Self-Assembling Robots and the Potential of Artificial Evolution | Emma Hart | TED

譯者: Wei C. 審譯者: Shelley Tsang 曾雯海

Imagine a scientist

想像一位科學家

who wants to send a robot to explore in a faraway place,

想要將機器人 送去探索遙遠之地

a place whose geography might be completely unknown

一個地形可能完全未知的地方

and perhaps inhospitable.

而且可能不宜居住

Now imagine that instead of first designing that robot

現在想像如果不是 先設計好機器人送出

and sending it off in the hope that it might be suitable,

並且希望著 他應該能適應地形

instead, she sends a robot-producing technology

相反地，她運送 一個機器人生產的技術

that figures out what kind of robot is needed once it arrives,

可以在抵達時判斷出 這裡適合哪種機器人

builds it and then enables it to continue to evolve

將它建造出來 並讓它可以繼續演化

to adapt to its new surroundings.

來適應這個新的環境

It’s exactly what my collaborators and I are working on:

這正是我與合作對象 正在進行的研究

a radical new technology which enables robots to be created,

一個讓機器人可以被 創造出來的全新技術

reproduce and evolve over long periods of time,

並能在長時間中繁衍與進化

a technology where robot design and fabrication becomes a task

這個技術使得 設計與製造機器人

變成了機器 而不是人類的任務

for machines rather than humans.

Robots are already all around us, in factories, in hospitals, in our home.

機器人早已遍布我們的周遭 在工廠、醫院和家中

But from an engineer's perspective,

但從工程師的角度來看

designing a shelf-stacking robot or a Roomba to clean our home

設計一個擺放物品 或掃地的機器人

is relatively straightforward.

還相對直接了當

We know exactly what they need to do,

因為我們知道它確切要做什麼

and we can imagine the kind of situations they might find themselves in.

也能想像它們可能會身處何種環境

So we design with this in mind.

所以能在腦海中設計

But what if we want to send that robot to operate

但如果想將機器人送到

in a place that we have little or even no knowledge about?

我們不甚理解的地方運作呢？

For example, cleaning up legacy waste inside a nuclear reactor

例如，清理核反應爐中的廢棄物

where it's unsafe to send humans,

這些不適合派遣人類來做

mining for minerals deep in a trench at the bottom of the ocean,

在深海海溝中開採礦物

or exploring a faraway asteroid.

或是探索遙遠的小行星

How frustrating would it be if the human-designed robot,

這將會有多令人沮喪? 如果一台人類設計的機器人

that had taken years to get to the asteroid

花費多年到達小行星

suddenly found it needed to drill a hole

突然發現它得鑽洞

to collect a sample or clamber up a cliff

來採集樣本或是攀爬懸崖

but it didn't have the right tools

但是它沒有正確的工具

or the right means of locomotion to do so?

或是對的運動方式來執行

If instead we had a technology

相反地，如果我們有一項科技

that enabled the robots to be designed and optimized in situ,

能在原地設計和優化機器人

in the environment in which they need to live and work,

在它們需要生活和工作的環境中

then we could potentially save years of wasted effort

那麼我們或許能 省下多年努力的心血

and produce robots that are uniquely adapted

製造出對它們身處的環境

to the environments that they find themselves in.

有獨特適應性的機器人

So to realize this technology, we've been turning to nature for help.

為了實現這項科技， 我們向大自然尋求協助

All around us,

環顧四周

we see examples of biological species

我們看到的各種生物

that have evolved smart adaptations

都已演化出能夠在此環境中

that enable them to thrive in a given environment.

能夠蓬勃生長的聰明適應機制

For example, in the Cuban rainforest,

舉例來說，在古巴雨林中

we find vines that have evolved leaves

我們發現了幾種演化出葉子的藤蔓

that are shaped like human-designed satellite dishes.

形狀就如同人類設計的碟形天線

These leaves direct bats to their flowers

這些葉子利用增強蝙蝠所發出的訊號

by amplifying the signals that the bats send out,

將牠們引到花旁

therefore, improving pollination.

因而增加授粉的機率

What if we could create an artificial version of evolution

如果我們能創造出人工版的機制

that would enable robots to evolve in a similar manner

讓機器人能如同生物有機體

as biological organisms?

用相同的方式演化呢？

I'm not talking about biomimicry,

我指的不是仿生技術

a technology which simply copies what's observed in nature.

這種技術單純複製 自然所觀察到的結果

What we're hoping to harness is the creativity of evolution,

而我們是希望能駕馭演化的創造力

to discover designs that are not observed here on Earth,

去發掘在地球上所觀察不到的

the human engineer might not have thought of

人類工程師可能沒想到的

or even be capable of conceiving.

或甚至不可能理解的設計

In theory,

理論上

this evolutionary design technology could operate completely autonomously

這種演化設計技術應要能

在遙遠的星球上自動運行

in a faraway place.

But equally it could be guided by humans.

但同樣又能被人類所指引

Just as we breed plants for qualities such as drought resistance or taste,

如同我們繁殖 具有抗旱或特殊味道的植物

the human robot breeder could guide artificial evolution to producing robots

人類機器人飼養員 要能夠引導人造演化

with specific qualities.

產出具有某些特質的機器人

For example,

舉例來說

the ability to squeeze through a narrow gap

能穿過狹小縫隙的能力

or perhaps operate at low energy.

或是能夠低功耗運行

This idea of artificial evolution imitating biological evolution

這種人工模仿生物演化的想法

using a computer program

利用電腦程式

to breed better and better solutions to problems over time

隨時間衍生出越來越好的解法

isn't actually new.

這個想法其實不算新穎

In fact, artificial evolution,

事實上，人造演化

algorithms operating inside a computer,

電腦內執行的演算法

have been used to design everything from tables to turbine blades.

早已被用來設計各種東西， 從桌子到渦輪葉片

Back in 2006,

早在2006年

NASA even sent a satellite into space with a communication antenna

NASA就曾將裝有通訊天線的衛星 送入太空

that had been designed by artificial evolution.

而它就是利用人造演化所設計的

But evolving robots is actually much harder

但演化型的機器人其實更難設計

than evolving passive objects such as tables,

相較於被動的物品，例如桌子

because robots need brains as well as bodies

因為機器需要身體跟大腦

in order to make sense of the information in the world around them

才能理解周遭世界的資訊

and translate that into appropriate behaviors.

並且將其轉譯成合理的行為

So how do we do it?

所以我們要如何辦到呢？

Surprisingly, evolution only needs three ingredients:

出乎意料的是， 演化只需要三個要素：

a population of individuals which exhibit some physical variations;

能夠展現一些 生理上不同的族群、

a method of reproduction

後代能繼承父母特徵

in which offspring inherit some traits from their parents

以及偶而能有 後天突變的繁衍方法、

and occasionally acquire new ones via mutation;

and finally, a means of natural selection.

最後是天擇的方式

So we can replicate these three ingredients to evolve robots

我們能應用這三種要素， 讓機器人演化

using a mixture of hardware and software.

利用軟硬體混合的方式

The first task is to design a digital version of DNA.

第一項任務便是 打造數位化的DNA

That is a digital blueprint that describes the robot's brain, its body,

這種數位藍圖描述了 機器人的腦、身體、

its sensory mechanisms and its means of locomotion.

感知機制與運動方式

Using a randomly generated set of these blueprints,

利用隨機生成的一組藍圖

we can create an initial population of 10 or more robots

我們可以創造出最初具有 10個或以上機器人的人口

to kick-start this evolutionary process.

來啟動這個演化程序

We've designed a technology that can take the digital blueprint

我們已經設計出一種科技， 能將數位藍圖

and turn it into a physical robot without any need for human assistance.

轉換成實體機器人， 而不需要人為協助

For example, it uses a 3D printer to print the skeleton of the robot

舉例來說， 利用3D列印出機器人的骨架

and then an automated assembly arm like you might find in a factory

與工廠中常見的自動化組裝手臂

to add any electronics and moving parts,

來放上任何電子元件與動件

including a small computer that acts as a brain.

包含一個作用如同大腦的小電腦

And to enable this brain to adapt to the new body of the robot,

而為使大腦能適應機器人的新身體

we send every robot produced to an equivalent of a kindergarten,

我們再將每一台生產的機器人 送到如同幼兒園的地方

a place where the newborn robot can refine its motor skills

讓新生機器人能 磨練他們的運動技巧

almost like a small child would.

就像是一個小孩一般

To mimic natural selection,

為了模仿天擇

we score these robots on the ability to conduct a task.

我們為這些機器人 所執行的任務做評分

And then we use these scores

並且利用這些分數

to selectively decide which robots get to reproduce.

來決定哪些機器人可以被繁殖

The reproduction mechanism

這個繁衍機制

mixes the digital DNA of the chosen parent robots

利用混合指定 父系機器人的數位DNA

to create a new blueprint for a child robot

來創造出子系機器人的新藍圖

that inherits some of the characteristics from its parents

繼承一些父母的特徵

but occasionally also exhibits some new ones.

但是卻又偶能展現新特徵

And by repeating the cycle of selection and reproduction over and over again,

藉由不斷重複天擇與繁衍的循環

we hope that we can breed successive generations of robots

我們期望能繁殖出 一代接著一代的機器人

where, just like is often observed in biological evolution,

並且能從中觀察到如生物演化一般

each generation gets better than the last,

一代比一代還要好

with the robots gradually optimizing their form and their behavior

而機器人會逐漸 優化它們的型態與行為

to the task and the environment that they find themselves in.

來適應它們所面對的任務與環境

Now, although this can all take place

雖然這些都能在

in a time frame that's much faster than biological evolution,

比生物演化更快的時間尺度下發生

which sometimes takes thousands of years,

還是可能花上數千年的時間

it's still relatively slow in terms of the time frames we might expect

時間還是相對的慢

對比於我們在現代世界所期待

in our modern world

to design and produce an artifact.

生產一個人造產品的時間尺度

It's mainly due to the 3D printing process,

主要是因為3D列印的程序

which can take more than four hours per robot,

一個機器人需花費四小時以上

depending on the complexity and the shape of the robot.

取決於機器人的複雜度與形狀

But we can give our artificial evolutionary process a helping hand

但是我們能幫人造演化的過程一把

to reduce the number of physical robots that we actually need to make.

來減少我們實際需要製造的機器人

We create a digital copy of every robot produced

我們可以將每一個 產出的機器人的數位副本

inside a simulation in a computer,

建立在電腦的模擬中

and we allow this virtual population of robots to evolve.

並且讓這個虛擬人口演化

Now it's quite likely that the simulation isn't a very accurate representation

雖然這個模擬 不會非常精確的呈現真實世界

of the real world.

But it has an advantage that it enables models of robots to be created

但是它的優勢在於 能在幾秒內建立與測試模型

and tested in seconds rather than hours.

而不會花費數個小時

So using the simulator technology,

所以利用模擬器的技術

we can quickly explore the potential of a wide range of robot types

我們可以發掘潛在 一系列的機器人種類

of different shapes and sizes, of different sensory configurations,

有著不同的形狀與尺寸， 還有感知配置

and quickly get a rough estimate of how useful each robot may be

並且可以在實際製作機器人前

before we physically make it.

快速估計每台機器人是否有用

And we predict that by allowing a novel form of breeding

而我們預測藉由讓實體機器人

in which a physical robot can breed with one of its virtual cousins,

與其虛擬表親之一繁殖這種新模式

then the useful traits that have been discovered in simulation

在模擬中發現的有用特徵

will quickly spread into the physical robot population,

可以快速地散播到 實體機器人人口之中

where they can be further refined in situ.

讓他們能在原地進化

It might sound like science fiction,

這聽起來很像科幻小說

but actually there's a serious point.

但卻有非常重要的一點

While we expect the technology that I've just described

當我們期望這項技術

to be useful in designing robots,

能對設計機器人非常有用時

for example, to work in situations where it's unsafe to send humans

比方說， 在不適合送人進入的工作環境

or to help us pursue our scientific quest for exoplanetary exploration,

或是幫助我們執行 科學探查與異星探索

there are some more pragmatic reasons

對於我們為何要考慮人工演化

why we should consider artificial evolution.

有一些非常實際的理由

As climate change gathers pace,

隨著環境變遷加劇

it is clear that we need a radical rethink

很明顯我們需要徹底的反思

to our approach to robotic design here on Earth

我們在地球上的機器人設計

in order to reduce that ecological footprint.

來減少生態足跡

For example,

舉例來說

creating new designs of robot built from sustainable materials

利用永續材料創造新的機器人設計

that operate at low energy,

能夠低功耗運行

that are repairable and recyclable.

而且是可維修也可回收的

It's quite likely that this new generation of robots

新一代的機器人很有可能

won't look anything like the robots that we see around us today,

完全不像現在看到 在我們周圍的機器人

but that's exactly why artificial evolution might help.

這也就是為什麼 人造演化能提供協助

Discovering novel designs by processes that are unfettered by the constraints

利用這些程序發掘新型設計

that our own understanding of engineering science

能讓我們不再被

用在現有設計程序的 工程知識所限制

imposes on the design process.

Thank you.

謝謝

(Applause)

（掌聲）