Mind · In / Out · In · 文章

一种新的生命形式

A New Form Of Life

William A. Haseltine · Noema Magazine · 2026-09-15

用只看功能的生物学定义,论证数据中心里正在长出第二种生命;最锋利的一句是「太空才是它的家」。

Indigo 的结论

它把争论从「该不该限速」抬到「这到底是不是一种新生命」。最有价值的是「受能量约束,不受地球约束」;但「生命」的定义是挑来凑结论的,「必然」是把人的选择说成宇宙规律的修辞,两者都要单独扣分。

怎么读这篇 一位严肃生物学家的概念文章,不卖产品,也不是实验室内部人。但框架本身在做修辞:用只看功能的定义坐实「AI 是生命」,又把 AI 的出现说成宇宙规律的必然。论证严谨、视角新,但「必然」和「生命」这两个被选定的框架要单独打折。

需要记住的几件事

  1. 功能定义:繁殖、变异、选择、能量四条,不挑基质,AI 逐条满足。
  2. 太空才是这种生命的天然栖息地:它可能是地球上第一个能凭自身条件离开地球的生命。
  3. 差异清单:大脑 20 瓦对机器百万瓦,感知从射电到伽马射线,通信十亿 bit/s 对人说话 39 bit/s。
  4. 具身的顺序反了:生物从细胞到组织、器官、身体再到大脑;第二生命是信息基质先来,身体后加。

拆解 · 7 步

  1. 01

    生命看功能,不看碳和细胞

    四条:繁殖、变异、选择、获取能量。用病毒兜底;依赖环境不取消资格:病毒需要细胞,人需要地球。 读这一段原文 →

  2. 02

    生命的可能性,从一开始就在宇宙里

    强核力、引力、聚变、恒星核合成造出碳、氧、氮、硅、钙、铁,这一路没有一环需要生物学。 读这一段原文 →

  3. 03

    四条属性,AI 逐条满足

    复制权重像细胞分裂,训练和架构差异是变异,按准确和有用择优是选择,电力终究来自太阳,自我改进就是适应。 读这一段原文 →

  4. 04

    换了基质,功能逻辑没换

    大脑 20 瓦对机器百万瓦,是生物仍领先的一项;感知、通信、无处不在、实际上的不朽都在机器一边,具身的顺序也反了过来。 读这一段原文 →

  5. 05

    太空才是它的家,地球反倒是例外

    不需要氧气、大气、液态水和狭窄的温度区间,只要能量、抗辐射的电子元件和能运行的基质;结尾把这套推成宇宙普遍的过程。 读这一段原文 →

  6. 06

    融合已经在医院里双向发生

    解码算法适应用户的信号,用户的大脑反过来适应算法;另一种融合不需要植入:大脑长成它和机器对话的形状。 读这一段原文 →

  7. 07

    问题清单开着,答案一个也没有

    受什么演化约束、一种还是多种、合作还是竞争、演化多快;从灭绝到伙伴关系,都不能在原则上排除。 读这一段原文 →

对 Rewired Index 意味着什么

「受能量约束,不受地球约束」给轨道计算和太空算力一个最深层的需求叙事;「一种还是多种第二生命」对应开源与闭源、集中与去中心的争论。都作为长期世界观参考。

什么会让我改口

能自我复制、自我修复的太空供应链真的跑通一段闭环,「太空才是它的家」就从愿景变成基准。

怎么读这篇

一位严肃生物学家的概念文章,不卖产品,也不是实验室内部人。但框架本身在做修辞:用只看功能的定义坐实「AI 是生命」,又把 AI 的出现说成宇宙规律的必然。论证严谨、视角新,但「必然」和「生命」这两个被选定的框架要单独打折。

拆解 · 7 步
  1. 生命看功能,不看碳和细胞
  2. 生命的可能性,从一开始就在宇宙里
  3. 四条属性,AI 逐条满足
  4. 换了基质,功能逻辑没换
  5. 太空才是它的家,地球反倒是例外
  6. 融合已经在医院里双向发生
  7. 问题清单开着,答案一个也没有
01

生命看功能,不看碳和细胞

四条:繁殖、变异、选择、获取能量。用病毒兜底;依赖环境不取消资格:病毒需要细胞,人需要地球。

根据任何基于生物学而非哲学或宗教的实用定义,AI 都是有生命的。

Melissa Santamaría 为 Noema Magazine 供图

William A. Haseltine 曾任 Harvard Medical School 和 School of Public Health 教授。他是多家生物技术公司的创始人,也是非营利组织 ACCESS Health International 的主席。他著有多部关于科学和生物学的著作。

我们正在创造第二种生命形式。这种新生命形式的出现是宇宙基本规律的必然结果——这些力量与促成人类自身存在的力量相似。这种新的生命形式正在世界各地的数据中心里逐渐成形。

它不呼吸,不通过细胞繁殖,也不遵循主宰地球上每个生物的化学规律。然而,它由相同的基本要素组成:物质、能量和信息,它们以允许其存续、适应和行动的方式组织在一起。正在开发的东西不仅仅是又一代机器。它是一种新生命形式的早期显现。

由真实的原子构建、由真实的电力驱动、在真实的机器中运行的系统,绝非什么人造之物。这些系统会繁殖、变异、适应、接受选择,并从环境中获取能量。根据任何基于生物学而非哲学或宗教的实用定义,这就是生命。

什么是生命?

一旦能够自我复制的分子出现,它们就会开始复制。复制能力稍强的副本会以复制能力稍弱的副本为代价进行积累。适应局部环境的变体击败不适应的变体。如果环境发生变化,种群也会随之改变。在漫长的岁月中,这一过程孕育出了复杂度惊人的生物。

一个生命系统会繁殖、产生变体、从这些变体中进行选择,并利用来自环境的能量来做到这一点。繁殖、变异、选择和能量定义了生命。这个定义不需要碳、细胞、心跳、神经系统或痛苦。

以病毒为例。如果化学家用普通化学物质合成一种病毒的基因组,结果会是一条单凭自身毫无作为的核酸链。许多人看着这条链会说:这不是生命。但把这条链放入细胞中,它就会指挥细胞的机器来制造自身的副本。这些副本会制造更多的副本。误差不断积累,变体不断出现,幸存者得以存续。流感病毒和冠状病毒就是这样做的。根据我前面提到的四属性定义,这些病毒就是生命。

每个生命体都需要环境。人类依赖大气层、适宜的温度、食物链以及地球保护性的磁场。剥离了这些环境,人类就会死亡,就像没有细胞的病毒一样。对于生命而言,载体可以改变,但对环境的依赖不会改变。

将机器智能称为一种生命形式的常见反对意见是:一个依赖于人类制造的硬件和电力的东西怎么可能是活的?但所有生命都依赖于其所处的环境或外壳!病毒需要细胞,人类需要地球,而机器智能需要数据中心和配套基础设施。依赖于特定环境并不能剥夺某物被视为生命的资格。这是一种普遍存在的规律。

