Rethinking Biology: A Conversation With Michael Levin
Summary
TLDRこのスクリプトは、生物学の新しい波を代表する生物学者であるMichael Levin教授と、高度技術移行の教授でアリゾナ州立大学の「ヒューマン・ビーイングの未来」イニシアチブのディレクターであるAndrew Maynard氏の対話を記録しています。Levin教授は、細胞がどのように集まってネットワークを形成し、それらのネットワークが集団知能を発揮して生物体の形を決定するのかについて研究しています。ゲノムは生物学的な「ハードウェア」をエンコードし、個々の細胞はそのハードウェアを再プログラミングして、集団として問題解決を行うと述べています。この対話では、生物学の可能性を広げる革新的な研究と、それが人工知能や人間の定義に与える可能性について触れています。
Takeaways
- 🧬 生物学の基礎知識に挑戦する新しい波の思想家と研究者がいて、私たちが生きる世界についての既定観念に問いを投げかけ、包括的な変革の可能性を広げている。
- 🧠 細胞は個々の存在として自分の位置を知らないが、細胞の集まりはネットワークとして情報を共有し、集団知能を発揮して形を作り出す。
- 🧵 ゲノムは細胞の「ハードウェア」をエンコードし、proteinsなどの具体的な機能を決定するが、生物体の形や記憶内容は直接指定しない。
- 🔋 生物の電気的ネットワークは細胞集団の集団知能を介して行動パターンを制御し、その_voltage pattern_を操作することで生物学的な形を変化させることができる。
- 🪲 プラリア(扁形虫)の研究から、生物電気回路のパターンを操作して、記憶を書き換え、_TWO-HEADED_(二頭)のプラリアを生み出すことができる。
- 🌱 進化の過程で生物は不安定なハードウェアと環境の変化に対処し、問題解決エージェントとして機能し、形質空間を通じて様々な形に適応する能力を発達させてきた。
- 🧬 遺伝子工学的なアプローチから一歩後退し、生物学の各レベルが持つ自己主張と問題解決能力を活用して、再生医療などの応用技術を開発している。
- 🧠 記憶と学習は脳だけでなく、分子ネットワークや単一細胞でも見られる。生物学的な記憶は組織を越えて移転することができる。
- 🤖 AIと同様に、機械も問題解決能力を持つが、意識や自己認識を持つかどうかは別問題であり、それが彼らに対する道徳的な扱い方にも影響を与える。
- 🌐 現代の科学は、私たちが世界の仕組みについて考える伝統的な考え方を根拠から再評価し、再定義し始めている。
- 🚀 未来の生物学と再生医療は、人間の体の可能性を大幅に拡大し、個々の目標に応じて自在に身体を変化させることができる未来を示唆している。
Q & A
細胞はどのようにして自分の体の位置を知るのですか?
-細胞自体はその位置を知りませんが、細胞集団はネットワークを形成し、それらのネットワークには集団知能というコンピュータティブな性質があります。このネットワークが、自分が構築すべき物体の粗い表現を記憶し、現在の状況とのデルタを減少させるために誤差を最小限に抑えます。
ゲノムはどのようにして私たちの形を決定するのですか?
-ゲノムはハードウェアをエンコードし、各細胞が持つべきマイクロハードウェアの種類を指示します。その後のことは、重要な意味でソフトウェアであり、そのハードウェアは再プログラマブルで、デフォルトで何かを実行し始め、高度に再プログラマブルです。
生物学的な「アウトオブザボックス」デフォルトとは何ですか?
-「アウトオブザボックス」デフォルトとは、例えば人間であれば、基本的な人間の形を指します。しかし、生物学的生き物は状況によって異なる道を進むことができます。例えば、早期胚を切っても、半分の体ではなく、同卵双子や三つ子ができるように、生物学は様々なperturbationsに対処できます。
プラリアの研究で何が見つかりましたか?
-プラリアの研究では、細胞集団のbioelectric circuitが発見され、その特定の電圧パターンが1つの頭を持つべきであるという事実をエンコードしていることがわかりました。研究者はこのパターンを書き換えて、2つの頭を持つように変更することができました。
プラリアのbioelectric circuitを操作することで何が可能です?
-プラリアのbioelectric circuitを操作することで、プラリアの形と構造を変化させることができます。これにより、2頭を持つプラリアを作成することができ、その2頭を持つ状態は永久的です。遺伝子的な変化なしで、切り離された後も2頭を持つことが可能です。
プラリアのbioelectric circuitの操作は、脳の記憶にどのように関連していますか?
-プラリアのbioelectric circuitの操作は、脳の記憶に類似しています。プラリアの2頭を持つ記憶はcounterfactual memoryであり、現在 trueではないため、未来のある時点で真になる可能性があるという能力を有しています。これは、脳で行われているmental time travelの能力に似ています。
記憶はどのようにして異なる生物間で転送されるのでしょうか?
-記憶は生物学的なネットワークを通じて転送されることができます。例えば、訓練されたアピスのRNAを無知のホストの脳に注入することで、情報を転送する実験が行われています。また、プラリアでは、頭を切り落としてから新しい頭と脳を再生させた後も、情報を記憶していることが示唆されています。
カタツムリからチョウに変わる際の記憶の転送はどのように行われますか?
-カタツムリからチョウへ変わる際には、記憶が保持されますが、カタツムリの記憶はチョウには無意味です。代わりに、記憶は一般化された食品のカテゴリーへとGENERALIZEされ、新しい体で駆動される異なるセットのエフェクターシグナルへと関連付けをリマップします。
生物学的なネットワークを操作することで、生物学と医療にどのような可能性が開かれるでしょうか?
-生物学的なネットワークを操作することで、出生欠陥、再生誘導、およびtorsの正常化などのアプリケーションが可能になります。形態を制御し、生物学的なシステムの各レベルが設定されたパラメータ内で最適化を試みる環境を創造することができます。
マシンやAIにおける知能とは何ですか?
-マシンやAIにおける知能は、問題解決能力です。これは、意識や自己認識とは別であり、異なる手段で同じ目標を達成するナビゲーション能力を意味します。これは、現在の機械学習の多くで開発されている能力です。
私たちはAIを構築しているので、その動作を完全に理解していると思いますが、なぜそれでも予測不可能な結果が発生するのでしょうか?
-私たちが構築するシステムには、私たちが完全に予測することができないbuilt-in agencyがあるからです。これは、システムが持つ問題解決能力や学習能力によるものです。たとえ私たちがシステムの各部分を知っている場合でも、システム全体の動作は予測不可能かもしれません。
未来のヒトになる意味とは何ですか?
-未来のヒトになることは、身体の自由度を意味します。私たちは、宇宙線がDNAに当たった結果、生まれた体にとどまる必要があるという偶然性にとらわれることなく、自分の目標に応じて異なる知能や長命を持ちたいと考えることができます。
Outlines
😀 未来人类研究所の紹介とゲストの迎え
アンドリュー・メイヤード教授は、亜利桑那州立大学の先端技術移行研究所の教授であり、「未来人类」イニシアティブのディレクターです。ゲストとして、生物学者で先駆的な研究を行っているマイケル・レヴァン教授を迎え、彼の研究は私たちの生物学に対する前提知識に挑戦し、私たち自身を含む世界を変革する可能性を広げています。
🧠 細胞の集団知能と遺伝子の役割
個々の細胞は自分の位置を知りませんが、細胞の集団はネットワークを形成し、集団知能を発揮します。レヴァン教授は、細胞のネットワークがどのように集団として情報を処理し、生物の形や振る舞いを決定するのか、また遺伝子の役割について語りました。遺伝子は生物の「ハードウェア」をエンコードし、その後のプロセスは「ソフトウェア」であり、再プログラマブルで、生物学的な形を決定するのではなく、デフォルトの動作を実行し始め、その後は高度に再プログラマブルとなります。
🐛 記憶の移転と生物学的な柔軟性
生物学的なcreatabilityについては、通常は非常に頼りになる発展過程がありますが、他の生物工学者が示すように、私たちの考えられる限界は非常に小さいです。特に、ガラの葉がハエの寄生蜂によって変貌される例や、フカケの再生能力に関する研究が、細胞の集団知能の電気的インターフェースについて語りました。彼らは、フカケが2つの頭を持つようにバイオ電気パターンを書き換え、その動物がその後も2つの頭を持つようにしました。
🧬 記憶と集団知能の連続性
レヴァン教授は、私たちの脳における記憶の蓄積方法については不明であると語りました。しかし、集団知能における記憶の概念は、私たちの脳の記憶と類似しています。彼らは、フカケにおける2つの頭の記憶をカウンターファクトゥアルな記憶と見なしており、将来的に真実になる可能性のあるものであり、それが動物が将来的に傷ついたときに行うべき行動です。
🦋 記憶のリマッピングと進化
カタツムリからチョウに変わる際に記憶が保持されるという過去の研究、特にダグラス・ブラックストンの经典的なカタツムリバタフライの研究について語りました。カタツムリは葉を食べるが、チョウは三维空間で飛び、蜜を吸う必要があります。そのため、記憶を一般化し、新しい体にリマップする必要があります。これは、進化や認知科学など、様々な分野で見られるリマッピングの例であり、非常に興味深いと感じています。
🧵 生物学のネットワークと目標達成
生物学のネットワークは、細胞が集団として記憶を持ち、私たちが生物学を実際に利用できるようになることについて話しました。短期間では、生物学を再指定する方法を理解し、出生欠陥や再生を誘導し、torsを正常化するアプリケーションがあります。また、生物学の各レベルは、生物学的なシステム全体が望む方向に向かって最適化しようとするパラメータを設定することで、カジュアルなトリックを実行するようにすることができます。
🤖 AIと機械学習の進化
AIと機械学習の進化について語り合い、レヴァン教授は、私たちが持っているアーキテクチャは、キープロパティを多く持っていないが、問題解決のインテリジェンスを持っていると語りました。また、現在のAIアーキテクチャでは、予測不可能なアウトカムを持つビルドシステムがあることを理解し、そのアプローチから離れることが重要です。
🧐 科学的なルネッサンスと意義の危機
レヴァン教授は、科学的ルネッサンスが到来していると感じています。多くの異なる分野で確立された考え方に挑戦している人々が多く見受けられます。また、意義の危機から脱却し、より科学的に支持される方法で考え方を再構築する必要があると語りました。
🚀 人間の未来と身体の自由
未来の人类について語り合い、レヴァン教授は、身体の自由とエンバディメントの自由が見られる未来を想像しています。人々は、宇宙線によってDNAに当たっただけで、生まれた体にとどまる必要があるとされる過去の物語に驚くでしょう。未来の世界では、人間の身体や能力に制約されることなく、自分自身の目標を追求することができるでしょう。
Mindmap
Keywords
💡細胞集団
💡ゲノム
💡集団知能
💡問題解決能力
💡bioelectric circuit
💡集団の記憶
💡進化
💡自己認識
💡人工知能
💡科学的なルネッサンス
💡自己形成
Highlights
Andrew Maynard, a professor at Arizona State University, discusses the future of human evolution with pioneering biologist Michael Levin.
Levin's research is redefining our understanding of how cells and organisms operate, challenging traditional biological assumptions.
The conversation explores the concept that individual cells don't know their location but cell networks do, indicating a form of collective intelligence.
Levin explains that the genome encodes the hardware of a cell, but the software, or functional expression, is shaped by the cellular collective.
The idea that organisms have an 'out of the box' default form is not fixed; biology can be manipulated to achieve different outcomes.
Levin's lab has demonstrated that by manipulating bioelectric circuits, they can change the physical form of organisms, such as creating two-headed flatworms.
The concept of cellular 'memory' is introduced as a way to understand how cells collectively remember and reshape their form.
Levin discusses the transfer of memories between different species and across generations, highlighting the complexity of biological memory.
The potential applications of these discoveries in regenerative medicine are significant, offering new ways to treat birth defects and induce regeneration.
Levin's work suggests a shift from deterministic science to a more nuanced understanding where biology is persuaded to achieve desired outcomes.
The discussion touches on the implications of artificial intelligence, drawing parallels between machine learning and biological intelligence.
Levin emphasizes the need for humility in understanding the capabilities of both biological and artificial systems, given their inherent complexity.
The conversation suggests we may be on the cusp of a scientific renaissance, reevaluating long-held beliefs across various disciplines.
Levin speculates about the future of humanity, envisioning a time where the constraints of our physical bodies are no longer a limitation.
The potential for a crisis of meaning in society is discussed, with a call for a new narrative that embraces the implications of modern scientific discoveries.
The interview concludes with a call to action for scientists and philosophers to provide a new perspective that honors both our humanity and the power of our intelligence.
Transcripts
[Music]
hello and welcome to the future of Being
Human unplugged my name's Andrew mayard
I'm a professor of advanced technology
transitions at Arizona State University
and I'm also the director of asu's
future of Being Human initiative today
it is my great pleasure to be joined by
the pioneering biologist Professor
Michael Levan so I first became aware of
Mike's work some time ago when one of my
grad students asked if I'd seen it and
what I thought of it so here I should
say I'm not a biologist so Mike hadn't
actually been on my radar but there were
enough hints from Mike student that
there was something profoundly
transformative going on in his lab that
I thought I should probably check it out
and I'm very glad I did so Mike is part
of a new wave of thinkers and
researchers who are challenging received
wisdom about the world we live in and
opening up radical new possibilities for
how we transform it including ourselves
and here I suspect that that phrase
profoundly transformative is an
understatement through his work on why
cells do what they do why organisms look
and behave as they do and why our genome
doesn't seem to be that great of an
indicator of much of that um it's clear
that Mike and his collaborators are
upending a lot of our assumptions about
how biology works and what that implies
about everything from intelligence
including AI to what it means to be
human so Mike it is great to have you
here for this conversation thanks so
much Andrew so happy um yeah and it's
it's one that I've actually been looking
forward to for a long time um and I'm
expecting that we're actually going to
cover a lot of ground and as always with
this unplugged format um Serendipity is
the name of the game so I have no idea
where this conversation is going to go
but I did want to start with um the
biology um and the science um and I want
to start off with a question that I
suspect some people will consider to be
very naive um and I'm allowed to because
as I said I'm a physicist and it's
something that actually has been
bothering me for years so the question
is you take a cell on the end of my nose
how does it know that on the end of my
nose how does it know what the shape of
my nose is and what its places in my
body because that cell if you look at
the genome there's no way it should know
where it is in space and time how on
Earth does this work yeah um great
question and uh like you uh I entered
this field from a different area I was a
computer scientist and and likewise um
asking similar similar questions um I as
as far as we know now the individual
cell does not know where it is it
doesn't know anything about a nose it
doesn't know anything about you but the
cellular Collective does so cells merge
into networks and these networks have
computational properties that we're only
beginning to understand and it's those
networks that um Implement a kind of
collective intelligence that solves a
number of problems one of those problems
that solves is storing a rough uh
representation of what it's supposed to
be building and then minimizing error SL
stress in order to reduce the Delta
between what it thinks it should be
building and what it thinks the current
state is so so so what we ask in this
lab all the time is what what does the
group know not you know sometimes what
does the cell know but mostly what does
the group know right and and I've got to
ask where where does that idea of what
the group should be knowing come from
presumably at some point that is coded
within the
genome well uh one way one way to think
about this uh and I know you know people
are not enamored of of computer
analogies in biology and and and in many
ways they're bad but but but this one I
think is a good one uh the genome what
the genome encodes is the hardware the
jome Tells every cell what uh kind of uh
microscopic Hardware it has to play with
those are the proteins that it has
everything that happens after that in an
important sense is software that that
Hardware is reprogrammable uh it uh it
it does representation it does a number
of interesting things that that that
Hardware does but but you have to treat
the genome as that the genome does not
directly specify your shape it doesn't
specify uh the content of the of the
memories of your of your body networks
um what it gives you is some amazing
Hardware that does some stuff out by
default you know sort of out of the box
and also is is is highly reprogrammable
yeah yeah so I just to to sort of follow
on with that so this idea of out of the
box so I'm assuming if you're looking at
at humans for instance and we're going
to start really complex and we'll
probably get more simple from here
but the the out of thebox default is a
rough human shape but what you're saying
it seems to be is that that doesn't have
to be the the end point we can begin to
sort of tweak the out ofth box sort of
default of the genome cor correct I mean
if you think about um the task that that
is set in front of biological creatures
and I can give you some examples of of
some amazing things that they do uh the
kind of architecture that we com that we
use in computer science where the
hardware is ex extremely reliable and
information is meant to be kept and we
make sure that it doesn't get get
altered and and we keep the noise down
and then the higher levels are sort of
insulated from the lower levels um that
that doesn't work in biology because the
hardware is fundamentally unreliable
everything that is in the cell is going
to be damaged turned over no the noise
is huge and during Evolution guaranteed
the hardware is going to change so
there's going to be mutations there are
going to be the environment will change
your own parts will change everything is
going to change so sticking with this
idea that um what you're going to try to
do is create a solution to a specific
problem the way that our current let's
say genetic algorithms do do um is I I
think is not what biology does at all
what I think what biology does is create
problem-solving agents that operate uh
and we could talk about some of the ways
that that this happens they operate in
various problem spaces like anatomical
space and uh if all all things being
equal they take the same journey through
anatomical morphis space and you go from
a being a a single cell oite to a human
shape but but they're perfectly willing
to take other Journeys if the situation
um requires them for example if you cut
an early embryo into pieces you don't
get half bodies you get monozygotic
twins triplets whatever so so they can
make up for all kinds of weird
perturbations some really strange ones
in certain species you know amazing
examples um and then uh and then you can
also you you could also get these cells
to do other things so we we've shown
examples of that of of of you know PL
area making flat worms making bodies
with the heads of different species and
and things like that there's a lot of
variety yeah so why why do we sort of go
to the that the flat won't work because
I think that actually sort of shows some
really interesting stuff where you can
effectively I I'm not sure that
reprogram is programing is the the right
term but you can actually get the the
the form and structure of flat worms to
change depending on how you actually
program these collectives of
cells yeah uh and and I guess the the
first thing I wanted to just um mention
is that this this idea the plasticity of
of the hardware is so uh hard for us to
see because development is normally very
reliable so so most of the time it does
exactly the right thing and so we get
this we get lulled into this false sense
of security that we know what these
genomes do and they produce a certain a
certain shape and that's that um there
are other bio then I'm going to I'll
I'll address the the plenaria point in a
second but there are other bioengineers
besides the human ones that show us how
limited that view is and uh one of the
things that uh that that that I show
sometimes in my talks are these Galls so
imagine uh you have a you have a a flat
Green Leaf that belongs to an oak and
you kind of know you say well the oak
genome makes this flat green thing you
know 100% of the time that's what it
makes well Along Comes A a wasp a
parasite that's a non a non-human
bioengineer and it prompts the leaf
cells with some some chemicals not
really known exactly how it works and
they go on to make this crazy beautiful
red spiky thing that looks absolutely
nothing like the leaf so we we would
have zero clue that those cells are
capable of doing that and you can bet
that the uh the WASP is not sitting
there micromanaging where the cells go
it's not micromanaging the gene
expression it's not doing genetic
editing it is uh communicating to the
cellular Collective some prompts that
take advantage of their morphogenetic
capacity so so nature is already doing
this but you know evolution is making
these these plastic things that can
respond in interesting ways and so in
plaria we took advantage of that and the
thing about planaria is that it's these
flat worms uh they're they've got a head
and a tail and lot you know true brain
lots of different organs you can chop
them into pieces the record is I don't
know 275 I think or something like that
and each piece will reliably regenerate
an entire worm so so so really
interesting to ask how does every piece
know what a correct worm is supposed to
look like and so we were studying this
uh and we long story short we discovered
that there's a there's a bioelectric
circuit and this is one thing my lab
does is it studies this really
interesting electrical interface to the
collective intelligence of these cells
that that interface is it's kind of like
an API that cells use to to hack each
other's behavior and um and and there's
a circuit that has a particular voltage
pattern that basically encodes the fact
that you should have one head and we
learn to rewrite that pattern to change
that into saying two heads instead of
one and those animals if you then if you
then cut them they will make two-headed
animals and so so a couple things
interesting there one is that uh we can
actually see the bi electrical pattern
so we now have the ability to directly
visualize the memories in the mind of
this collective intelligence you can you
can see them the way that
neuroscientists try to read the brains
um that's a b uh we can at least begin
to uh decode them so that now you can
rewrite them see these two-headed worms
are permanently two-headed meaning that
if you keep cutting them in plain water
with no more manipulation of any kind
they will continue to be two-headed for
forever with no genetic change right
right so you look at that I actually I
love the analogy of an API between cells
by the way um but you also use this this
term memory so effectively by
manipulating these bioelectrical
Networks you're EMB in effectively a new
Collective memory um Within These um
these plaria um how analogous is that to
sort of the memory we think of in in our
brains is is there a Continuum there um
well a couple of things so so
mechanistically actually no one really
knows how memories are stored in our
brain um there's a con there's a sort of
a conventional story having to do with
something about uh synaptic structures
that story has a lot of cracks in it
there are some folks that have been
challenging that in a in a in a strong
way it it it really isn't clear the
biggest thing about memory is not just
the storage medium but the
interpretation because uh there have
been all kinds of experiments on moving
memories from one animal to another and
uh in in fact across radically different
architectures so let's say from
caterpillar to butterfly memories
persist they have to be remapped you
know the straight up memories of a
caterpillar don't don't are useless to a
butterfly everything is is different and
so so so these we don't know how it's
handled in brains I mean I I have
suspicions but but we don't know but but
the thing that the thing that is
um homologous here is a couple things
first of all a lot of the Machinery is
the same so ion channels um
neurotransmitters electrical synapses
all of that stuff is there and being
used much like in in brains and the idea
that that that two-headed memory that we
that we first incept into these animals
is a counterfactual memory meaning that
it isn't true right now so you can put
in a a two a a you know kind of a
bipolar memory pattern into a
anatomically normal worm and that
pattern is not what the worm is right
now so I think of it as the beginnings
of that kind of mental time travel that
we have meaning the ability to conceive
of and remember things that are not true
right now but might be true at some
later time so that pattern right that
that pattern is what the animal is going
to do if it gets injured in some future
time okay okay so so this actually takes
us into
weird territory and I I'm going to push
on this I I'm not sure how comfortable
you are going there but it's the the way
that you're actually challenging sort of
our concepts of memory um all the way
through sort of memories from Collective
collectives of cells all the way up to
what we understand about sort of the
memory in our heads um it seems like
you're implying that in principle we can
actually sort of shift memories around
we can actually sort of put new memories
maybe it's just me memories of sort of
physical form but we can put memories
into actually I'm going to move away
from humans but but into the the brain
of an organism but you can also transfer
those memories either between
generations of organisms or Beyond there
I how am I going outside the the bounds
of reality here no no I don't think you
are I mean it's it's been done people
there there are lots of papers on moving
uh memories from from one body to
another so so some of the best modern
work is David lansman um at UCLA who um
injects RNA ground up from trained Appia
into the brains of of of naive hosts and
the and the information transfers
there's a long history of of that work
in plaria and and you know this was
discovered in the 60s but we actually
confirmed it ourselves in 2013 um if you
train the worm chop off their heads and
wait for the tail to regenerate a brand
new head with a brand new brain they
still show recall of the information
which means that Not only was it
partially stored in the tail but also
somehow imprinted onto the new brain as
the new brain develops so this idea of
Behavioral memories uh moving through
tissue moving across tissues being
transferred in molecular uh you know
molecular media um I yeah I think that's
I think all of that exist and and if you
could talk a little bit more about the
caterpillar butterfly example because I
that work of yours just blew my mind in
terms of the the progression from the
catabella to the butterfly with retained
memories so so I want to be really clear
that isn't my work so we weren't that so
so there was there was old work that
that did it in um uh various kind of
larv and beetles and things like that
and then uh the classic caterpillar
butterfly stuff uh was done by um
Douglas Blackiston who's a staff
scientist in my lab that's kind of a
coincidence I hired him um you know a
long time ago uh and uh you know I
didn't realize at first that he had done
that amazing work but but anyway uh
the the the the results basically go
like this um you train a caterpillar to
eat food which for the caterpillar is
leaves on a particular color disc the
caterpillar uh under goes metamorphosis
because what it needs to do is shift
from a softbed kind of creature which
requires a particular controller because
you know in the soft body there's
nothing you can push on right so so then
it becomes a butterfly that has to uh
that has to um live in a
three-dimensional world now and um and
so because of that the brain is largely
dissolved a lot many of the connections
are broken most of the uh most of the
cells are killed off there's some
there's some debate now as to you know
whether everything is killed off or
whether some things remain but the
interesting thing is not just the
Persistence of the memory the
interesting to me the more interesting
thing is this if you learn as a
caterpillar to crawl in a particular way
to receive um uh leaves which is your
food that memory is completely useless
to the butterfly the the butterfly
doesn't crawl it lives in a
three-dimensional space it has to fly
and it doesn't eat leaves it drinks
nectar so so so that memory is is is
useless so what has to happen is an
interesting kind of remapping which we
still really don't understand very well
although I have some some thoughts about
um what's going on but what's happening
is that it first has to generalize the
idea of from from leaves into a generic
category of food so generalizing from
specifics to to General categories is a
kind of intelligence so first it has to
generalize
and then it has to remap that
relationship the the the L the
association between the color and the
and the food concept onto a completely
new body which is driven by a completely
different set of effector signals so
that that to me is the more interesting
part is this remapping of information
and I think that that is just um just
the beginning I think I think once you
start looking for it that remapping is
everywhere it's there in evolution it's
there in um you know human um Co you
know cognitive science it's it's all
over the place yeah um and that I think
is is where this this strand of work
becomes particularly interesting when we
begin to look at how it begins to sort
of apply and have relevance and
resonance across so many different areas
and I I actually want to sort of come
back to this in a in a moment looking at
how we extend it but just sticking with
this um and the idea of these endogenous
bcal networks so I if I'm getting this
right if you're looking at sort of the
these Notions of memory that they're
embedded in this idea of of these um
these networks which are sort of
remembered by these clusters of cells
but then moving away from memory this
this seems to be profoundly important
that we can actually um or or biology is
effectively determined to a certain
degree by these these networks and we
can begin to engineer these networks so
then where does this this knowledge take
us in terms of understanding biology and
how we can actually utilize
biology um okay there's a there's a
short kind of short-term version and a
in a in a bigger picture here the the
short-term version is basically that now
that we understand how to respecify at
least a little you know we're beginning
to understand how to respecify these
pattern memories it means that we have
applications and and in my group we've
been we've been going after some of
these applications in birth defects in
um inducing regeneration and normalizing
tors um we can we can control morphology
uh at a at a at a much uh higher level
and what I mean by that is look imagine
let's just let's just take it into
behavior for a second imagine you had a
rat and you wanted this rat to do a
circus trick you know sit on a little
bicycle or something one thing you could
do if you if you took the bottom up
approach that's that's basically what
all of molecular medicine these days is
is about if you were you know wanted to
um micromanage all of this from the
hardware end you could try to figure out
exactly what the muscle motions need to
be to get the rat to sit on this thing
and then uh try to work it up to to see
upwards to see which neurons would have
to fire Trace that into the brain all
the circuits figure out uh what would
have to happen there and then figure out
what pixels on the rat's retina you
would have to activate with light
signals in order to get it to do the
behavior if you if you do that uh you'll
be here Till The Sun Burns Out right and
and that feels very much like
deterministic sort of science that we do
at the moment in fact it feels like an
awful lot of computer science how do you
build stuff out from scratch yeah y
exactly but but you know the the the
good news is that computer science kind
of has shown us the way the reason that
uh we you don't get out your soldering
iron when it's time to switch from
Photoshop to Microsoft Word is that we
Now understand that that the hardware is
only a part of the story and then you
commun if if your Hardware is good
enough you can communicate to it with
signals with reprogrammability all all
all that fun stuff so so what the thing
with the rat is that instead of that
what you can do is you can just train
the rat because the rat offers you this
amazing interface that does all the hard
work of translating your goals to the
rat's goals you're getting the buyin of
the of the agent the organism and it
does all the hard work of organizing its
downward component parts into um a set
of activities that are going to get with
your joint goals met so we've been
taking exactly that approach in biom
medicine to say that I I I don't want to
control all the cells I don't want to
talk to stem cells I don't want to
control gene expression I want the cells
to be motivated to take a journey in
anatomical space that goes from a wound
to a limb being regenerated versus um
scar tissue and this is exactly what
we've done for example in the Frog where
uh we can show that adult frogs which
normally don't regenerate their legs um
24 hours of stimulation with a with a
particular treatment that we came up
with gives you a year and a half of leg
growth after that we don't touch it at
all during that time the idea is not to
micromanage the process the idea is to
convince the cells that this is what
they want to do and they have all the
competencies uh about how to do it and
and of course this brings in this idea
of a gentle system which I I actually
find quite compelling um the the
understanding that are multiple layers
within biology you have systems with
agency and you're basically creating the
environment where that agency leads to
what you want yeah the the amazing thing
about um biology is that your whole body
every every level seems to be deforming
the the energy landscape for the level
below it to take advantage of their
competencies in navigating that that
landscape but to get it to go where you
want it to go um you know and uh and and
this is this is exactly what what we can
take advantage of not not because we're
we're so smart but but because that's
what the uh the the hardware is already
primed to do you know every every level
is already primed to do all these
interesting things if you can get the
incentives right right right so when you
look at this I this feels like it's it's
really powerful in terms of rethinking
biology so I know you've already hinted
at at this that that so much of what we
do is trying to engineer biology from
the the ground up whether it's the the
individual proteins or the genome
upwards but if and I'm paraphrasing here
but if we can persuade biology to do
what we want at a higher level I from
what my understanding is you're
cascading down so you sort of set the
top level parameters and each level
within the biological system will then
try and optimize within those
parameters yeah I think I think that's
right I think I think all of these uh
levels are made of uh it's sub agents
that solve problems in various uh spaces
um anatomical space physiological space
whatever and they have different
competencies and different agendas of
doing it and each layer is taking
advantage of this is I call it an
agential material because you have to
engineer it very differently than you
would uh you know engineer um the
passive or even active matter and and it
goes It goes even below cells I mean
we've been studying the um learning
capacities of molecular Networks so
never mind whole cells even the
molecular networks have probably at
least six different kinds of learning
capacity yeah and I I know you've used
the term intelligence with these these
systems um talk a little bit about how
you define intelligence because it feels
a little weird to talk about sort of
molecular systems having intelligence
yeah uh okay two two two important
things there one one is that um we need
to have some kind of way of talking
about molecular systems with
having intelligence because we have to
be able to tell a story of scaling we
all Start Life as a as a as an
unfertilized Osa a little blob of
chemistry and physics and if you if you
don't want to tell any kind of story
about intelligence with that system you
you're going to owe a uh some kind of a
a claim on when during embryonic
development this intelligence shows up
and there is no magic light lightning
bolt that at some point says okay you
weree physics but now you're a real mind
you know so that that does happen so so
we need we we know already given given
our origin as a collection of cells uh
and then then the Single Cell before
that we know we we have to come up with
some kind of a scaling Paradigm for how
intelligence scales from simpler forms
now the kind of intelligence uh that I'm
talking about is uh the kind that
William James defined as same goal by
different means so it's really a
navigational intelligence it's it's it's
a publicly observable uh perfectly sort
of empirically test able problem solving
capacity so this is not I am not talking
about Consciousness I am not talking
about um self-aware meta intelligence
where you know how intelligent you are
I'm not talking about any of that I'm
talking about the ability to navigate a
problem space and get your goals met
despite various new things that are
going to happen like how much competency
do you have at that and so the framework
that I'm developing is called Tam
technological approach to mind
everywhere uh this this framework the
most basic thing about it says the any
kind of intelligence claim two two
things about intelligence claims first
of all it it isn't a philosophical claim
it is an empirical testable experimental
claim so if you think some kind of
system has some kind of intelligence
what you're going to do is make a
hypothesis about a problem space uh
about the goal that it has about the
competencies you think it has and then
you're going to do perturbative
experiments to see if the the type and
amount of intelligence you've you've
ascribed it helps you uh have a a more
efficient relationship with that system
so it's it's not anything goes and we
don't just paint hopes and dreams on
rocks we we have very specific
hypotheses about problem solving
capacities and and that means that you
know yeah you you you can't just sort of
Imagine a spirit under every Rock but at
the same time you can't just assume that
cells don't have it uh you have to do
experiments and when we do experiments
we find you know we find amazing things
the other the other an unexpected we get
surprised which is what you want from a
scientific theory the other right the
the the other of that is when you make a
um a claim about the intelligence of
some system you're basically taking an
IQ test yourself because what you're
saying is as an observer as an external
Observer here is what I have noticed
this system can do and that doesn't mean
you didn't miss a whole bunch of other
things that you didn't notice just
because you don't see it doing things
doesn't mean that that it that
definitely doesn't do them right right
but even just talking about that you're
you're taking an approach where you say
at some point in the system there is
enough way awareness to be aware of of
yourself and what other systems are
doing so that must be a transition I I'm
assuming that that we're not looking at
intelligence in a way where cells have
got a concept of what molecules do it's
just that they have a specific type of
agency so is there a differentiator I
mean and I know we're getting into
Consciousness here we don't really want
to go down that down that path but it it
feels as you begin to talk about humans
sort of having that intelligence to
understand what's happening further down
the hierarchy that there needs to be
some degree of self-awareness or is that
just a an emergent property which
actually isn't that important well I'm
certainly not going to say
self-awareness isn't important I'm sure
it's important uh look uh the the thing
the thing is that I you you can you can
definitely have Bonafide intelligence
without uh without self-awareness right
you can have interesting learning
capacities uh you can be able to use uh
the tools that you have in novel and
creative ways you can you have delayed
gratification the capability of uh a
context sensitive attention you can have
all of these things without having that
kind of metacognitive self-awareness and
I think and I think it's perfectly good
intelligence I think self-awareness is
is a slightly different thing but I also
think that we have to have a continuous
notion of all of these things in other
words I don't believe there's a binary
um way that you can say yes self
awareness no self-awareness because uh
we're inevitably going to get to
questions of when and how during
embryogenesis and during Evolution that
supposed self-awareness shows up all of
these processes are extremely slow they
uh they go step by step there are no you
know people talk about phase transitions
but I have yet to hear other than
certain Quantum events I have yet to
hear an actual phase transition that's
really sharp when you sort of zoom in to
the to the key parameters I don't
believe any of these things are are
sharp face transitions yeah um and I
want to take a quick far into Artificial
Intelligence on that point before coming
back to the the biology because this
seems to be really important with
conversations and discussions around the
nature of intelligence with machines um
simply because what you've just
described seems to make it very easy to
to talk um in Practical terms about the
nature of intelligence with machines if
you forget about Consciousness and
self-awareness that ability to solve
problems along multiple Pathways seems
to be actually what we're developing
with a lot of machine learning at the
moment and to me it seems to simplify
those questions around what is
intelligence versus not when it comes to
AI does that make sense yeah it makes it
makes perfect sense I I don't think
there's any way I mean we can have some
arguments about Consciousness and so on
but I don't think there's any way to
argue that we do not have machines that
have that that have a a considerable uh
in some cases human level degree of
operational intelligence and so in the
case of problem solving I think what's
what's new nowadays is that in the past
that level of intelligence always went
along with a very long evolutionary
journey and uh and and and you know
certain other um properties that certain
are the cognitive properties and now
we've managed to to dissociate them
because I think current architectures
actually don't have a lot of the Key
Properties but but I absolutely think
they have uh the you know problem
problem solving intelligence and
whatever I mean I think it's important
to know that whatever the differences
between us and some kind of um AI
architecture whether it be the current
one or some some future one the answer
is not going to be what people often say
is that's just so so here are some bad
answers that's just a machine it
operates on the laws of physics and
chemistry well guess what so so do you
right that's you know and uh and um I I
know what it is because I built it and
it's just linear algebra you know I hear
I hear that sometimes too I you know I a
couple of couple of key key things here
which is that you know we we find
learning and memory in systems as simple
as a few genes that turn each other on
and off that's it a network of of of
differential equations that turn that
that represent genes turning each other
on and off never mind a whole cell never
mind you know the genome nothing already
can do pavlovian conditioning that's it
this this stuff starts very early on and
we found unex
expected problem solving capacities and
behaviors in something as as dumb as a
sorting algorithm you know so we're
talking bubble sort selection sort if
you look at them the right way you and
these things are deterministic six lines
of code there's there's nowhere to hide
there's no magic there's nothing there
people people you know people have been
studying these for for many decades and
if you look at them the right way you
find things that you did not know they
could do and you find things that are
literally not in the algorithm so that
tells me that we need to have a lot of
humility about saying that we know what
something does or or what something is
is capable of just because we know the
parts or just because we made it if a if
if a stupid bubble sword can do things
we didn't see coming you know what what
are these other things GNA do that and
it seems that that actually completely
changes the framing around how we think
about machine learning and and AI um
from our very deterministic perspective
where we say as you say we're building
the things so we know to the nth degree
what it does to understand understanding
we're Building Systems um which have got
inbuilt agency that it's it's not that
easy to to predict the outcomes of but
it also strikes me and I I correct me if
I'm wrong here because and you you
hinted at this earlier that quite often
when it comes to building machines or or
computer science we we try and predict
everything to the nth degree we try and
create the the perfect system and from
everything you're saying it seems like
that is totally the wrong approach and I
I think we're beginning to move away
from there but if we we approach these
systems as hierarchical um systems of of
AG gental sort of algorithms or whatever
um how far along are we in terms of
saying create a system where we create
the parameters that persuade the
subsystems to do what we want or is that
something we really need to focus more
on yeah um I think that if if we were if
we I I'm not sure we I'm not sure we
need to or want to do that the problem
is well well two things number one is we
we may get it without um without uh even
trying because um while we know I mean
people have been studying complexity and
emergence for a really long time it's
very you know or perverse instantiation
these kinds of things it's very easy to
make systems that follow simple rules
and then generate a bunch of complexity
that's not what I'm talking about I'm
not talking about unexpected complexity
or side effects or or or or any of that
I'm talking about emergent agency that
emergent goal directedness now uh I I I
will say that the architect that we have
today at least the ones of which I'm
aware um do do not maximize the kinds of
uh the kinds of dynamics that would lead
to that but but there are unexpected
ways of getting it that I I think we
need to be really careful about and the
bigger picture for me is that um I think
we have to be really careful about this
in the sense that like like I started uh
a few months ago I started writing a
paper to to lay out very clearly what
are the half a dozen things about
biology that are really critical for
making a true agents that matter in a
moral sense and what's different you
know how here's what biology is doing
that none of our computer architectures
are doing and and I stopped and I'm not
going to write that paper because I
think that um well not that it'll help
because somebody else will do it
eventually but but right so somebody
will catch on to the stuff but but but
but I don't want to be responsible for
it the the thing is that to whatever
extent I'm right in in uh having found
some of the key features that I think
make true um sensient beings that we're
going to need to take care of uh to that
extent uh I I don't want to be
responsible for creating you know
trillions of them and and having no
control over how they so but but
interesting that that's your thinking
it's not that this isn't something
that's possible it's that this is
something that's possible and we've got
to be really careful about what we do in
that space I I think it's absolutely
possible because the idea that what is
special about um Minds can only be
produced by a blind uh you know a
tinkering agent that makes mutations and
selects for certain things I I don't see
why that process would have a monopoly
on creating real minds I think that U
and I know there there are people that I
respect a lot who dis who disagree with
that you know Richard Watson I think is
one but to me uh I think that there are
many roads to doing this and at some
point we are going to figure out and I
think we already have a good uh basis
for figuring out what are the actual
policies and and components that are
that are necessary and they have nothing
to do with being made of protoplasm or
um any of the things that that we assume
uh you know are are tied to the biology
so so I absolutely think they can be
reinstantiate in other media and and
that resonates with with conversations
I've had with other people where you're
looking at almost substrate agnostic
systems it's it's the agency within the
systems um which is important so on that
I we've only got a few minutes left but
I wanted to sort of build on that and
extend out into much broader territory
um and I've have no idea sort of how
you'll respond to this but it's
something that fascinates me at the the
moment I looking at your work and
looking at work in in other disciplines
it feels very much as if we're on the
age of a on the edge of a scientific
Renaissance um I'm beginning to see
people challenging established Notions
of of how the world Works in many many
different areas and I don't know whether
that's just me seeing things through my
narrow perspective or whether it really
is the case that we're at a point in
human history where we're beginning to
rethink a lot of what we've assumed is
constant and and true um how I am I off
totally off base here I I would love to
uh agree with you and and my personal
experience is that as well but but I
have to correct for the fact that I
mostly hang around with you know with
people who who who like to think in
these big directions I mean I think
purely statistically you know when I
give talks about this stuff to a to a
generic random audience in conventional
Fields most of this is things they've
either never heard of or that sound
completely wrong to them on from a
philosophical level so I'm not sure how
um where we are in that transition from
this is impossible to this is completely
obvious I'm sure I I I think that's the
journey we're on I'm not sure where we
are there but but I agree with you that
that is that is where we're going this
is assuming we we all live long enough
this is going to overturn everything and
and it should because right our future
and and I one of the reasons I say that
is I I come across the same sorts of
conversations in multiple different
disciplines so if you're looking at at
Neuroscience if you're looking at at
Psychology if you're looking at at
physics even conversations where I'll
talk to respective physicists will say
you know I'm not sure the second law of
Thermodynamics is actually sort of
holding true um and it just feels like
there are cracks in the the real that
we've built over the last um sort of
several sort of decades or centuries I
and your work as part of that um and
there's definitely a sense that that
something is Shifting it it may take a
while before sort of broader audiences
actually see it um but it does feel as
if we're at one of those really
interesting points in history where
we're rewriting what we know about the
world and how it works I think that's
true and and the part of it that I'm the
most excited about in addition to the
Practical implications of you know for
biom medicine and and thing which I
which I think are huge uh is is I think
we have to start climbing out of the
what what some people have called um a
crisis of meaning so so so this is work
that I'm doing with with a number of
collaborators but you know you you can
sort of you can sort of see this this
this parabolic shape where uh
Neuroscience has told us some some
things that that things we thought about
um what we are and how Free Will Works
that are that are wrong uh evolutionary
theory has told us about this this sort
of every man for himself uh kind of
fundamental idea physics is telling us
about um determinism and and things like
that and so this progressively the the
loss of of these important ways of
thinking about the world has taken us
down in into a uh situation where a lot
of people are actually very disturbed by
it and we scientists and philosophers
need to now climb out we need to provide
provide the the other side of that
Parabola which I think does exist right
for recovering the things in a better in
a better way not in the you know sort of
incorrect way that that we've been
thinking about them but in a better um
more scientifically defensible way yeah
so final question before we we finish
then just on that um from your very
speculative perspective what does all
this mean to the future of Being
Human okay uh 60 seconds 60 seconds the
the the future the future is uh freedom
of embodiment and the future that I see
are children who are told uh stories of
the past where they say you've got to be
kidding me you mean you mean somebody
was born and just because of the
vagaries of some cosmic ray that hit
some DNA they had to stay in the body
that they were born with maybe they
wanted you know maybe their goals
required more IQ or longer lifespan but
no they got lower back pain and
stigmatism and then they you know they
died at 70 that that can't be right like
nobody can live like that that that's
the future I see that where that that
where we are now that becomes ridiculous
and and it should be it is ridiculous I
I I love that I perfect place to and a a
beautiful challenge um and I'm very very
aware we we need to to wrap up um there
are so many other areas we we could have
touched on um that the whole
regenerative medicine area um that not
being constrained by what we think our
bodies can do um we're going to have to
have you back at some stage happy but we
just looking at the time we should wrap
up so Mike thank you so much for this
this has been an incredible conversation
thank you and to those of you watching
that's it for this episode of the future
of Being Human unplugged obviously um
thank you so much again for Mike for
joining us and challenging our thinking
and what I think are some quite profound
and unexpected ways um if you want to
know more about Mike's work I should
check out the the links in the the blur
to this video or simply Google him and
you'll get a wealth of information and
finally if you're interested in joining
us for future conversations um please do
sign up for updates from asus's future
of beinghuman initiative at Future ofbe
human. asu.edu that's future of
beinghuman all one word.
asu.edu and with that thank you again
for joining us and have a great day
[Music]
Weitere ähnliche Videos ansehen
THE BIG IDEA (3): The Journey
Mary Evelyn Tucker 0101 - To See The Universe Not Just As A Place, But As A Story
【前編】次期ノーベル賞候補×成田悠輔 「寝不足だと嫌な奴になる?」「ほとんどの日本人の脳は低パフォーマンス状態?」数々の睡眠の落とし穴に成田絶句!世界的権威柳沢正史教授が語る快眠術とは?
「東大史上最高の天才」が挑む”究極の謎”。解明すればノーベル賞?「なぜ生命は動くのか」の答えとは…【ホリエモン×岡田康志】
Biosystems Informatics 1-1: Information Processing
Mary Evelyn Tucker 0203 A Communion Of Subjects, Not A Collection Of Objects
5.0 / 5 (0 votes)