红迪热议：AI...是泡沫吗？“与2000年初的互联网泡沫有些相似！”“总会出现新的炒作热潮或泡沫”，“AI取得重大突破,泡沫将被消化掉.”
AI... is it a bubble?
译文简介
网友：在我看来，这与2000年初的互联网泡沫有些相似。一些知名人士支持AI，加上媒体对AI及其潜力的持续报道。这并不意味着AI最终不会实现其承诺，就像互联网最终发展成为今天的样子一样......
正文翻译
Discussion
Everyone knows about AI at this point and its nearly impossible to find any economic or business news coverage that isn't talking about it. I completely understand how influential AI can be over the coming decades and I'm not doubting the practicality that it will increasingly have in our society, but do these tech stocks that have been fueling this recent rally really have enough substance to be able to back their current stock prices. To me it seems similar to the internet bubble in the early 2000s. A couple of prominent figures backing AI along with constant media coverage of the subject and its potential. This doesn't mean that AI won't eventually live up to the billing, the same way that the internet eventually grew into what it is today. Most who are heavy into AI seem like those who are seeking to get rich quick, which is of course their prerogative, it just seems like a dangerous combination to me. To me it seems somewhat speculative and I'm curious what others think about the subject.
目前，几乎所有人都了解人工智能（AI），并且在经济或商业新闻报道中很难找到不涉及AI的话题。我完全理解在未来几十年内AI可能产生的巨大影响力，也不怀疑它将在我们的社会中发挥越来越大的实用性作用。但是，推动近期股市上涨的这些科技股真的有足够的实质内容来支撑它们当前的股价吗？
在我看来，这与2000年初的互联网泡沫有些相似。一些知名人士支持AI，加上媒体对AI及其潜力的持续报道。这并不意味着AI最终不会实现其承诺，就像互联网最终发展成为今天的样子一样。
大多数热衷于AI的人似乎都在寻求快速致富，这是他们的权利，但在我看来，这种组合似乎有些危险。在我看来，投资AI领域似乎带有一定的投机性，我很想知道其他人对此有何看法。
评论翻译
很赞 ( 3 )
收藏
No real way to tell without a crystal ball, but as a business literate programmer and systems analyst.... to a degree.
Most of the startups are garbage. They'll get some hype and fizzle out over time. It's a modern day gold rush and the majority of the gold is garbage.
However the techniques that are being developed are irreplaceable and can achieve things that were literally impossible to program even 10 years ago. Techniques like turning photocopies into parseable written text was a pipe dream. Language comprehension, large scale pattern recognition, that sort of thing is absolutely gold. These technologies will gradually hook in with automation tools to meaningfully reduce man-hour requirements in office like roles.
For examples of some of the recent uses, recently in material science it was used to attempt to find hypothetical materials that can outperform modern materials(and "jumped the field ahead 600 years"), or the recent use to find a new antibiotic to fight MRSA, these are absolutely insane benefit wise and the first person to develop AI to produce inventions in any category reaps huge rewards.
The techniques that have already been developed are hypothetically extremely valuable in a wide variety of fields, it's good at time series forecasting for businesses, connecting patterns in data that didn't exist, outputting weird nonsense like using dead actors in movies, etc...
But these tools will become common place and things will settle down and their value will fall to that of a normal tool I believe. My assessment is that the thing worth investing in now is "the cloud". AI is so expensive to train that almost everyone will be outsourcing their computational needs for a long time and the companies providing the computational power are in an insanely strong position to capitalize on it.
This is your obvious Nvidia, AMD, google, microsoft, amazon ABCDEFG tech stack. I don't think you can go wrong buying stock in them, properly diversified. Nvidia is the obvious example and I do not believe they're truly in a bubble compared to the revenue growth we've seen quarter over quarter as AI quickly outpaced their entire classic revenue streams and is beating forecasts constantly. Pretty much every major cloud provider is using them, and they are just starting to try to pivot to providing similar services. Will nvidia collapse in price a lil? Maybe, probably even. But I think it'll keep going up in general for several years, barring massive recession(which seems pretty likely too lol)
Riskier options... any kind of technology that makes AI easier to use or cheaper could fly up in value more dramatically. AI is still a pain in the ass to learn or code for inexperienced developers as the tools are generally written by people with less industry experience in libraries for the masses.
Classification based AI is amazing, but Generative AI in my opinion is a massive bubble. Useful, but more of a parlour trick that I don't believe is possible to ever quite be "perfect". The inherent hallucinations and semi-randomness in the generation means you can never actually trust the output, so it prevents itself from being used in any mass business automation where "pretty good" means losing a lot of money. The obvious thing is it can save time in producing art or movies or music, but I don't think it has significant value beyond the last revolutionary CGI technology.
Just my two cents.
作为一个具备商业素养的程序员和系统分析师，虽然无法通过水晶球准确预测未来，但我可以基于现有的理解分享一些看法。
当前大部分初创公司涉足AI领域，但其中很多可能只是炒作一时，最终会逐渐消失。这就好比现代版的淘金热，大多数“黄金”其实并不成色。然而，在这一过程中开发出的技术是无可替代的，它们能够实现十年前几乎无法编程的任务，比如将复印件转化为可解析的文字，语言理解，大规模模式识别等，这些技术无疑是真正的“黄金”。随着这些技术与自动化工具的逐步结合，办公室类角色的工作时间需求将得到有意义的减少。
近期在材料科学、药物研发等领域，AI已展现出惊人的应用价值。例如，利用AI尝试寻找超越现有材料性能的假设性材料，使该领域的研究进度提前了数百年；或是发现对抗MRSA的新抗生素，这些都是巨大的突破。首个成功开发出能在任何类别中产生发明的AI的企业将收获巨大回报。
目前已经开发出的技术在众多领域都具有极大的潜在价值，如用于商业的时间序列预测、挖掘数据中未曾存在的模式，以及生成诸如让已故演员出现在电影中的奇特内容等。然而，随着时间推移，这些工具将成为常态，其价值将趋于平稳并回归到普通工具水平。我认为目前值得投资的是“云计算”。由于训练AI的成本极高，预计在未来很长一段时间内，几乎所有人都会选择外包计算需求，而提供计算能力的公司将处于极其有利的地位来利用这一机遇。
显而易见的投资选择包括英伟达（Nvidia）、AMD、谷歌（Google）、微软（Microsoft）、亚马逊（Amazon）等大型科技公司的股票。通过适当分散投资，我认为投资这些公司不会犯大错。以英伟达为例，尽管其股价可能会有所波动，但从季度同比增长来看，AI业务已经迅速超过其传统收入来源，并且持续超出预期。几乎所有主要云服务提供商都在使用其产品，而英伟达也开始转向提供类似的服务。即使出现大幅衰退（这也似乎是很有可能的情况），但在未来几年内，我认为英伟达股价总体上仍会上涨。
风险较高但可能带来更大收益的投资选项，是那些能让AI更易于使用或降低成本的技术。对于经验不足的开发者而言，AI的学习和编码仍然是一项挑战，因为目前的工具通常由行业经验较少的人为大众开发的库编写而成。
基于分类的AI非常出色，但在我看来，生成式AI可能存在一个大的泡沫。生成式AI虽有用，但某种程度上更像是一个吸引眼球的小把戏，我认为它很难做到“完美”。生成过程中的内在幻觉和半随机性意味着你无法完全信任输出结果，这就限制了它在需要高度精确的商业自动化场景中的应用，因为在这些场景中，“相当好”的结果也可能导致大量金钱损失。生成式AI明显的优势在于节省艺术创作、电影制作或音乐创作等方面的时间，但在超越上一次革命性的CGI技术之外，我不认为它拥有显著的价值。
以上仅为个人观点。
Great insight especially that sentence about the difference between AI for classification and gen AI. I think the main issue is everyone is competing for a piece of the pie and margins will suffer, but that will take a while. The catalyst I guess will be a good CUDA competitor.
很好的见解，特别是关于分类AI与生成式AI之间差异的那个论断。我认为目前的主要问题在于大家都在争夺市场份额这块“蛋糕”，而这将会导致利润空间受到压缩，不过这一过程可能会持续一段时间。我猜测，推动这一变革的催化剂很可能是出现一款能够与CUDA相抗衡的优秀竞争产品。
I think that's a great point and I 100% agree with you!
AMD and Apple are both currently starting to try to flex at Nvidia that their technology is competitive/better, but right now the big thing isn't technological superiority, it's infrastructure at scale, and Nvidia is the only game in town. But when the other companies do get a legitimate counter offer on the table I think you're right on the money, it'll reign nvidia's growth and margins in pretty hard and that's probably a great time to start backing off on investments as it becomes a more stable/competitive market.
I wouldn't be surprised to see AMD grow pretty majorly if they do get proper rent to use infrastructure in place though, just through getting a piece of that massive pie.
我认为这是一个非常好的观点，并且我100%赞同你的看法！
AMD和苹果目前都在试图向英伟达展示，他们的技术同样具有竞争力甚至更优，但当前的关键问题并非技术上的绝对优势，而是规模化基础设施的建设，而在这方面英伟达是唯一的主导者。然而，一旦其他公司确实提出了有竞争力的对抗方案，我认为正如你所说，这将对英伟达的增长速度和利润空间形成相当大的制约，那时可能是开始减少相关投资的好时机，因为市场将变得更加稳定且竞争激烈。
不过，如果AMD能够成功建立起相应的基础设施租赁使用模式，我也不会感到惊讶，仅仅是通过分享这块巨大蛋糕的一小部分，他们就可能实现大幅度增长。
That was told about Tesla too
关于特斯拉（Tesla）也曾有类似的说法。
did they not lower prices and will now have to compete against BYD?
他们不是降价了吗，现在还要和比亚迪竞争吗？
most of the startups in AI field are fragile. Upcoming AI integration into the OSes will obliterate most of the startups and the app ecosystem to some extent.
AI领域的大多数初创企业都较为脆弱。即将到来的AI与操作系统（OSes）的深度融合将在一定程度上摧毁大部分初创企业和应用程序生态系统。
A single caution here that when it will get obvious that AI (actually machine learning based on past knowledge and statistical methods) is not so profitable it will repeat fate of any bubble - wins first dropper ;)
这里有一个需要注意的警告，那就是当人们明显意识到AI（实际上是基于过往知识和统计方法的机器学习）并非如预期般盈利时，它将会重蹈任何泡沫破灭的命运——率先退出者将获胜。
Not sure that it's a "bubble" yet, but tech companies are 100% overselling its current abilities. In my opinion, tech companies are using AI as a cover to continue shedding pandemic overgrowth. They get to keep cutting staff while saying they're pouring resources into the fastest-growing sector to keep investment flowing.
目前尚不确定是否已形成“泡沫”，但科技公司无疑在过分夸大AI当前的能力。在我看来，科技公司正在利用AI作为幌子，以继续削减疫情期间过度扩张的人员规模。他们一边裁员，一边声称正将资源投入到增长最快的领域，以此保持投资流入。
Bubbles are only recognized in hindsight when they pop, by definition.
泡沫的定义本身就包含了事后才能被识别的特点，即只有当它们破裂时，我们才能确定曾经存在过一个泡沫。
Time to sell nvidia
是时候卖掉nvidia了
NVDA growing at 80% YOY.. how many years into the future do you think that trend will last?
When people extrapolate into infinity horrible things will usually follow
NVDA的年增长率高达80%...你认为这种趋势还能持续多少年呢？
当人们将趋势无限制地外推时，通常会导致糟糕的结果。
People need to look at the dot com bubble and realize you can be right that a certain tech will change the world, but you can also be WAY too early.
人们需要回顾互联网泡沫，并认识到尽管你可能准确地预见到某种技术将会改变世界，但你也可能会过早地押注。
And also just because AI will change the world doesn’t mean semiconductor companies should be worth trillions of dollars. By that same logic Cisco should be worth trillions from all the internet routers.
同时，仅仅因为AI将会改变世界，并不意味着半导体公司的市值就应该达到数万亿美元。按照相同的逻辑，所有互联网路由器加起来也应该让思科（Cisco）价值数万亿美元。
What is even weirder is that NVDA is 3x the market cap of TSMC while TSMC is the one producing all the chips..
更奇怪的是，尽管台积电（TSMC）是生产所有芯片的厂商，但NVDA（英伟达）的市值却是台积电的三倍。
The money appears to be in design - AMD is another company that stopped making their own chips and focused only on design. It served them well.
However I wonder how the Chips Act will impact this industry in the next 10 years.
资金似乎集中在设计环节——AMD是另一家停止自主制造芯片、仅专注于设计的公司，这一战略为它们带来了良好的效益。
然而，我很好奇在未来10年内《芯片法案》将如何影响这个行业。
That all comes down to what the competition can offer. Nvidia has significant demand and pricing power right now.
这完全取决于竞争对手能提供什么。目前，英伟达拥有巨大的市场需求和定价权。
Internet Bubble of early 2000s Made a lot of billionaires. Broke a lot of people.
2000年代初的互联网泡沫造就了许多亿万富翁，但也使许多人破产。
Exactly. Picture yourself back then in “the bubble”…what do you do? What was the billionaire vs broke strategy?
确实如此。设想一下当时身处“泡沫”之中，你会怎么做？亿万富翁和破产者的策略有何不同？
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
See it as the dot com bubble. Every business putting .com in their name and nothing to do with the web. It burst but companies we know and use today still survived cause they had an actual business.
Same with AI. Lots of shovelware and AI name plaster but there are actual AI businesses out their too
可以将当前AI领域的情况比作互联网泡沫时代。那时很多公司都在名字里加上.com，但实际上与互联网业务并无实质关系。尽管泡沫最终破灭了，但如今我们所熟知并仍在使用的一些企业之所以能够生存下来，是因为它们确实拥有实实在在的商业模式。
同样地，在AI领域也存在许多空壳产品和盲目贴上AI标签的现象，但不可否认的是，其中也有真正从事AI技术开发及应用，并具备实际商业价值的企业。
'94 to '99 was a hell of a ride if you invested well. We are in the 1995 of AI. I expect a bubble pop, but not for a few years.
'94年到'99年间，如果你投资得当，那将是一段疯狂的旅程。我们现在正处于AI发展的1995年阶段。我预计会有一个泡沫破裂的情况出现，但可能还需要几年时间才会发生。
Yup same. Were at the beginning and it shows cause i don’t think weve had that one AI company thats makes something actually valuable .
没错，情况相同。我们现在正处于AI发展的初期阶段，这一点可以从目前还没有出现一家真正创造出具有极高价值产品的AI公司这一现象中看出。
I invested in NVDA as soon as I saw ChatGPT a year ago, and later MSFT when I found out they were involved, although I plan to sell my NVDA next year sometime. I saw it as a once-in-a-decade event (like the Internet, iphone, etc).
当我一年前看到ChatGPT时，我立即投资了NVDA（英伟达），后来发现微软（MSFT）也参与其中后，我又追加了投资。不过，我计划明年某个时候卖出我的NVDA股票。我认为这是一个十年难得一遇的重大事件（就像互联网、iPhone等一样）。
Yeah, I was kind of stupid there when Meta fell into the 80s. I know Pytorch was created by Meta and it's widely use in coding for AI. But I didn't focus on that. All I thought was that Facebook was stupid and I wouldn't want to invest in Facebook. Because of that bias, I totally overlooked Pytorch.
是的，当Meta股价跌至80美元左右时，我确实犯了个傻。我知道Pytorch是由Meta开发的，并且在AI编程中得到了广泛应用。但我没有把注意力放在这点上。我当时满脑子只觉得Facebook没前途，不值得投资。由于这种偏见，我完全忽视了Pytorch的重要性及其潜在价值。
They are about to launch more AI products in the coming months. I'm not sure if they are monetizing it well enough, but they are innovating a lot in the space. If they make the right decisions over the next year, they could take over.
他们计划在接下来几个月内推出更多的AI产品。我不确定他们在盈利方面是否做得足够好，但他们确实在这个领域进行了大量创新。如果在未来一年中做出正确的决策，他们有可能实现市场的主导地位。
Nvidia feels like the AI equivalent to the dotcom darling Cisco. It powers many AI applications just like Cisco enabled the internet with their routers and networking gear. But I think ultimately most of the winners in AI will be companies that integrate it into products and services.
英伟达在AI领域的地位就如同互联网泡沫时期的宠儿思科。正如思科通过其路由器和网络设备推动了互联网的发展，英伟达也在为众多AI应用提供动力支持。但我认为，在人工智能领域最终的赢家很可能是那些将AI成功整合到产品和服务中的公司。
I don't paint with a broad brush. I would just say look at individual companies. Some companies are up to some pretty interesting things. Other companies might just be using the term AI as a smokescreen.
我不一概而论。我只想说，要仔细研究各个公司。有些公司在做一些非常有趣的事情，而其他一些公司可能只是在利用AI这个术语作为幌子。
Yes and no. Remember all the blockchain projects? NFTs?
是也不是。回想一下所有区块链项目，以及NFTs（非同质化代币）吧。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
I don't think you can compare NFTs with AI. most of the people knew from the beginning NFT was a scam.
我不认为可以将NFT与AI相提并论。大多数人从一开始就意识到NFT存在骗局的成分。
Hype is hype
炒作就是炒作。
Now it’s AI, it was EVs, then shitcoins, then bitcoin, 5G, internet of things…etc etc etc
AI is just the favor of the month until proven otherwise.
现在是AI（人工智能），之前是电动汽车（EVs），然后是山寨币、比特币，再到5G、物联网等等。
在被证明并非如此之前，人工智能仅仅是当前的热门焦点。
EVs - many investors have done well in it over the long run, notably Tesla investors. That said the margins on car companies have never been anything like those in tech companies.
Shitcoins, bitcoin - has no useful functionality. Just a medium for pump and dump.
5G - high capex from telecom's companies for barely any revenue growth. Race to the bottom for investors.
AI - already changing the nature of labour and obviously has the potential to make huge efficiency improvements in almost all areas of white collar work and creative work. The potential profit margin on it is very high. Is it due for a pullback? Yeah maybe. Is it comparable to those other scams and bad investments? No.
电动汽车（EVs）——从长期来看，许多投资者在此领域获得了丰厚回报，尤其是特斯拉的投资者。然而，汽车公司的利润率历来无法与科技公司相提并论。
山寨币、比特币——它们并无实际用途，仅作为一种炒作和抛售的媒介存在。
5G——电信公司在几乎没有带来多少收入增长的情况下投入了大量资本支出，对于投资者而言，这是一场争相触底的竞争。
人工智能（AI）——已经在改变劳动力的本质，并明显拥有提升几乎所有白领工作和创造性工作效率的巨大潜力。其潜在的利润空间非常高。是否即将出现回调？或许有可能。但它是否能与那些骗局和其他不良投资相提并论呢？答案是不能。
Everything you listed did extremely well for early investors. It's early for AI.
你所列举的所有领域都为早期投资者带来了极为丰厚的回报。对于AI来说，现在还是早期阶段。
Yes, but:
1.Not sure how much further you can bend gigacaps with trillion stock; think MSFT can sustain 20% growth? It's already huge.
2.How will AI be capitalized on? Subscxtion software? Licensing fees?
是的，但是：
对于拥有数万亿美元市值的超大型公司来说，不确定其股价还能被推高多少；比如微软（MSFT），它能否继续保持20%的增长率？毕竟它已经是一个庞然大物了。
关于如何从AI中获取收益的问题，会通过订阅软件、授权费用等方式实现吗？这仍是一个有待探讨的关键问题。
Yes. Pump and dumps often do that. Doesn’t mean the grand majority stuck around or got acquired for much.
是的，炒作和抛售通常会带来短期暴利。但这并不意味着大多数公司能够长期坚持下来或以高价被收购。
And in some cases, they are still doing well.
而在某些情况下，这些公司至今仍表现良好。
In terms of shooting valuations through the roof, yes. There's always a new hype train/bubble. The trick is to ride it and know when enough is enough. Not easy to do.
从估值飙升的角度来看，确实如此。总会出现新的炒作热潮或泡沫。关键在于要懂得在其中寻找机会并知道何时应该适可而止，但这并不容易做到。
Some thoughts about this: https://www.chaotropy.com/stock-market-why-you-might-not-see-an-ai-bubble-once-its-formed/
关于这篇文章的一些想法：https://www.chaotropy.com/stock-market-why-you-might-not-see-an-ai-bubble-once-its-formed/
Here is what the 2000’s tech boom was:
Sharing information at 0 marginal cost. Think of everything that made possible from maps to stocking shelves and all the efficiencies it generated.
Here is what the AI boom is:
Generation of information at zero marginal cost. We need some time to figure what to do with it. But the lesson corporate America learned from the dot com boom is that if you have a ceo who is proud to not use email your company dies. The hesitation to adopt new technology just isn’t there anymore because those companies died, got cheap enough to be bought out, or eventually adapted.
I forget which companies, but there were a couple that said they held down labor costs using AI for some not insignificant things. Give it another few quarters.
2000年代科技繁荣的核心在于： 以零边际成本共享信息。这使得地图、库存管理等各种应用成为可能，并极大地提升了效率。
当前人工智能繁荣的关键在于： 以零边际成本生成信息。我们需要一些时间来探索如何利用这些信息。但美国企业从互联网泡沫中学到的教训是，如果一个CEO骄傲地宣称自己不使用电子邮件，那么这家公司很可能会走向衰亡。如今企业对采用新技术的犹豫已经大幅减少，因为那些不愿接受新科技的公司要么破产倒闭，要么被低价收购，要么最终选择了适应和转型。
我记不清具体是哪些公司了，但有几家表示他们通过在某些重要领域运用AI技术成功降低了劳动力成本。再过几个季度，我们将看到更多此类应用的实际效果。
So, do you use Tensorflow, Pytorch, Keras? Or are you running projects with teams doing it? Do you know which companies are serious when they stand on a stage and say "our company is moving into AI" or not? Have you explored the degree to which adding more processing adds more value?
I can't answer that, and I work in software. But what I can see is every CEO standing on a stage saying "AI", every midwit journalist writing about AI that doesn't know the first thing about it. And that's just like what happened with the 2000s tech boom. Lots of garbage, build-to-flip companies marketing terrible ideas, and the odd Amazon and Google amongst them.
所以，你是否使用Tensorflow、Pytorch或Keras？或者你正在与团队合作运行相关项目？你知道哪些公司在舞台上声称“我们公司正转向人工智能”时是认真的，哪些不是吗？你有没有探索过增加更多处理能力能在多大程度上增加价值？
我无法回答这个问题，尽管我从事软件行业。但我看到的是每个CEO在舞台上高喊“AI”，每个对此一无所知的中庸记者都在撰写关于AI的文章。这就像2000年代科技热潮时期的情形一样。大量基于糟糕理念、旨在快速转手的垃圾公司进行市场营销，而其中偶尔会出现像亚马逊和谷歌这样的优秀企业。
People always say look for the next Amazon. But Amazon and Google and Microsoft right now are the Amazons and Google's of the earlier tech boom.!These cloud providers are developing the models and selling the servers and making the SASS products that every other fortune 500 company will be using for AI. Their insane equity valuations mean they can buy out almost any new competitor in its infancy. This is the justification for the multiples on these stocks.
人们常说要寻找下一个亚马逊。但如今的亚马逊、谷歌和微软其实就是先前科技热潮时期的“亚马逊”和“谷歌”！这些云服务提供商正在开发模型，销售服务器，并制造SaaS产品，而这些产品将被几乎所有财富500强公司用于人工智能技术。他们令人瞠目的股权估值意味着它们能够在新生竞争对手尚处于起步阶段时就将其收购。这正是这些股票高市盈率背后的逻辑依据。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
Lol im working for an “AI” company using pytorch for LLM inferences and all i can say, i dont believe in AI to do anything great outside of a few niches yet. I currently believe that we have to hire people to review the LLM outputs. They get hallucinated too often and only good for tasks that can tolerate some false positives, false negatives without much consequences. I think the current main use case can be for productivity. Ie: quicker searches and more on point for testing - rather than scouring the web at first
Inference scales up computing power with a longer inputs, so without caching, it’s a linear cost growth also.
哈哈，我正在为一家所谓的“AI”公司工作，使用Pytorch进行LLM推理，我能说的就是，目前除了少数特定领域之外，我不相信AI能做出什么伟大的事情。我认为我们必须雇人来审查LLM的输出结果。它们常常会产生错误的推断，只适用于那些能够容忍一定比例假阳性、假阴性结果且不会带来严重后果的任务。我个人认为当前主要的应用场景是提高效率，比如：更快捷地搜索信息以及在测试时提供更准确的结果，而不是一开始就去遍历整个网络。
随着输入内容的增长，推理过程会相应增加计算能力需求，如果没有缓存机制，其成本将呈现线性增长态势。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
They get hallucinated too often and only good for tasks that can tolerate some false positives, false negatives without much consequences.
This is something I've generally observed since I started looking at machine learning. It either gets used for very unimportant things, or it is more of an assistant to a person. Which, yes, has some value, but how much?
Like ChatGPT can write some code for me, but it has bugs, or it isn't well-optimised. Doesn't make use of well-known libraries like it should. If I have to fix it, it's mostly saving me some bits of typing. Which honestly just isn't worth that much. And cutting down the errors with better models doesn't really make much difference because I still have to check and fix it.
And the investment thing is companies selling dreams to people. No-one from Microsoft or Oracle is going to say that this has limited value. They want their stock rising.
它们常常产生错误的推断，仅适用于那些对假阳性、假阴性结果有一定容忍度且不会带来严重后果的任务。这是我从开始研究机器学习以来普遍观察到的现象。它要么被用于非常不重要的事情上，要么更多地充当人的辅助工具。虽然确实有一定的价值，但实际价值有多少呢？
比如ChatGPT可以帮我编写代码，但它可能会有bug，或者不够优化，并未充分利用已知的库。如果我需要修复这些问题，它其实只是节省了我敲击键盘的一些时间。老实说，这一点价值并不算太大。而且即使通过更好的模型来减少错误，也不会带来实质性的改变，因为我仍然需要检查和修正这些代码。
至于投资方面，很多公司都在向人们兜售梦想。来自微软或甲骨文等公司的人都不会说这项技术的价值有限。他们希望股价上涨。
This is why the next phase and push in Gen AI or LLM is AGI or a "group" of AI agents with specialized tasks. I had that idea, but I'm sure so did millions of other people too.
Basically AI that can monitor and guide other AI to minimize the necessary human component needed currently.
The difference with AI vs. anything else we've seen is that since it's computers / software training or improving other software, the growth is exponentially higher than anything we have witnessed.
Just think about it, AI became the hot topic to mainstream people last year. I know a lot of people who still never even used ChatGPT and barely even understand it but among the industry we went from just chatbots to Cha bots that can read documents, see images and videos, respond back with voice to fully working voice cloning and custom voice profile generation, infinite language translation in the same voice pattern and tone, photo generation and THEN solving the text issue for full graphic design creation in seconds, and we're on like phase 5 of that already and full music generation with vocals too and video generation and we're already in like phase 2 of that. They are now working on real world simulation and being able to teach other AI and robots how to do years of evolution, learning, and training within days to be even more efficient and impactful.
Keep in mind while a lot of this probably has been in the works, we already have WORKING applications for ALL of this in just one year and it still hasn't even been adopted by the majority yet.
By the time it is actually fully adopted by the average person it will be like using LCARS or J.A.R.V.I.S. or asking a genie to create you personal customized content on the fly.
This is like dot com, social media, the industrial age being rolled into one.
Once it becomes a thing, it will be used everyday and will completely change how society operates.
We went from social media not being a thing to some people are addicted and use it daily or even hourly or even in a higher frequency than that. Now imagine how people use social media now but with AI and robots that actually can physically manipulate or create.
这就是为什么在生成式AI或LLM的下一阶段和推进方向是AGI（通用人工智能）或由具有专门任务的“群体”AI代理组成的原因。我有过这样的想法，但我相信也有数百万其他人同样有此想法。
基本上，这是一种能够监控并指导其他AI以尽可能减少当前所需人力投入的AI。
与我们所见过的任何其他技术相比，AI的独特之处在于它是通过计算机/软件训练或改进其他软件，其发展速度比我们已见证过的任何事物都要快得多。
试想一下，AI去年才成为主流人群关注的热点话题。我知道很多人至今甚至从未使用过ChatGPT，对其也知之甚少，但在业内，我们已经从简单的聊天机器人发展到了能够阅读文档、查看图像和视频、用语音回应并实现完全的语音克隆及定制化语音配置生成、无限语言翻译且保持相同的语音模式和语调、照片生成，然后解决了文本问题，能在几秒钟内完成完整的图形设计创作；同时，在音乐生成方面，包括人声部分，以及视频生成领域，我们也已经进入第二阶段的发展。现在，研究者正在致力于真实世界模拟，并教导其他AI和机器人如何在几天内完成数年的进化、学习和训练，从而变得更加高效和有力。
请记住，虽然很多技术可能已经在研发中，但仅仅一年时间里，所有这些技术就已经有了可实际应用的产品，而且尚未被大多数人采用。
等到这项技术真正被普通人全面采纳时，它的体验将会像使用科幻片中的LCARS界面、J.A.R.V.I.S.智能助手，或是向精灵许愿实时创造个人定制内容一样神奇。
这就像互联网泡沫时代、社交媒体时代和工业革命时代的结合体。
一旦它成为现实，将被日常使用，并彻底改变社会运作方式。
从社交媒体尚未存在到如今一些人对其上瘾，每日甚至每小时甚至更高频率地使用，想象一下人们现在使用社交媒体的方式，如果再加上能进行物理操作或创造的AI和机器人，会是怎样一番景象。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
All I know is NVDA is selling the shovels
我所知道的是，NVDA在当前的AI技术浪潮中是售卖工具的角色。
So were Sun Microsystems for a while. And then a couple of years later, they weren't.
https://finance.yahoo.com/news/what-were-you-thinking-180823527.html
Sun Microsystems在一段时间内也曾处在这样的位置。但几年后，他们就不再是了。
链接文章：https://finance.yahoo.com/news/what-were-you-thinking-180823527.html
Yes this is why it's a bubble. Certainly won't last forever but AI hype is still going very strong and the need for highly efficient GPU compute isn't letting up anytime soon.
是的，这就是为什么说这是一个泡沫。当然这个热潮不会永久持续，但AI炒作目前仍然非常猛烈，并且对高性能GPU计算能力的需求在短期内并无减弱迹象。
I'm not saying I'm smart at all. But I'm not investing in ai and I use ai services.
我并没有说自己很聪明。只是说我现在并未投资AI领域，但我在日常中确实会使用AI服务。
Yes. The amount of promises being made in the industry VASTLY outnumber the actual deliverables. I work with some of these companies, the claims I see are 90% bullshit; maybe its different with other companies but I doubt it.
In time itll crash (maybe 3 years), then go back to gradual development while it changes the world over the course of decades.
是的，在这个行业里，所做出的承诺数量远远超过了实际可交付成果。我在与一些此类公司合作时发现，其中90%的声明都是空话；也许其他公司在这一点上有所不同，但我对此表示怀疑。
随着时间推移，这一热潮终将会破灭（可能在3年内），然后回归到渐进式发展状态，并在数十年的时间里逐渐改变世界。
I love how every product is powered by AI or Machine learning.
我确实很喜欢现在每款产品都宣称自己是基于AI或机器学习技术驱动的这一现象。
AI is a revolutionary tech on par with computers and internet.
But with the exception of chip (NVDA, AMD, TSMC) and cloud oligopoly (AMZN, MSFT, GOOG), I doubt other companies share price will benefit from it.
There's a lot of open source models so the knowledge is widely dispersed. Its hard to build a proprietary product.
In fact, in racing to incorporate AI into their products, companies may be forced to incur extra expenses just to maintain competitiveness.
Just like today, every company has a website. Website is now required expense not a differentiating factor. Same with AI.
人工智能是一项与计算机和互联网相提并论的革命性技术。
但是，除了芯片（NVDA、AMD、TSMC）和云服务寡头（AMZN、MSFT、GOOG）之外，我怀疑其他公司的股价是否能从中受益。
由于存在大量开源模型，相关知识已经得到了广泛传播，因此构建专有产品颇具挑战。
事实上，在竞相将AI融入其产品的过程中，企业可能被迫承担额外开支，仅仅是为了保持竞争力。
就像现在这样，每家公司都有网站。网站已经成为必需的成本支出，而非区别于竞争对手的独特优势。同样地，AI也可能会变成这样的常规成本支出，而非竞争优势。
AI is a bubble like the Internet was in a bubble in 1995, and like the iPhone was in a bubble in 2009.
AI就像1995年互联网那样处于泡沫阶段，也类似2009年iPhone所经历的泡沫时期。
I think AI could replace any job that’s done at a computer especially if it’s repetitive to some degree. Does seem to be able to at least simulate like it’s fluent in the language and I think AI programs could actually be used to design websites and do computer programming the same way they master the English language. AI will be able to chat and generate websites for people based on minimal input. I think AI could also be used to moderate and regulate social media platforms without the use of people or administrators.
我认为人工智能能够取代任何在计算机上完成的、尤其是具有一定重复性的工作。AI似乎至少能模拟出流利的语言表达能力，我相信AI程序可以像掌握英语那样用于设计网站和进行计算机编程。AI将能够基于少量输入与人们聊天，并自动生成网站。同时，我认为AI还可以用于社交媒体平台的管理和监管工作，无需人工或管理员介入。
The issue will be as it always is, greed. People that missed NVIDA and the other big boys are desperately looking for the next AI stock to go . Eventually it will fall apart and only the real legitimate companies will survive but I suspect over the next 5 years there will be opportunity to make money. Just have to set a limit and sell when appropriate. Most people though will be left holding bags of worthless shit. It's going to be very interesting.....
问题始终如一，那就是贪婪。那些错过英伟达和其他大公司股票机会的人正急切地寻找下一个有望爆发式增长的AI概念股（用符号形象地表示其潜力巨大）。最终市场泡沫会破裂，只有真正有实力和合法合规的公司才能生存下来，但我推测，在接下来的五年里依然会有赚钱的机会。关键在于设定好止盈止损点，并在适当的时候卖出。然而，大多数投资者可能会手持一堆毫无价值的股票而无法脱身。这一过程将会非常有趣……
Imo no, I agree with everything you said and there will likely be a bubble but this is nothing. Maybe early innings of the bubble.
我的观点是不完全赞同，虽然我同意您所说的关于泡沫存在的可能性，并且确实可能会出现泡沫现象，但现在的情况并不严重。也许这只是泡沫初期阶段而已。
Of course it's a bubble. The amount of crap written on the topic of AI is ridiculous at this point, as are the stock narratives / valuations. The tendency to extrapolate a couple years of strong growth well into the distant future never ceases to amaze.
当然，这确实是一个泡沫。当前关于人工智能话题的大量言论简直是荒谬可笑，与此相关的股票叙事和估值也是如此。人们总是一味地将几年的强劲增长趋势线性外推至遥远的未来，这种倾向令人惊讶不已。
@IHeartLife
When there is a gold rush sell pickaxes when there’s AI hype get in on the good chips
当出现淘金热时，要出售的是铲子；而当AI热潮兴起时，则应投资于优质的芯片。
Do you believe it’s not going to be an asset in the foreseeable future?
你认为它在未来可预见的时期内不会成为一项有价值的资产吗？
Not a Bubble.
https://yardeni.com/charts/stock-market-p-e-ratios/
Look at the p/es then vs now
NVDA sells as 30x forward earnings, that’s high but not near a boiling point
不是泡沫。
[参考链接：https://yardeni.com/charts/stock-market-p-e-ratios/]
请查看那时与现在的市盈率对比。
NVDA（英伟达）股票的前瞻性市盈率为30倍，虽然这个数字相对较高，但还未接近沸腾点。
Can it sustain that growth?
它能否持续保持这样的增长态势呢？
AI as a technology is absolutely positively not a bubble.
The market speculation surrounding future profits for business is speculative, therefore possibly a bubble in 10 years.
We’re at the bottom of an exponential increase in machine intelligence. Just getting warmed up for the game. Haven’t even started.
人工智能作为一种技术绝对不是泡沫。但是，围绕企业未来利润的市场投机具有一定的投机性，因此在10年内可能存在泡沫风险。目前我们正处于机器智能指数级增长的早期阶段，就像是游戏才刚刚热身，甚至可以说还未真正开始。
I was working for a startup from 1998 to 2001. There were thousands of them. The Tech Bubble was the entire market.
This is nothing like that. This is a few companies sticking out above the rest over a new technology. I've owned Nvidia and Microsoft for years. I'll own them after this blip, and until the next one.
Cute story: I held stock options worth half a million in 2000. When they vested in 2001, I sold them for $600.
我曾在1998年至2001年为一家初创公司工作。当时有成千上万的初创企业，科技泡沫涵盖了整个市场。而现在的情况与那时截然不同，现在只是少数几家公司因为在新技术领域领先而脱颖而出。我已经持有英伟达和微软的股票多年，并且在这次波动后仍将继续持有，直到下一次转变。
有趣的故事是：我在2000年拥有价值50万美元的股票期权。当它们在2001年到期可行使时，我以600美元的价格将其卖出。
No—AI is invaluable to the future economy and can offset labor shortages and streamline supply chain issues/management. Just like every new industry there will be winners and losers, but it is not a bubble.
不，人工智能对于未来经济来说是不可或缺的，并且可以抵消劳动力短缺问题，优化供应链管理。就像任何新兴行业一样，会有赢家和输家，但它绝不是一个泡沫。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
Doesnt mean it cant be a bubble. The internet revolutionized the world, the stock market was still overvalued as fuck in 2000 resulting in the dot-com bubble
这并不意味着它不可能成为一个泡沫。互联网确实改变了世界，但在2000年时股市仍然严重高估，从而导致了 dot-com 泡沫的发生。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
Sure, but I don’t think AI is overvalued in light of the key areas it services. Dot-Com was the convergence of innovation in the communications and commerce space in a 10 year period. The excess that promulgated was simply because every business or trade necessitated the use of these tools and they all came to being within essentially the same time frx (mobile phones, accessible high speed internet, and PayPal/ecommerce). AI is much more of a general integration tool rather than a tool of dominance and its value is far less accessible across the board. Sure, somebody may use it for their resume and some HR firm may implement anti-AI tools, but the real value and focus in the industry is to automate labor in highly mundane-detailed workforces. This is the very thing which the economy is starved for. I’m not talking about the cool sexy AI where we have a robot butler, I’m talking about weapons, tracking, information synergy, and validation of deliveries. This is where the bulk of the AI industry is focused. There is a heavy need for this at a very high level and companies are invested here. AI doesn’t impact the broader consumer market and so too volatility in that market as the Dot Com Bubble—also an era (unlike this) facing a decline in interest rates and a surge in consumer demand. The fed is trying to cool consumer demand at the moment and keeping interest rates high. This increase in pressure helps to keep risky investing at bay.
当然，但我认为考虑到AI服务的关键领域，它并未被高估。互联网泡沫是通信和商业空间在10年内创新融合的结果，当时所有业务或行业都迫切需要使用这些工具，并且它们几乎在同一时间框架内迅速涌现（如移动电话、可访问的高速互联网以及PayPal/电子商务）。
相比之下，人工智能更像是一种通用整合工具而非主导性工具，其价值在整个行业内远没有那么广泛易得。确实，有人可能会在简历中提到AI技能，一些人力资源公司也可能会实施反AI工具，但行业真正关注的价值和焦点在于自动化高度繁琐精细劳动的工作岗位。这是当前经济极度渴望的东西。
我谈论的不是那种拥有机器人管家的酷炫性感AI，而是指武器系统、追踪技术、信息协同及交付验证等应用。这才是AI产业集中关注的核心部分。在高级别上对这类AI有着强烈的需求，许多企业都在这一领域进行投资。
与互联网泡沫时期不同的是，AI并未直接影响到广大消费者市场及其波动性，那时正面临利率下降和消费需求激增的时代。目前美联储正试图抑制消费者需求，并维持较高利率水平，这种压力有助于抑制风险投资。
Nasdaq-100 is barely over its ATH from 2021 and everyone thinks we're in a bubble. If you want to compare this to the Dotcom bubble, then we're in the first inning.
纳斯达克100指数目前仅略高于其2021年的历史高点，但每个人都认为我们正处于泡沫之中。如果要拿当前的市场与互联网泡沫时期相比的话，那我们现在可能才刚刚进入初级阶段。
No
I wish I had the risk tolerance to go hard on it
IMO it's massively underpriced
It will create a bubble though
Like we're approaching the end of all labor
Wtf does that even mean
I have no idea
I'm a Boglehead but I own some AI stocks on the side despite worse odds statistically
不，我倒希望我能有承受高风险的能力，加大对它的投资力度。我认为AI领域的定价目前严重偏低。不过它确实可能会催生一个泡沫。
就像我们似乎正走向所有劳动的终结，这到底意味着什么呢？说实话，我自己也搞不清楚。
虽然从统计学角度来说持有AI股票可能胜算较低，但我还是遵循博格尔投资理念的同时，持有一些AI相关的股票作为辅助投资。
Basically yes but the "bubble" could get eaten up if AI has a major break through which it is somewhat likely to do.
基本上是的，但如果AI取得重大突破（这种可能性是存在的），那么这个“泡沫”可能会被消化掉。
Most of the companies that are developing AI will not be the ones marketing it. Its too large to centralize. The ones that capitalize on it will mostly be your chain stores and manufactures.
大多数开发AI技术的公司不会是直接将其推向市场的主体，因为AI技术太过庞大，难以集中由单一实体进行推广。真正从AI中获益并利用其优势的，大部分将是连锁商店和制造商等企业。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
And who will sell them the models and services? Cloud providers.
那么，谁将向这些企业出售模型和服务呢？答案是云服务提供商。
I do not see a substantial market for subscxtion service. I do see one for devices, toys, manufacturing equipment, maybe even the medical industry. People hate to pay for services, but love buying items.
我认为订阅服务市场不会有很大的发展空间，但我确实看到在设备、玩具、制造设备，甚至是医疗行业等领域存在市场需求。人们通常不喜欢为服务付费，但却热衷于购买实物产品。
Depends on the future. The future growth and performance determines if this is a “bubble.” I’m not going to touch it directly but I have large positions in companies like AMZN and META.
Overall since this is the value investing subreddit, I will simply say that when you have companies this (seemingly) expensive, the risk goes up. Again, I use the word seemingly because the stocks may be undervalued if we underestimate the growth of AI. Taking risk into account and my lack of industry knowledge, I am not willing to play AI directly.
这取决于未来的发展。未来的增长和表现将决定这是否是一个“泡沫”。我不会直接涉足这一领域，但我持有像AMZN（亚马逊）和META这样的大公司的大量股份。
总的来说，鉴于这是价值投资相关的论坛，我要说的是，当公司估值（看似）如此之高时，风险也随之增加。再次强调，“看似”这个词是因为如果我们低估了AI的增长潜力，这些股票可能实际上是被低估的。考虑到风险因素以及我自己对行业的了解不足，我不愿意直接投资于AI领域。
yanno, this q has levels. with regard to mega caps, i dont think AI is a bubble at all. But there is a lot of trash out there too.
你知道的，这个问题有多个层面。对于大型公司（巨头企业）来说，我认为AI并不是一个泡沫。但同时市场上确实也充斥着很多质量不高的东西。
AI does have great potential, but it's important to distinguish between the hype surrounding startups and the actual value of established companies driving AI development. It's actually a good move to invest in big companies like Nvidia and cloud services that have a strong foundation for AI. Just be careful not to get too carried away with the excitement and expect endless growth.
人工智能确实具有巨大的潜力，但重要的是要区分围绕初创公司的炒作与实际推动AI发展的老牌企业的价值。投资于像英伟达这样在AI领域有坚实基础的大公司以及云计算服务实际上是一个明智之举。不过，要注意不要过于沉迷于兴奋之中，期待无止境的增长。
Yes, it's definitely a bubble, when will it pop? Nobody knows. It could last 2 more years... It could end tomorrow.
是的，这确实是一个泡沫，但没人知道它何时会破裂。这个泡沫可能还会持续两年……也可能明天就会结束。
原创翻译：龙腾网 https://www.ltaaa.cn 转载请注明出处
Yea its overhyped. Buy companies that want to utilize ai but have other moats. Like buying Microsoft is a pretty good deal because they have tons of good businesses that are not dependent on whether or not ai succeeds.
Ai problem is expectations. People assume this new technology will overthrow everything. As usual with technology, we will find its strength and weaknesses, society will progress with some jobs being eliminated and others being created.
是的，AI确实被过度炒作。应该投资那些打算利用AI但同时拥有其他竞争优势的公司。比如购买微软就是一个相当不错的交易，因为他们有很多优质的业务并不依赖于AI是否成功。
AI面临的问题在于人们的过高期待。人们认为这种新技术将颠覆一切。一如往常，随着技术的发展，我们会发现它的优点和局限性，社会将会进步，一些工作岗位会被淘汰，而新的职位也会随之产生。
Of course it is ... Internet is the most crazy thing humanity invented and guess what ? 2001 crash.
当然，人工智能领域存在泡沫现象。互联网是人类发明的最疯狂的事物之一，而回想一下2001年的互联网泡沫破裂。
I once heard something along the line of „we overestimate impact of new technologies in short term but underestimate their impact in the long term”. I think it really is a nice summary of current AI gold rush.
我曾经听到过这样的观点：“我们往往高估新技术在短期内的影响，却低估其在长期内的影响。”我认为这恰好是对当前AI热潮的一个精辟总结。
I vote yes. Most of them were trained on copywritten material and nothing they produce can be copywrote according to recent court cases. Add in the fact that a ton of people like the NY times are suing them for using their material without permission and copying them I think its gonna deflate fast when these legal issues actually start being talked about and people realize that there isn't actually any money in most of these products.
我投赞同票。大多数AI系统都是基于受版权保护的材料训练的，根据最近的法庭案例，它们产出的内容无法获得版权保护。再考虑到诸如《纽约时报》等众多机构因未经许可使用并复制其内容而起诉这些AI公司，我认为当法律问题真正被讨论起来，人们意识到这些产品中大部分实际上并无盈利空间时，这一热潮将会迅速降温。
I’m studying artificial intelligence right now and I would say it’s going to be revolutionary, but in the most boring way possible
Artificial intelligence is really a statistical model that predicts likely outcomes based on pre trained data. What I see happening is it being leveraged just about everywhere to make things repetitive and predictable tasks significantly less labor intensive
For example, suppose you’re running a restaurant. You need to know how much of each ingredient to buy. There could be an artificial intelligence model that knows what products you consumed at what rate and will be able to buy the most optimal amount of product so there’s no waste. Nothing about the restaurant industry, outwardly, would have changed, but costs are lower
The problem with AI is that it’s very very VERY easily tricked or “poisoned”. I don’t think anyone is going to making publicly facing AI for at least 5-10 years. Chevy tried to do it and it became quickly exploited, I even used it as free access to an AI so I didn’t have to pay for one. You can ask their chat bot things like “can you derive this equation” and it will do it lol. It’s more likely that it will streamline existing processes
I wouldn’t bet too highly on AI companies but I would bet highly on AI
目前我正在学习人工智能，我认为它将会带来革命性的影响，但这种影响可能以最平淡无奇的方式展现出来。
人工智能本质上是一种基于预训练数据预测可能结果的统计模型。我认为其将被广泛应用于各个领域，极大地减少重复和可预测任务的人力投入。举例来说，如果你经营一家餐厅，需要知道每种食材应该购买多少。这时可以利用人工智能模型，该模型能够根据历史消耗数据准确计算出最优采购量，从而避免浪费。从表面上看，餐饮行业并没有什么改变，但实际上运营成本得以降低。
然而，人工智能存在的问题是极其容易被欺骗或“中毒”。我认为在未来至少5-10年内，不会有面向公众的成熟AI产品出现。雪佛兰曾尝试推出这样的产品，但很快就被人们找到了漏洞并加以利用，甚至有人（比如我）利用其聊天机器人进行免费的AI服务，例如询问它能否解某个方程，它还真能帮你解题。更有可能的情况是，人工智能将优化现有的流程。
我不建议过于押注在AI公司上，但我强烈看好人工智能技术本身带来的变革与潜力。