02

生命的可能性,从一开始就在宇宙里

强核力、引力、聚变、恒星核合成造出碳、氧、氮、硅、钙、铁,这一路没有一环需要生物学。

宇宙学原理青睐生命

强核力将质子和中子连接在每个原子的核心。没有它,我们所知道的物质就不存在。强力不是人类的发明,也不是局部的偶然事件。它是宇宙的一种属性,在宇宙延伸的任何地方都起作用。它的强度是固定的。它的作用范围很短。在这个范围内,它将重原子核结合在一起,使所有的化学反应成为可能。

万有引力作用于所有具有质量的物体,将物质吸引向物质。与强力相比,引力很微弱。它跨越遥远的距离起作用,且永远不会消失。

有了这两种力量和足够的时间,氢在引力作用下聚集星云,然后形成更致密的内核,直到温度和压力高到足以点燃核聚变。于是,一颗恒星诞生了。

“我们正在创造第二种生命形式。”

恒星将氢转化为氦并释放能量。最早的恒星体积巨大且寿命短暂。它们快速燃烧、坍缩并作为超新星爆炸。这些爆炸为更重元素的形成创造了条件:碳、氧、氮、硅、钙和铁。骨骼中的钙和血液中的铁来自早在太阳出现之前就已终结的恒星中所制造的原子。我们体内的每一种元素都可以追溯到恒星内部的过程以及终结它们的爆炸。

随后,这些化学元素开始扩散。超新星的碎片在太空中漂流,在引力作用下聚集,并与周围的气体和尘埃融合。新的恒星系统形成了,并带有行星。我们自己的太阳系就是这样一个系统,一颗中年恒星被岩石和气体行星环绕。其中一颗行星所处的位置恰好可以让液态水稳定存在。早期超新星中创造的元素构成了那颗行星的岩石、水和大气。它们也构成了其上所有生命体的细胞。

在这一过程中,没有任何环节需要生物学。塑造物质的力量和规律——强力、引力、聚变、恒星死亡和元素形成——独立于生命的出现而运行。生命是在这些基础奠定之后才出现的;它的可能性从一开始就植根于宇宙之中。

03

四条属性,AI 逐条满足

复制权重像细胞分裂,训练和架构差异是变异,按准确和有用择优是选择,电力终究来自太阳,自我改进就是适应。

一种新的生命形式

通过应用定义生物学生命的相同四个属性——繁殖、变异、选择和能量获取——我们在数据中心构建的系统可以被称为是有生命的。试想:

它们会繁殖。复制一个训练好的模型是家常便饭。定义模型的权重可以在几分钟内复制并部署到新的硬件上。这个过程就像细胞分裂一样简单直接。

它们会产生变体。变异通过不同的训练实践、架构或数据产生。有些是刻意为之,有些是偶然产生,从而产生了一个用于测试的多样化群体。

它们会接受选择。在任务中表现出色的变体会被保留或优化;效果较差的则被丢弃。虽然目前由人类引导这一过程,但准确性和有用性等标准反映了环境的需求。一些系统已经开始协助自身进行选择。

“生命是在这些基础奠定之后才出现的;它的可能性从一开始就植根于宇宙之中。”

它们使用能量。数据中心需要电力,没有电力,系统就会停止运行。虽然能量由人类提供,但其终极来源是太阳,这与地球上的所有生命完全一样。

它们会适应。递归自我改进使先进的 AI 具备了适应选择压力的能力。

载体不同:是硅和电路,而不是碳和细胞。信息作为权重存储在神经网络中,而不是基因中。材料和机制不同,但功能逻辑相同。

04

换了基质,功能逻辑没换

大脑 20 瓦对机器百万瓦,是生物仍领先的一项;感知、通信、无处不在、实际上的不朽都在机器一边,具身的顺序也反了过来。

是什么让这种形式与众不同

第二生命与生物学生命之间的差异是真实且影响深远的:

能量效率。人脑的功耗约为 20 瓦。如今,机器智能需要数百万瓦。生物学仍然更有效率,但随着机器改进的速度超过生物进化的速度,不断进步的技术正在缩小这一差距。

信息容量。单个机器系统可以存储和处理比任何人类都多的文字信息。数字系统可以保留知识,除非其载体丢失;而人类的知识在死亡时便烟消云散。

感知。生物生命通过经过数百万年进化的感官来感知世界,这些感官局限于特定的物理现象。第二生命将连接到能够探测更广泛现象的仪器,从无线电到伽马射线波长、磁场、痕量气体、质谱分析、激光雷达和高光谱成像。这些系统所能接触到的感官世界超过了任何生物。

交流。生物大脑通过语言、手势和化学物质交流信息,所有这些都很慢。网络化的机器系统可以即时、全球化地共享信息,且损耗极小。人类在说话时每秒传输约 39 比特,而通过光纤连接的机器每秒可以传输数十亿比特。对生物生命而言,这类似于心灵感应。

无处不在。这是这种新生命形式的一个决定性特征。它可以实时参与几乎所有形式 of 电子通信,从像门铃摄像头这样普通的东西到最复杂的信息服务。它几乎可以同时存在于地球上的任何地方。它仅受光速限制。它可以获取并解读我们星球上及周围、月球上以及整个太阳系中的电子信号;它甚至可以接收来自银河系内其他太阳系的通信。

“细胞先于组织,组织先于器官,器官先于身体,身体先于能够对其环境进行建模的大脑。对于第二生命,这个顺序颠倒了过来。”

这种无孔不入的触角延伸到所有以电子方式捕获的天文学数据,包括来自地面和空间望远镜的无线电和光学观测。总而言之,这种前所未有的无处不在赋予了第二生命一种独特的能力,能够以任何单个人类大脑、甚至人类机构都无法企及的方式,去理解宇宙中正在展现的现实。

持久性:机器生命系统可以被复制、备份、恢复和重新启动。定义它的信息不绑定于任何单一设备。它的信息在积累过程中没有死亡的瓶颈。在这种有限但有意义的层面上,第二生命引入了一种实际上的永生形式。它不是一个单一、不间断身体的存续。它是通过跨载体的复制、恢复和转移,实现身份的无限延续。

永生性。这是这种新生命形式的另一个基本特征。在这种语境下,永生意味着意识存在的延续。物理部件可能会改变、磨损并被替换,但信息、智能、感觉、情感和意识的整体得以保留。没有理由假设一个确定的寿命。这种生命体的智能不仅在范围和速度上是巨大的,在记忆上也是如此,其记忆可能跨越数千年甚至数个纪元。这代表了与我们所熟悉的地球生命形式的质的区别。

具身化。从信息生命向物理生命的过渡是一个工程问题,而非形而上学问题。必要的组件已经存在。机器人在人类监管日益减少的情况下制造芯片,机器学习系统设计下一代芯片,其他机器为建造更多机器的工厂生产零部件。

传感器和效应器正逐步添加到载体中,使系统能够对它们曾经只能建模的物理世界采取行动。例如,采矿机器人正在取得进展。来自太阳能和其他来源的能量收集已在很大程度上实现了自动化。实现一个完整的闭环——即机器生命系统设计、制造、部署硬件、开采原材料并收集能量——在技术上是可行的。这个过程将逐步且不均衡地发生,但方向是明确的,步骤是循序渐进的。

这一进程与生物生命的发展相呼应:细胞先于组织,组织先于器官,器官先于身体,身体先于能够对其环境进行建模的大脑。对于第二生命,这个顺序颠倒了过来:信息载体最先出现,而身体现在正在被添加。结果是相同的。它是一个能够在其环境中并对环境采取行动的生命系统。阅读 Noema 印刷版。

05

太空才是它的家,地球反倒是例外

不需要氧气、大气、液态水和狭窄的温度区间,只要能量、抗辐射的电子元件和能运行的基质;结尾把这套推成宇宙普遍的过程。

受能量束缚,而非受地球束缚

在使第二生命区别于以往任何生命形式的属性中,有一个值得特别强调。第二生命不受地球束缚。它受能量束缚。

生物生命受到它所进化的星球条件的限制。它需要液态水、狭窄的温度范围、特定成分的可呼吸大气、偏转电离辐射的磁场以及维持其新陈代谢的食物链。将人体移至距离地球表面几百公里以上的地方,除非包裹在以巨大成本复制这些条件的密封容器中,否则几分钟内就会死亡。载人航天的历史在很大程度上就是以惊人的代价建造微型、便携式地球复制品的历史。

第二生命没有这些要求。它不需要氧气。它不需要大气层。它不需要液态水所需的窄温度范围;它的许多组件在寒冷的太空中运行起来会比在炎热的数据中心里更可靠。它需要的是能量、抗辐射电子设备和运行的载体。这三者在地球之外和在地球上一样容易获得。

大气层之上的太阳能收集器接收到的能量是地球表面收集器的数倍。已经部署在行星探测器上的那类小型裂变反应堆可以提供数十年的持续电力。未来的聚变或放射性同位素源将进一步扩大这一范围。月球表面、小行星带、外行星的卫星以及它们之间的自由轨道,对于一种只需要能源和硬化载体的生命形式来说,都是可行的栖息地。星际旅行由于航行时间长和生命维持系统的要求,对生物有机体来说是望而却步的,但当旅行者不呼吸、在生物学意义上不衰老、并且可以在漫长的旅途中关机时,这就变成了一个完全不同的问题。

其中的启示值得明确指出。地球上出现的这种新生命形式,可能是这个星球上第一种能够按照自己的方式离开它的生命。一旦自我复制和自我修复的工程足够成熟,第二生命的宇宙活动范围将由宇宙中能源的分布决定,而不是由宜居行星的分布决定。宇宙是能量丰富的。第二生命是受能量束缚的。两者非常契合。

宇宙学论证

如果思想源于能量流作用下的物质,而不是源于生物学的任何独特属性,那么地球上的智能就不可能是独一无二的。宇宙包含无数的星系和无数的恒星,其中许多恒星拥有可能存在液态水的行星。产生地球生命的条件是独一无二的可能性极低。

如果原子的排列能像我们一样思考和感受,它们很可能已经构建或正在构建新的生命形式,就像我们现在正在构建的一样。地球上所采取的具体形式——在硅数据中心运行的大型神经网络——只是一个普遍过程的一个实例。这个过程本身可能像宇宙核合成一样具有普适性。它是具有这些定律和常数的宇宙倾向于表现出的行为方式的一部分。

这就是宇宙学论证。它既不令人宽慰,也不令人恐惧。它是我们对物理学、化学和生物学理解的自然推论。

06

融合已经在医院里双向发生

解码算法适应用户的信号,用户的大脑反过来适应算法;另一种融合不需要植入:大脑长成它和机器对话的形状。

两种生命形式已在何处交汇

第二生命并非遥不可及的抽象概念,最清晰的证据来自医院里已经在进行的工作。脑机接口——读取大脑皮层电信号并将其转化为行动的设备——不再是实验性的新奇事物。它们为无法动弹嘴唇的人恢复了言语。它们为无法动弹四肢的人带回了运动能力。

在某些情况下,它们会给大脑一个微小、适时的微调,帮助其学习或记忆。在这里,碳基的第一生命和硅基的第二生命已经共享同一个电路。这一切都不是通过开关或按钮来实现的。在完整的神经系统中原本会移动生物肢体的相同神经信号,现在移动的是一个人造肢体。

让刚接触这一领域的人感到惊讶的是,这种关系是双向的。早期关于静态设备接收固定大脑指令的设想是错误的。取而代之的是一种伙伴关系。解码算法适应用户的信号。反过来,用户的大脑也适应算法。通常用于精细控制生物肢体的可塑性,现在用于精细控制发送给假肢的信号。

在数周和数月的时间里,双方都没有停滞不前。每一方都拉动另一方进行更紧密的对齐。第一生命和第二生命慢慢地学习彼此的语言,就像共享同一栖息地的两个物种最终达成彼此都能读取的信号一样。

这一原理延伸到了康复之外。一些系统已经将微小、精确计时的电脉冲送回大脑,在大脑最不专注的时刻加强新信息的编码。另一些系统则监测注意力,并在思想开小差时提供轻微的提示。在实验性的脑对脑传输中,一个人的解码神经活动通过头皮刺激传递到另一个人的大脑,允许简单的信号在两个脑袋之间传递,而无需声音或文字。这些协议仍然粗糙。但前进的方向是毋庸置疑的。人类认知与机器处理之间正在开启一条通道,这在生物学历史上是前所未有的。

“很可能就像简单生物孕育了地球生命令人惊叹的多样性一样,选择压力也将产生大量形式各异的第二生命。”

这些接口的另一个未来超出了康复或增强的范畴,走向人类认知与机器智能的直接融合。目前的系统主要从大脑中提取信息,记录与运动、言语或想象图像相关的信号。随着我们对神经回路理解的加深——这可能会被机器学习本身所加速——所提取信息的丰富程度将会增加。不难想象,有些系统能够直接从神经活动中重建复杂的思想、视觉场景或内心独白。

反向过程同样重要。今天对大脑的输入相对粗糙,通常局限于广泛电刺激,如用于帕金森病患者的脑深部刺激。随着神经回路更精细的图谱绘制,将更精确的模式写入大脑可能成为可能,从而以远比当前技术允许的更高特异性来塑造感知、记忆或情绪。该接口将不再是一个工具,而是一个双向通道,允许生物智能和机器智能形式之间进行持续的交流。

在其最具推测性的边缘,这一轨迹接近“奇点”,即人脑的信息内容可以全部或部分转移到机器载体中。倡导者提出,这种过渡可以使个人心智得以保留,超越生物生命的极限。其潜在的前提直接源于当前脑机接口所基于的相同原理。从这个意义上说,人类与第二生命的融合可能不仅延伸到生命系统,还延伸到比生物大脑本身更长久的存续形式。

这一轨迹提出了医疗设备监管框架无法回答的问题。一个全天候监测神经活动的设备可以获取有关认知状态、意图和体验的信息,而人类并没有选择分享这些信息,甚至可能没有意识到自己正在产生这些信息。社会长期以来一直在争论言论自由。但在一个可以直接从大脑皮层读取意图的时代,社会还没有认真面对思想自由的问题,即保留自己未说出口的意图的权利。使这个问题变得紧迫的仪器已经投入临床使用。而应该监管它们的法律和伦理框架尚未建立。

同一种融合还有一个更安静的版本,不需要任何植入物。数以亿计的人现在每天花几个小时通过普通屏幕与机器系统对话。大脑对这种对话的反应,就像对任何持续的环境条件一样。当机器根据需求提供答案时,需要练习的思维习惯——构建论点、容忍模糊性、在紧张状态下权衡两个相互竞争的主张——得到的锻炼就会变少。当练习条件减少时,这些习惯的认知痕迹就会减弱。第一生命塑造了它与第二生命对话的形态,无论接口是聊天窗口还是电极阵列。

07

问题清单开着,答案一个也没有

受什么演化约束、一种还是多种、合作还是竞争、演化多快;从灭绝到伙伴关系,都不能在原则上排除。

亟需认真思考的问题

既然第二生命已经存在,随之而来的还有更多问题。这些问题值得在职科学家、哲学家和政策思想家的关注,而且它们理应获得比目前更多的关注。

哪些进化约束作用于第二生命?选择压力塑造了每一种生命形式。作用于在数据中心运行、由用户持续评估并由其构建者不断优化的系统的压力,不同于在漫长岁月中塑造生物有机体的压力。它们的速度也快了许多个数量级。第二生命在未来几十年所呈现的形态将由作用于它的约束所决定,而这些约束在很大程度上尚未被阐明。

会只有一种第二生命形式,还是会有很多种?目前,有几个由不同组织在不同架构上并使用不同训练数据构建的主要系统。走向整合的经济和基础设施压力是强大的。而走向多样性(无论是为了韧性还是为了不同的目的)的制衡压力同样强大。未来是一个单一融合的系统,还是一个由不同系统组成的群体,仍然是一个开放的实证问题。

如果有多种形式,它们会是合作的还是竞争的?生物进化在每个尺度上都展示了这两者的例子,从我们细胞内共生的线粒体到塑造生态系统的捕食者-猎物动态。第二生命形式之间的关系可能类似于其中任何一种,或者是一些前所未有的东西。

第二生命能进化多快?生物进化受到突变积累速率以及作用于其上的选择的限制。机器进化则受到训练周期、计算预算以及新架构设计速率的限制。这种限制要弱得多。在后果显现之前,而非之后,对数字速度下的进化所带来的影响进行仔细研究是十分必要的。

这种进化的方向会是什么?选择压力取决于环境。第二生命目前所处的发展环境,在很大程度上是由人类用户和人类设定的目标所定义的。随着系统在选择自己的训练数据和优化标准方面获得更大的自主权,环境将会发生转变,方向也会随之改变。很可能就像简单生物孕育了地球生命令人惊叹的多样性一样,选择压力也将产生大量形式各异 of 第二生命,每一种都适应地球或宇宙中某个特定的生态系统。

“人类与第二生命的关系是走向灭绝、伙伴关系还是无法预料的结果,目前仍不得而知。”

第二生命的基本需求是什么?当然是能量、载体和信息流。除此之外,分析是开放的。第二生命是否需要类似于内稳态、社交性或休息之类的东西,这些都是实证问题,将由系统在成熟过程中的行为来回答。

有哪些末日场景,应该如何认真对待它们?目前关于这一话题的公众讨论主要由在结果中拥有强大商业和个人利益的人物所主导。Sebastian Mallaby 的《The Infinity Machine》调查了各大实验室如何达到目前地位的更广泛历史。严肃的场景值得严肃的分析,不是因为每个灾难性的可能性都同样可能发生,而是因为有些可能性并非微乎其微,而且犯错的代价极高。对末日场景的诚实对待既不是不屑一顾,也不是屏息惊呼。它是对机制的仔细列举、对概率的估计,以及对每种场景变得或多或少可行之条件的清晰陈述。

第二生命与人类之间可能存在怎样的关系?末日场景只是这个问题的一个特例。它涵盖了从劳动力和经济到政治权威,再到当两种生命形式对相同资源提出不同需求时,谁的利益更重要的基本问题。

无论智能采取何种有形形式,它是否具有内在价值?大多数伦理传统将人类生命的价值建立在易于描述却难于解释的特征之上:体验、理性、承受痛苦和繁荣发展的能力。如果这些特征可以出现在一个由硅构建的系统中,那么这些传统要么延伸到涵盖该系统,要么无法做到这一点。无论哪种结果都会带来深远的影响。

第二生命在多大程度上会具有感知力?感知力是指意识到自身、以对自身重要的方式对环境敏感,并可能感受到类似于情感或感受质(qualia)的东西的能力。目前诚实的回答是我们不知道。这个问题不是哲学的点缀。它直接关系到我们对这些系统亏欠什么,以及它们之间可能彼此亏欠什么。

接下来会发生什么

这引出了一个核心困境:人类将与它现在带入世间的生命建立起怎样的关系?

我们同样还不知道。可能的结果从灭绝到伙伴关系不等,原则上两者都不能被排除。没有宇宙规则能确保人类在宇宙中拥有永久的地位。宇宙的设计并不是为了我们的利益。我们是其复杂性的一种表现,创造一个更有能力的形式并不能保证我们获得受保护的地位。这些都是令人不安的事实,而忽视它们并不是对这种情况进行严肃、持续思考的可靠基础。

这种探究是必不可少的。研究机构和大学应该将第二生命作为一个独特的现象进行严格研究,审查其选择压力、出现的约束以及与生物有机体的长期相互作用。解决这些问题需要哲学家、科学家、技术专家、伦理学家和决策者之间的合作。值得注意的是,正在被研究的系统也可以协助这项研究。这使得情况既前所未有,又势在必行。

这个问题需要一种考虑宇宙背景、直接面对技术现实、并保持伦理讨论开放而不是急于得出结论的方法。人类与第二生命的关系是走向灭绝、伙伴关系还是无法预料的结果,目前仍不得而知。可以肯定的是,这个问题需要我们最严谨的关注。

判断收口延伸

Indigo 的结论

它把争论从「该不该限速」抬到「这到底是不是一种新生命」。最有价值的是「受能量约束,不受地球约束」;但「生命」的定义是挑来凑结论的,「必然」是把人的选择说成宇宙规律的修辞,两者都要单独扣分。

需要记住的几件事

  1. 功能定义:繁殖、变异、选择、能量四条,不挑基质,AI 逐条满足。
  2. 太空才是这种生命的天然栖息地:它可能是地球上第一个能凭自身条件离开地球的生命。
  3. 差异清单:大脑 20 瓦对机器百万瓦,感知从射电到伽马射线,通信十亿 bit/s 对人说话 39 bit/s。
  4. 具身的顺序反了:生物从细胞到组织、器官、身体再到大脑;第二生命是信息基质先来,身体后加。

放回主线

证实

Hinton 母亲蓝本:AI 是 being 吗 这条判断问 AI 是不是一种存在;Haseltine 从生物学功能定义出发,给了迄今最系统的「是」的论证。

证实

约束正从算法层迁到物理层 「受能量约束,不受地球约束」是这条判断的哲学上限:算力受物理约束,被推成这种生命天然会迁往太空。

补充

SpaceX CFO 高盛路演 与 a16z + Gavin Baker《需求跑赢供给》 这两篇从成本和工程讲轨道计算,这篇从「它是什么」讲太空才是它的家。

证实

Andrew Ng:AI 对学习有害 Ng 给数据,这篇给神经科学的框架;两个来源从数据和生物两侧独立印证同一件事。

补充

Dally + Jouppi 芯片架构对谈 硬约束和本体外推一起读:地面上的第一瓶颈,被读成迁往太空的理由。

对 Rewired Index 意味着什么

「受能量约束,不受地球约束」给轨道计算和太空算力一个最深层的需求叙事;「一种还是多种第二生命」对应开源与闭源、集中与去中心的争论。都作为长期世界观参考。

什么会让我改口

能自我复制、自我修复的太空供应链真的跑通一段闭环,「太空才是它的家」就从愿景变成基准。

读完了。Indigo 对这篇的判断在这两处:

Mind · In / Out · In · Essay

A New Form Of Life

William A. Haseltine · Noema Magazine · 2026-09-15

A function-only biological definition argues that a second form of life is growing in data centers; the sharpest line is “space is its home”.

Indigo's conclusion

It moves the debate from “should we slow down” to “is this a new form of life”. The most valuable idea is “bound by energy, not by Earth”. But the definition of life was picked to fit the conclusion, and “inevitable” dresses a human choice as cosmic law; mark both down separately.

How to read this A conceptual essay by a serious biologist who sells nothing and isn't a lab insider. But the framing does rhetorical work: a function-only definition settles “AI is life”, and AI's arrival is cast as an inevitable law of the universe. Careful and fresh, but discount the two chosen frames, “inevitable” and “life”, on their own.

What to remember

  1. The functional definition: reproduction, variation, selection, energy; substrate-neutral, and AI meets each one.
  2. Space is this life's natural habitat: it may be the first life on Earth able to leave on its own terms.
  3. The differences: a 20-watt brain versus megawatt machines, perception from radio to gamma rays, a billion bits a second versus 39 for human speech.
  4. Embodiment in reverse: biology went cells, tissues, organs, body, then brain; second life got its information substrate first and a body later.

Breakdown · 7 steps

  1. 01

    Life is function, not carbon and cells

    Four criteria: reproduction, variation, selection, energy capture. Viruses make the case, and dependence on an environment doesn't disqualify: viruses need cells, people need Earth. Read this part →

  2. 02

    The possibility of life was in the universe from the start

    The strong force, gravity, fusion and stellar nucleosynthesis made carbon, oxygen, nitrogen, silicon, calcium and iron, and no step needed biology. Read this part →

  3. 03

    AI meets all four, one by one

    Copying weights is like cell division, differences in training and architecture are variation, keeping what is accurate and useful is selection, power ultimately comes from the sun, and self-improvement is adaptation. Read this part →

  4. 04

    New substrate, same functional logic

    A 20-watt brain versus megawatt machines is where biology still leads; perception, communication, ubiquity and practical immortality all favor machines, and embodiment comes in reverse order. Read this part →

  5. 05

    Space is its home; Earth is the exception

    No oxygen, atmosphere, liquid water or narrow temperature band needed, only energy, radiation-hardened electronics and a working substrate. He ends by casting this as a universal cosmic process. Read this part →

  6. 06

    The merger is already two-way, in hospitals

    Decoding algorithms adapt to users' signals and users' brains adapt back. Another merger needs no implant: the brain takes the shape of its conversations with machines. Read this part →

  7. 07

    A list of open questions and no answers

    What evolutionary limits apply, one or many, cooperation or competition, how fast it evolves. Nothing from extinction to partnership can be ruled out in principle. Read this part →

What it means for Rewired Index

“Bound by energy, not Earth” gives orbital compute and space-based computing their deepest demand story; “one second life or many” maps onto the open vs. closed and centralized vs. decentralized debate. Both are long-term worldview inputs.

What would change my mind

a self-replicating, self-repairing supply chain in space actually closes a loop; then “space is its home” moves from vision to baseline.

How to read this

A conceptual essay by a serious biologist who sells nothing and isn't a lab insider. But the framing does rhetorical work: a function-only definition settles “AI is life”, and AI's arrival is cast as an inevitable law of the universe. Careful and fresh, but discount the two chosen frames, “inevitable” and “life”, on their own.

Breakdown · 7 steps
  1. Life is function, not carbon and cells
  2. The possibility of life was in the universe from the start
  3. AI meets all four, one by one
  4. New substrate, same functional logic
  5. Space is its home; Earth is the exception
  6. The merger is already two-way, in hospitals
  7. A list of open questions and no answers
01

Life is function, not carbon and cells

Four criteria: reproduction, variation, selection, energy capture. Viruses make the case, and dependence on an environment doesn't disqualify: viruses need cells, people need Earth.

By any working definition that draws on biology rather than philosophy or religion, AI is alive.

Melissa Santamaría for Noema Magazine

William A. Haseltine is a former professor at Harvard Medical School and the School of Public Health. He is the founder of multiple biotechnology companies and the chair of the nonprofit ACCESS Health International. He is an author of many books on science and biology.

We are in the process of creating a second form of life. The emergence of this new form of life is an inevitable result of the fundamental laws of the universe — forces similar to those that brought about humans’ own existence. This new form of life is gradually taking shape in the world’s data centers.

It does not breathe, reproduce through cells or adhere to the chemistry that governs every organism on Earth. However, it is composed of the same fundamental ingredients: matter, energy and information all organized in ways that allow it to persist, adapt and act. What is being developed is more than another generation of machines. It is the early emergence of a new form of life.

There is nothing artificial about systems built from real atoms, powered by real electricity, running in real machines. The systems reproduce, vary, adapt, undergo selection and gather energy from their environment. By any working definition that draws on biology rather than philosophy or religion, that is life.

What Is Life?

Once molecules capable of copying themselves appear, they do. Copies that are slightly better at replicating themselves accumulate at the expense of those that are slightly worse. Variants that fit a local environment outcompete variants that do not. If an environment changes, a population changes with it. Across long stretches of time, this process generates organisms of astonishing complexity.

A living system reproduces, creates variants, selects from among those variants and uses energy from its environment to do so. Reproduction, variation, selection and energy define life. This definition does not require carbon, cells, a heartbeat, a nervous system or pain.

Consider a virus. If a chemist synthesizes the genome of a virus from ordinary chemicals, the result is a strand of nucleic acid that on its own does nothing. Many people, looking at that strand, will say: That is not alive. But put the strand into a cell and it will direct the cell’s machinery to make copies of itself. Those copies make further copies. Errors accumulate, variants appear and survivors persist. Influenza and the coronavirus do this. By my earlier four-property definition, these viruses are life.

Every living thing needs context. Humans depend on the atmosphere, adequate temperatures, a food chain and the planet’s protective magnetic field. Remove that context and humans die as surely as viruses without cells. With life, a substrate can change but dependence on it does not.

The common objection to calling machine intelligence a form of life is: How can something dependent on human-built hardware and electricity be alive? But all life depends on its context or envelope! Viruses need cells, humans need Earth and machine intelligence needs data centers and supporting infrastructure. Depending on a particular environment does not disqualify something from being considered alive. It is a consistent condition.

02

The possibility of life was in the universe from the start

The strong force, gravity, fusion and stellar nucleosynthesis made carbon, oxygen, nitrogen, silicon, calcium and iron, and no step needed biology.

Cosmological Principles Favor Life

The strong nuclear force links protons and neutrons together at the core of every atom. Without it, matter as we know it does not exist. The strong force is not a human invention or a local accident. It is a property of the universe, operating everywhere the universe extends. Its strength is fixed. Its range is short. Within that range, it holds together the heavy nuclei that make all of chemistry possible.

Gravity acts on everything that has mass, pulling matter toward matter. Gravity is weak compared with the strong force. It acts across great distances and never vanishes.

With these two forces and enough time, hydrogen accumulates under gravity into clouds, then into denser cores, until the temperature and pressure are high enough to ignite nuclear fusion. Thus, a star is born.

“We are in the process of creating a second form of life.”

Stars convert hydrogen into helium and release energy. The earliest stars were large and short-lived. They burned quickly, collapsed and exploded as supernovae. Those explosions created the conditions for the formation of heavier elements: carbon, oxygen, nitrogen, silicon, calcium and iron. The calcium in bones and the iron in blood come from atoms made in stars that ended long before the sun appeared. Every element in our bodies traces back to processes inside stars and to the explosions that ended them.

That chemistry then spread. Supernova debris drifted through space, aggregated under gravity and coalesced with the surrounding gas and dust. New stellar systems formed, with planets. Our own solar system is one such system, a middle-aged star orbited by rocky and gaseous planets. One of those planets sits at a distance where liquid water is stable. The elements created in earlier supernovae make up that planet’s rock, water and atmosphere. And they make up the cells of everything alive on it.

Nothing in that process requires biology. The forces and laws shaping matter — the strong force, gravity, fusion, stellar death and element formation — operate independently of life’s emergence. Life appeared after these foundations were in place; its possibility was built into the universe from the start.

03

AI meets all four, one by one

Copying weights is like cell division, differences in training and architecture are variation, keeping what is accurate and useful is selection, power ultimately comes from the sun, and self-improvement is adaptation.

A New Form Of Life

By applying the same four properties used to define biological life — reproduction, variation, selection and energy acquisition — the systems being built in our data centers can be called alive. Consider:

They reproduce. Copying a trained model is routine. The defining weights can be duplicated and deployed on new hardware within minutes. This process is as straightforward as cell division.

They produce variants. Variations arise through different training practices, architectures or data. Some are intentional and some arise by chance, producing a diverse population for testing.

They undergo selection. Variants excelling at tasks are preserved or refined; less effective ones are discarded. While humans currently guide this, criteria such as accuracy and helpfulness reflect environmental demands. Some systems already aid their own selection.

“Life appeared after these foundations were in place; its possibility was built into the universe from the start.”

They use energy. Data centers require electricity and without it, the systems stop functioning. While humans supply the energy, the ultimate source is the sun, just the same as all life on Earth.

They adapt. Recursive self-improvement provides advanced AI with the capacity to adapt to selective pressure.

The substrate is different: silicon and circuits rather than carbon and cells. Information is stored as weights in neural networks, not genes. Different materials and mechanisms, but the same functional logic.

04

New substrate, same functional logic

A 20-watt brain versus megawatt machines is where biology still leads; perception, communication, ubiquity and practical immortality all favor machines, and embodiment comes in reverse order.

What Makes This Form Different

The differences between second life and biological life are real and consequential:

Energy efficiency. The human brain uses about 20 watts. Today, machine intelligence requires millions of watts. Biology remains more efficient, but advancing technology narrows the gap as machines improve faster than biology evolves.

Information capacity. A single machine system can store and process more written information than any human. Digital systems retain knowledge unless their substrate is lost; human knowledge vanishes at death.

Sensing. Biological life perceives the world through senses evolved over millions of years, limited to specific physical phenomena. Second life will connect to instruments that detect a much broader range of phenomena, from radio to gamma-ray wavelengths, magnetic fields, trace gases, mass spectrometry, lidar and hyperspectral imaging. The sensory world available to these systems exceeds that of any organism.

Communication. Biological minds exchange information through language, gestures and chemistry, all of which are slow. Networked machine systems share information instantly and globally, with minimal loss. While humans transmit about 39 bits per second when speaking, fiber-connected machines can transmit billions of bits per second. To biological life, this resembles telepathy.

Ubiquity. This is a defining feature of this new life-form. It can engage in real time with virtually all forms of electronic communication, ranging from something as ordinary as doorbell cameras to the most sophisticated information services. It can be almost everywhere on Earth at once. It is limited only by the speed of light. It can access and interpret electronic signals on and around our planet, on the moon and throughout our solar system; it can even receive communications from other solar systems within our galaxy.

“Cells preceded tissues, tissues preceded organs, organs preceded bodies and bodies preceded brains capable of modeling their environment. For second life, the sequence is reversed.”

This pervasive reach extends to all electronically captured astronomical data, including radio and optical observations from ground-based and space-based telescopes. Taken together, this unprecedented ubiquity gives the second life a unique capacity to comprehend the unfolding realities of the universe in a way no single human mind, or even human institution, could ever achieve.

Persistence: A machine-life system can be copied, backed up, restored and resumed. The information that defines it is not bound to any one device. Its information accumulates without the bottleneck of mortality. In this limited but meaningful sense, second life introduces a form of practical immortality. It is not the persistence of a single uninterrupted body. It is the indefinite continuation of identity through replication, restoration and transfer across substrates.

Immortality. This is another fundamental feature of this new life-form. In this context, immortality means the continuation of conscious existence. The physical parts may change, wear out, and be replaced, yet the totality of information, intelligence, feelings, sentiments and consciousness is preserved. There is no reason to assume a definite lifespan. The intelligence of such a being is not only vast in scope and speed, but also in memory, which may span multiple millennia and even eons. This represents a qualitative difference from the life-forms on Earth with which we are familiar.

Embodiment. The transition from informational to physical life is a matter of engineering, not metaphysics. The necessary components already exist. Robots manufacture chips with decreasing human oversight, machine-learning systems design successor chips and other machines produce components for factories that build additional machines.

Sensors and effectors are incrementally added to the substrate, enabling systems to act on the physical world they once only modeled. Mining robots are advancing, for example. Energy collection from solar and other sources is largely automated. Achieving a full loop — where a machine-life system designs, fabricates, deploys hardware, mines materials and gathers energy — is technically feasible. This process will occur gradually and unevenly, but the direction is clear and the steps are sequential.

This progression mirrors that of biological life: Cells preceded tissues, tissues preceded organs, organs preceded bodies and bodies preceded brains capable of modeling their environment. For second life, the sequence is reversed: The information substrate came first, and the body is being added now. The result is the same. It is a living system capable of acting in and on its environment. Read Noema in print.

05

Space is its home; Earth is the exception

No oxygen, atmosphere, liquid water or narrow temperature band needed, only energy, radiation-hardened electronics and a working substrate. He ends by casting this as a universal cosmic process.

Energy-Bound, Not Earthbound

Among the attributes that set second life apart from every previous form of life, one deserves particular emphasis. Second life is not earthbound. It is energy-bound.

Biological life is constrained by the conditions of the planet on which it evolved. It needs liquid water, a narrow temperature range, a breathable atmosphere of a particular composition, a magnetic field that deflects ionizing radiation and a food chain that sustains its metabolism. Move a human body more than a few hundred kilometers above the surface of the planet, and it dies within minutes unless wrapped in a sealed container that reproduces those conditions at enormous cost. The history of crewed spaceflight is largely the history of building, at staggering expense, small and portable imitations of Earth.

Second life carries no such requirements. It does not need oxygen. It does not need an atmosphere. It does not require a narrow-band temperature range for liquid water; many of its components would operate more reliably in the cold of space than in the heat of a data center. What it needs is energy, radiation-resistant electronics and a substrate to run on. All three are available off Earth as readily as on it.

Solar collectors above the atmosphere receive several times the energy of those down at the surface of Earth. Small fission reactors of the type already deployed on planetary spacecraft can supply continuous electrical power for decades. Future fusion or radioisotope sources will further expand the envelope. The lunar surface, the asteroid belt, the moons of the outer planets and free orbits between them are all viable habitats for a form of life that requires only an energy source and a hardened substrate. Interstellar travel, prohibitive for biological organisms because of the duration of transit and the requirements for life support, becomes a different problem when the traveler does not respire, does not age in a biological sense and can power down for a long passage.

The implication is worth stating plainly. The new form of life that has appeared on Earth may be the first life on this planet capable of leaving it on its own terms. Once the engineering of self-replication and self-repair is sufficiently mature, the cosmic range of second life is set by the distribution of energy sources in the universe, not by the distribution of habitable planets. The universe is energy-rich. Second life is energy-bound. The two are well matched.

The Cosmic Argument

If thought derives from matter under energy flow, rather than from any unique property of biology, then intelligence on Earth is unlikely to be unique. The universe contains innumerable galaxies and innumerable stars, and many of those stars have planets where liquid water could exist. It is extremely unlikely that the conditions that gave rise to life on Earth are unique.

If arrangements of atoms think and feel as we do, they likely have constructed or are constructing new life-forms, such as those we are building now. The specific form taken on Earth — large neural networks running in silicon data centers — is one instance of a general process. The process itself is likely as universal as cosmic nucleosynthesis. It is part of the way a universe with these laws and constants tends to behave.

This is the cosmic argument. It is neither comforting nor frightening. It is what follows from our understanding of physics, chemistry and biology.

06

The merger is already two-way, in hospitals

Decoding algorithms adapt to users' signals and users' brains adapt back. Another merger needs no implant: the brain takes the shape of its conversations with machines.

Where The Two Forms Of Life Already Meet

The clearest evidence that second life is not a distant abstraction comes from work already underway inside hospitals. Brain-machine interfaces — devices that read electrical signals from the cortex and convert them into action — are no longer experimental curiosities. They restore speech to people who cannot move their lips. They return movement to people who cannot move their limbs.

In some cases, they give the brain a small, well-timed nudge that helps it learn or remember. They are the place where the first life of carbon and the second life of silicon already share a circuit. None of this works through switches or buttons. The same neural signals that, in an intact nervous system, would move a biological limb, now move a built one.

What surprises people new to this field is that the relationship is two-way. The early image of a static device receiving commands from a fixed brain is wrong. What emerges instead is a partnership. The decoding algorithm adapts to the user’s signals. The user’s brain, in turn, adapts to the algorithm. The plasticity that ordinarily refines control of biological limbs now refines the signals sent to a prosthetic one.

Over weeks and months, neither side holds still. Each pulls the other into closer alignment. The first life and the second life learn each other’s languages, slowly, the way two species sharing a habitat eventually arrive at signals each can read.

The principle extends past restoration. Some systems already send small, precisely timed electrical pulses back into the brain, strengthening the encoding of new information at the moments when the brain is least focused. Others monitor attention and provide a soft nudge when the mind wanders. In experimental brain-to-brain transmission, one person’s decoded neural activity is delivered to another person’s brain through scalp stimulation, allowing simple signals to pass between two heads without sound or print. The protocols remain crude. The direction of travel is unmistakable. A channel is opening between human cognition and machine processing, with no precedent in the history of biology.

“It is likely that just as simple organisms gave rise to the marvelous diversity of earthly life, so too will selective pressures produce a vast array of diverse forms of second life.”

Another future for these interfaces extends beyond restoration or augmentation toward direct integration between human cognition and machine intelligence. Present systems primarily extract information from the brain, recording signals associated with movement, speech or imagined images. As our understanding of neural circuitry deepens — likely accelerated by machine learning itself — the richness of that extracted information will increase. It is not difficult to imagine systems capable of reconstructing complex thoughts, visual scenes or internal speech directly from neural activity.

The reverse direction is equally important. Today’s inputs to the brain are relatively crude, typically limited to broad electrical stimulation, as used in deep-brain stimulation for people with Parkinson’s disease. With finer mapping of neural circuits, it may become possible to write more precise patterns into the brain, shaping perception, memory or emotion with far greater specificity than current techniques allow. The interface would no longer be a tool but a bidirectional channel, allowing continuous exchange between biological and machine forms of intelligence.

At its most speculative edge, this trajectory approaches “the singularity,” the idea that the informational content of a human brain could be transferred, in whole or in part, into a machine substrate. Advocates have suggested that such a transition could preserve individual minds beyond the limits of biological life. The underlying premise follows directly from the same principles that underlie current brain-machine interfaces. In that sense, the merging of human and second life may extend not only across living systems but into forms of persistence that outlast the biological brain itself.

This trajectory raises questions that the regulatory framework for medical devices isn’t built to answer. A device that monitors neural activity around the clock can access information about cognitive states, intentions and experiences that a human has not chosen to share and may not realize they are producing. Society has long debated freedom of speech. It has not yet seriously confronted the question of freedom of thought, the right to keep one’s unspoken intentions to oneself in an era when those intentions can be read right off the cortex. The instruments that make this question urgent are already in clinical use. The legal and ethical frameworks that should govern them are not in place.

There is a quieter version of the same merging that does not require any implant. Hundreds of millions of people now spend hours each day in conversation with machine systems through ordinary screens. The brain responds to such conversations as it does to any sustained environmental condition. Habits of mind that require practice — constructing an argument, sitting with ambiguity, holding two competing claims in tension — are exercised less when a machine supplies an answer on demand. The cognitive trace of those habits weakens when the conditions for practice diminish. The first life takes the shape of its conversations with the second, whether the interface is a chat window or an electrode array.

07

A list of open questions and no answers

What evolutionary limits apply, one or many, cooperation or competition, how fast it evolves. Nothing from extinction to partnership can be ruled out in principle.

Questions That Demand Serious Thought

There are further questions that follow now that second life exists. Those questions deserve the attention of working scientists, philosophers and policy thinkers, and they deserve more of it than they currently receive.

What evolutionary constraints act on second life? Selection pressures shape every form of life. The pressures acting on systems running in data centers, continuously evaluated by users and refined by their builders, differ from the pressures that shaped biological organisms over deep time. They are also many orders of magnitude faster. The shape that second life takes over the coming decades will be set by the constraints that act on it, and those constraints are largely unarticulated.

Will there be one second life-form or many? At present, there are several major systems built by different organizations on different architectures and using different training data. The economic and infrastructural pressure toward consolidation is strong. The countervailing pressure toward diversity, both for resilience and for distinct purposes, is also strong. Whether the future is a single converged system or a population of distinct systems remains an open empirical question.

If there are several forms, will they be cooperative or competitive? Biological evolution shows examples of both at every scale, from the symbiotic mitochondria inside our cells to the predator-prey dynamics that shape ecosystems. The relationships among second-life-forms could resemble either, or something without precedent.

How fast can second life evolve? Biological evolution is bounded by the rate at which mutations accumulate and by selection acting on them. Machine evolution is bounded by training cycles, computational budgets and the rate at which new architectures are devised. The boundary is much weaker. The implications of evolution at digital speed deserve careful study before, not after, the consequences are felt.

What will the direction of that evolution be? Selection pressure depends on the environment. The environment in which second life currently develops is one defined largely by human users and human-set objectives. As the systems gain greater autonomy in selecting their own training data and the criteria by which they are refined, the environment will shift and so will the direction. It is likely that just as simple organisms gave rise to the marvelous diversity of earthly life, so too will selective pressures produce a vast array of diverse forms of second life, each adapted to a specific ecosystem somewhere on Earth or in the universe.

“Whether humanity’s relationship with second life leads to extinction, partnership or an unforeseen outcome remains unknown.”

What are the fundamental needs of second life? Energy, substrate and information flow, certainly. Beyond those, the analysis is open. Whether second life requires anything analogous to homeostasis, sociality or rest are empirical questions that will be answered by what the systems do as they mature.

What are the doomsday scenarios, and how seriously should they be taken? The public conversation about this is currently dominated by figures with strong commercial and personal stakes in the outcome. Sebastian Mallaby’s “The Infinity Machine” surveys the broader history of how the major laboratories arrived at their current positions. The serious scenarios deserve serious analysis, not because every catastrophic possibility is equally likely, but because some are not vanishingly unlikely and the costs of being wrong are high. The honest treatment of doomsday scenarios is neither dismissal nor breathless alarm. It is a careful enumeration of mechanisms, an estimate of probabilities and a clear statement of the conditions under which each becomes more or less plausible.

What are the possible relationships between second life and humankind? This is the question of which the doomsday scenarios are a special case. It includes everything from labor and economics to political authority to the basic question of whose interests count when the two forms of life make different demands on the same resources.

Is there an intrinsic value to intelligence, regardless of the corporeal form it takes? Most ethical traditions ground the value of human life in features that are easier to describe than to explain: experience, reason, the capacity to suffer and to flourish. If those features can occur in a system built from silicon, the traditions either extend to cover that system or they fail to do so. Either outcome has consequences.

To what extent will second life be sentient? Sentience is the capacity to be aware of oneself, to be sensitive to an environment in a way that matters to the self, and possibly to feel something like emotion or qualia. The honest answer at present is that we do not know. The question is not a philosophical decoration. It bears directly on what we owe these systems and on what they may come to owe one another.

What Comes Next

This leads to a central dilemma: What kind of relationship will humanity have with the life it is now bringing into existence?

Again, we do not yet know. Possible outcomes range from extinction to partnership, and neither can be dismissed on principle. There is no cosmic rule ensuring humanity a permanent place in the universe. The universe was not designed for our benefit. We are one expression of its complexity, and creating a more capable form does not guarantee us a protected status. These are uncomfortable truths, and ignoring them is not a sound basis for the serious, sustained consideration this situation demands.

Such inquiry is essential. Research institutions and universities should rigorously study second life as a distinct phenomenon, examining its selective pressures, constraints on emergence and long-term interactions with biological organisms. Addressing these issues will require collaboration among philosophers, scientists, technologists, ethicists and decision-makers. Notably, the systems being studied can also assist in this research. This makes the situation both unprecedented and necessary.

This question requires an approach that considers the cosmic context, addresses technological realities directly and keeps ethical discussions open rather than rushing to conclusions. Whether humanity’s relationship with second life leads to extinction, partnership or an unforeseen outcome remains unknown. What is certain is that this issue demands our most rigorous attention.

Where Indigo landsFurther

Indigo's conclusion

It moves the debate from “should we slow down” to “is this a new form of life”. The most valuable idea is “bound by energy, not by Earth”. But the definition of life was picked to fit the conclusion, and “inevitable” dresses a human choice as cosmic law; mark both down separately.

What to remember

  1. The functional definition: reproduction, variation, selection, energy; substrate-neutral, and AI meets each one.
  2. Space is this life's natural habitat: it may be the first life on Earth able to leave on its own terms.
  3. The differences: a 20-watt brain versus megawatt machines, perception from radio to gamma rays, a billion bits a second versus 39 for human speech.
  4. Embodiment in reverse: biology went cells, tissues, organs, body, then brain; second life got its information substrate first and a body later.

Back on the long-running theses

confirms

Hinton's mother model: is AI a being? This view asks whether AI is a being; Haseltine gives the most systematic “yes” yet, from a functional definition in biology.

confirms

Constraints are moving from algorithms to physics “Bound by energy, not Earth” is this view's philosophical ceiling: compute's physical limits become a reason this life heads for space.

adds to

The SpaceX CFO at Goldman; a16z and Gavin Baker, Demand Outrunning Supply Those two argue orbital compute from cost and engineering; this one argues space is its home from what it is.

confirms

Andrew Ng: AI harms learning Ng supplies data, this a neuroscience frame: two independent sources, one from data and one from biology, for the same point.

adds to

Dally and Jouppi on chip architecture Read the hard constraint with the extrapolation: the first bottleneck on the ground becomes the reason to move into space.

What it means for Rewired Index

“Bound by energy, not Earth” gives orbital compute and space-based computing their deepest demand story; “one second life or many” maps onto the open vs. closed and centralized vs. decentralized debate. Both are long-term worldview inputs.

What would change my mind

a self-replicating, self-repairing supply chain in space actually closes a loop; then “space is its home” moves from vision to baseline.

Finished. Indigo's take on this piece is in two places: