2024年怎么过?这是Sam Altman希望你早些知道的事

币圈资讯 阅读:52 2024-04-22 11:28:38 评论:0
美化布局示例

欧易(OKX)最新版本

【遇到注册下载问题请加文章最下面的客服微信】永久享受返佣20%手续费!

APP下载   全球官网 大陆官网

币安(Binance)最新版本

币安交易所app【遇到注册下载问题请加文章最下面的客服微信】永久享受返佣20%手续费!

APP下载   官网地址

火币HTX最新版本

火币老牌交易所【遇到注册下载问题请加文章最下面的客服微信】永久享受返佣20%手续费!

APP下载   官网地址

这个家伙正在塑造世界,如果他学到了什么,我们应该注意。

「日子很长,但几十年很短。」

Sam Altman 九年前写下这篇箴言体博客时,不知是否预感过自己会成为那个领导 AI 革命的人。
2023 年行将结束之际,他时隔多年重拾箴言体,发布了一篇题为《那些我希望有人告诉我的(事)》的博客,将过去一年所学凝结在了十七条短句中。(下文黑体部分,对黑体部分的解读来自编辑。)

过去一年激动又疯狂。

Altman 领导的 OpenAI 推出了一个世人真正喜欢并从中受益的 AI 工具,世界也开始认真对待 AI 。

从积极的一面来看,今年我学到了很多东西,他写道。

如果你正在认真思考 2024 年要做什么,希望这篇博客能对你有所帮助。

1、乐观、沉迷、自信、原始动力(raw horsepower)和与人建立联结,就是事情的打开方式。
OpenAI 最初几年很艰难,部分原因是连他们自己都不知道这是不是在追求一个极其昂贵的死胡同。

复制一些东西很容易,做真正的创新却真的很难。
什么是对的感觉?是正确的打开方式?Altman 总结出上面几个关键词。
这也和纽约客、大西洋月刊等刊物对他本人的描述非常相似——他是一个乐观主义者,而且极度自信,即使以硅谷的标准来看。
和马斯克一样,对自己未来愿景抱有坚定信念。与乔布斯一样,有一种救世主般启迪他人的能力。

YC 联合创始人曾如此评价当时仅 20 出头的 Altman,「你可以把他空降到一个满是食人族的小岛上,五年后再去看,他是那里的王了。」

他曾在接受采访时说过,我和大多数 AI 公司不同之处在于,我认为 AI 是好的。
不会私下怨恨自己的事业,我认为它很棒。


对了,在「与人建立连结」上,他早在九年前就写过,「多与人交谈。多阅读长内容,少阅读推文。少看电视。花更少的时间上网。」
2、一个有凝聚力的团队、平衡冷静和紧迫感,还有超乎情理的投入(unreasonable commitment )——这就是完成事情的方式。眼光长远(长期导向)并不常见;尽量不要担心人们短期内的想法,随着时间推移,你会更容易做到这些。
完成事情的方式是什么样的?这是 Altman 的回答。
他曾说,一支更强大、更团结、更专注、更忠诚的团队帮忙 OpenAI 渡过了难关。
飓风的狂暴只是外在,穿过飓风墙抵达飓风眼,你会发现这里反而很平静。
「我认为,我们之前有很强的信念和专注力,现在,我们有更多。」他说。
谈及自己对OpenAI 的用心——过去四年半,我全身心地投入到这家公司,但实际上比这个时间长,我的绝大部分时间都投入进去了。公司也在安全且有益的 AGI 上取得了巨大进展。
在接到董事会重返公司的邀约时,他花了几分钟才振作起来,克服了自我和情绪,然后说:「是的,我当然想这样做。」
3、对一个团队来说,做一件真正重要的困难的事情比做一件无关紧要的简单的事情要容易得多,大胆的想法能激励人。
确实。2021年,他写下《万物摩尔定律》,一份乌托邦宣言,预言 AI 革命(他在领导)将泽被地球。在那篇宣言里 AI 将创造惊人财富,改变工作性质,减少贫困。

作为核聚变的狂热粉丝,他认为,这种技术会与生成式 AI 一起,让知识和能源的成本以「漂亮的指数曲线」下降。正如我们看到的,过去一年,他以传教士般的狂热推动着 OpenAI 往前飞奔。

4、激励措施具有超能力,请谨慎设定它。
5、将资源集中在少量有高度信心的赌注( high-conviction bets)上,说易行难。可以被删除的选项比你想象的要多。
毫无疑问,这里的赌注应当算上过去一年OpenAI 在安全且有益的 AGI 上取得的巨大进展。要知道OpenAI 最初几年很艰难,部分原因是连他们自己都不知道这是不是在追求一个极其昂贵的死胡同。
6、沟通尽量清晰、简洁。
7、遭遇官僚主义和废话,请与之抗争,也让其他人与之抗争。不要让组织架构妨碍大家高效合作。
连同上面第 6 条,他再次重申了多年前的经验——做一个实干家,而不是一个空谈家。
8、结果才是最重要,好的过程不能成为坏结果的借口。
9、花更多时间招聘。敢于投资那些潜力巨大且进步迅速的人。除了智力,也要寻找对方能否胜任工作的证据。
Altman 曾经公开讲过,慧眼识人,就不要局限于他人过往的工作经验和当前成就,需要发掘那些有潜力且能在短时间激发潜能的人。
每当遇到新人,Altman 都会扪心自问「这个人有异于常人的能力吗?」对于渴求人才的人来说,这个问题很值得思考。
驱动力也是他了解他人时最先考察的点。
10、超级明星比他们看起来更有价值,但评估他们还要看对组织绩效的净影响(net impact)。
11、快速迭代可以弥补很多不足;如果迭代快速,犯错通常是可以接受。计划应该以十年为单位,执行应该以周为单位。
Altman 认为,经验主义是进步关键。


伟大的创新出现时,它几乎肯定会以一种混乱、不完整和令人困惑的形式出现。
对于发现者本人来说,它只能理解一半;对其他人来说,这将是一个谜。
处理高度不确定时,选择不做什么可能很困难,决策变得棘手。这时,不要低估「完成胜于完美」建议的价值。
马斯克 SpaceX 可谓是典范。
为了快速建造大型宇宙飞船,每一次新的迭代都基于上一次迭代的改进——无论是更流畅的制造工艺、减少不必要的重量、提高性能等。
建造许多火箭可以实现逐渐逼近目标。马斯克曾用一个简单的方程比喻这个过程:任何给定技术的进步都是迭代次数乘以迭代之间的进步。
这和传统航空航天机构的做法不同,后者通常花费数年时间去完善设计,虽然这样会减少爆炸的几率,但也需要耗费更多时间和资金。

这一条也是 Altman之前一条推文的的再现。
至于计划应该以十年为单位,不妨看看他在 2019 年做的预测,特别是第二条。


12、不要与近乎物理定律的商业法则(business equivalent of the laws of physics)对着干。
我们推测这句话是指一些影响商业运作的基本假设,比如人性的短视、厌恶改变。
举个例子。牛顿第一定律讲述了运动中的物体有维持运动的趋势,静止中的物体有保持静止的趋势,除非受到其它物体的作用力迫使它改变这种状态。
人性也是如此。Altman 曾谈到一些在谷歌、Meta 长期工作的创始人存在的问题:
舒适的生活、稳定的工作和无论做什么都会取得成功的名气,会让他们很难抛弃这些,即使他们真的离开公司了,也非常有可能再回来。与长期利益相比,短期诱惑和便利往往更具吸引力,也更符合人的天性。
再比如熵增原理。在自然界,熵增是一个自发的、由有序向无序发展的过程。从这个角度来看,ChatGPT 目前带来的无序,让信息控制变得非常困难,也是我们必须面对的现象。
13、灵感稍纵易逝,人生如白驹过隙。不作为是一种特别隐蔽的风险。
企业、市场能够创造指出增长的奇迹,但时间不能,它对每一个人都是公平的。
Altman 对时间非常看重。这次也再度共鸣了九年前那篇博客的一些想法。
比如,「时间极其有限,而且过得很快。」  

「世界上所有的钱都无法挽回逝去的时光。」

「不要浪费时间。大多数人都浪费了大部分时间,尤其是在生意上。」
所以,「做让你快乐和满足的事情——无论如何,很少有人在死后几百年后还被人记住。
不要做不会让你快乐的事情(这种情况最常发生在别人希望你做某事时)。
不要花时间试图与你不喜欢的人维持关系,将消极的人从你的生活中剔除掉。
消极情绪真的很糟糕。不要让自己为不做自己想做的事情找借口。」
14、 规模(scale)通常具有令人惊讶的涌现特性。(Scale often has surprising emergent properties.)
对于这一点,OpenAI 研究者、思维链提出者 Jason Wei 应该是最有感悟的人之一。
Jason Wei 等人的论文是第一篇发现当使用链式思维进行提示时,存在一个相变现象,表明大型模型在很大程度上优于较小的模型,这进一步导致了涌现能力的发现。
有意思的是,他也特别针对这一条做了回应,规模与涌现特性的关系,不仅存在于语言模型,还有「人类当中」。


15、复利指数(Compounding exponentials )很神奇。特别是,你真的想建立一个通过规模获得复利优势(a compounding advantage with scale)的企业。
在那篇广为流传的文章how to be successful 里,Altman 谈到过复利具有神奇的魔力,指数曲线是创造财富的关键。世界上少有企业具有真正的网络效应和高度的可扩展性,但是随着技术进步,这种情况会逐渐改变,这值得我们不断为之努力。
16、重新站起来,继续前进。
在他 38 年的人生历程中,OpenAI 那场人事动荡并非他第一次陷入漩涡中心。他也会在社交媒体上说「整夜梦见失败。」
他也说过,「只要有足够的时间,几乎任何事情都可以适应,无论好坏。人类在这方面是很了不起的。」


17、与优秀的人一起工作是生活中最美好的部分之一。
历经 OpenAI 震荡,这也是 Altman 最大的感受之一。

他说过,如何在努力工作的同时,又不至于透支身体?有条几乎不会出错的黄金准则——与相处愉快的人一起从事喜欢的工作。
一如九年前写过的,「竭尽全力与聪明、有趣、雄心勃勃的人在一起。为他们工作并雇用他们(事实上,工作中最令人满意的部分之一就是与真正优秀的人建立深厚的关系)。
尝试与那些在其所在领域处于世界最佳水平或非常有前途但完全不为人所知的人共度时光。确实,你会达到与你相处最多的人的平均水平。」
参考链接
https://blog.samaltman.com/what-i-wish-someone-had-told-me


This guy is shaping the world. If he has learned something, we should pay attention to the fact that the days are long, but decades are short. When he wrote this proverbial blog nine years ago, I don't know if he had a hunch that he would become the person who led the revolution. At the end of the year, he regained the proverbial style after many years and published a blog entitled "What I want someone to tell me", which condensed what he had learned in the past year into seventeen short sentences. The interpretation of the bold part below came from the editor's excitement and madness in the past year. On the positive side, I have learned a lot this year. He wrote that if you are seriously thinking about what to do next year, I hope this blog will help you. Optimism, obsession, self-confidence, original motivation and connection are the ways to open things. The first few years are difficult, partly because they don't even know if they are pursuing an extremely expensive dead end. It is easy to copy something. It's really hard to make a real innovation, but what is right and what feels right is the right way to open it. Summarize the above keywords, which is very similar to the description of himself in The New Yorker Atlantic Monthly and other publications. He is an optimist and extremely confident. Even by the standards of Silicon Valley, he has the same firm belief in his future vision as Musk, and he has the same messianic ability to enlighten others as Jobs. The co-founder once commented that you were only in the early days. You can parachute him into a place. I will go to see him on an island full of cannibals five years later. He once said in an interview that the difference between me and most companies is that I think it's good and I don't hate my career privately. I think it's great. By the way, nine years ago, he wrote too much to talk to people, read too much, read too much, watch less tweets, watch TV and spend less time surfing the Internet. A cohesive team balances calmness, urgency and unreasonable investment. This is the way to get things done. Long-term vision and long-term orientation are not common. Try not to worry about people's short-term thoughts. What are the ways to accomplish these things more easily over time? This is the answer. He once said that a stronger, more United, more dedicated and more loyal team helped tide over the difficulties. The fury of the hurricane only crossed the hurricane wall and reached the hurricane eye. You will find it calm here. I think we had strong beliefs and concentration before, and now we have more. He said that he talked about his intentions in the past. For four and a half years, I devoted myself wholeheartedly to this company, but in fact, I devoted most of my time to it, and the company made great progress in safety and benefit. When receiving the invitation from the board of directors to return to the company, it took him a few minutes to get up and overcome himself and emotions, and then said, yes, of course I want to do so. It is much easier for a team to do a really important and difficult thing than to do an insignificant and simple thing. Bold ideas can inspire people. In 1998, he wrote Moore's Law of Everything, a utopian declaration predicting revolution, and he was leading the revolution. In that declaration, he would create amazing wealth, change the nature of work and reduce poverty. As an avid fan of nuclear fusion, he believed that this technology would reduce the cost of knowledge and energy with a beautiful exponential curve together with generative technology. As we have seen, in the past year, he pushed forward with missionary fanaticism, and the incentives were super-powerful. Please set it carefully and concentrate resources in a small amount. There are more options that can be deleted than you think. There is no doubt that the bet here should count the great progress made in safety and benefit in the past year. You know, the first few years are very difficult, partly because they don't even know if they are pursuing an extremely expensive dead end. Communicate as clearly and concisely as possible and encounter bureaucracy and nonsense. Please fight it and let others fight it. Don't let the organizational structure hinder everyone's efficient cooperation. Together with the above item, he will try again. I reiterated the experience of many years ago, being a doer rather than an empty talker. The result is the most important thing. A good process cannot be an excuse for a bad result. We should spend more time recruiting people who dare to invest and make rapid progress. Besides intelligence, we should also look for evidence that the other party is competent for the job. We have publicly said that we should not be limited to other people's past work experience and current achievements. We need to explore those who have potential and can stimulate their potential in a short time. Whenever we meet new people, we will ask ourselves. Does this person have any ability different from ordinary people? For people who are eager for talents, this question is worth thinking about. The driving force is also the first point when he gets to know others. Superstars are more valuable than they look, but evaluating them depends on the net impact on organizational performance. Rapid iteration can make up for many shortcomings. If iteration makes mistakes quickly, it is usually acceptable. The plan should be implemented in units of ten years. Empiricism is the key to progress. When great innovation appears, it will almost certainly be based on. A chaotic, incomplete and confusing form appears. For the discoverer himself, it can only be understood by half. For others, it will be a mystery. When dealing with high uncertainty, it may be difficult to choose what not to do, and decision-making may become difficult. At this time, don't underestimate the value of completion over perfect advice. Musk can be described as a model. In order to build a large spacecraft quickly, every new iteration is based on the improvement of the last iteration, whether it is smoother manufacturing technology, reducing unnecessary weight and improving performance. Many rockets can approach the target gradually. Musk once used a simple equation to describe this process. The progress of any given technology is the number of iterations multiplied by the progress between iterations, which is different from the traditional aerospace institutions, which usually spend several years to improve the design. Although this will reduce the probability of explosion, it will also take more time and money. This article is also a reappearance of the previous tweet. As for the plan, it should be in units of ten years. Let's take a look at the prediction he made in 2008. It's the second rule, don't go against the business rules that are close to the laws of physics. We guess this sentence refers to some basic assumptions that affect business operations, such as short-sighted aversion to change of human nature. For example, Newton's first law tells that objects in motion have a tendency to maintain motion, while objects in rest have a tendency to remain stationary unless they are forced to change this state by other objects. The same is true of human nature. Some founders who have worked in Google for a long time have talked about problems, such as a comfortable life, a stable job and a reputation that will make it difficult for them to abandon them even if they really leave the company. 比特币今日价格行情网_okx交易所app_永续合约_比特币怎么买卖交易_虚拟币交易所平台

文字格式和图片示例

注册有任何问题请添加 微信:MVIP619 拉你进入群

弹窗与图片大小一致 文章转载注明 网址:https://netpsp.com/?id=62162

美化布局示例

欧易(OKX)最新版本

【遇到注册下载问题请加文章最下面的客服微信】永久享受返佣20%手续费!

APP下载   全球官网 大陆官网

币安(Binance)最新版本

币安交易所app【遇到注册下载问题请加文章最下面的客服微信】永久享受返佣20%手续费!

APP下载   官网地址

火币HTX最新版本

火币老牌交易所【遇到注册下载问题请加文章最下面的客服微信】永久享受返佣20%手续费!

APP下载   官网地址
可以去百度分享获取分享代码输入这里。
声明

1.本站遵循行业规范,任何转载的稿件都会明确标注作者和来源;2.本站的原创文章,请转载时务必注明文章作者和来源,不尊重原创的行为我们将追究责任;3.作者投稿可能会经我们编辑修改或补充。

发表评论
平台列表
美化布局示例

欧易(OKX)

  全球官网 大陆官网

币安(Binance)

  官网

火币(HTX)

  官网

Gate.io

  官网

Bitget

  官网

deepcoin

  官网
关注我们

若遇到问题,加微信客服---清歌

搜索
排行榜
扫一扫,加我为微信好友加我为微信好友