org
[40]Poe: https://poe.com
[41]Groq: https://groq.com
[42]X: https://x.com
[43]HuggingFace 日报: https://huggingface.co/papers
[44]GitHub Trending: https://github.com/trending
[45]arxiv: https://arxiv.org
[46]ar5iv: https://ar5iv.labs.arxiv.org
[47]Prompt Engineering Guide: https://www.promptingguide.ai
[48]OpenAI Cookbook: https://cookbook.openai.com
[49]Anthropic Prompt engineering: https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering