Hod Lipson: Robots that are "self-aware"

Translator: Jordina Nogués Graell Reviewer: Elisabet Delgado Mas

So, where are the robots?

On són els robots?

We've been told for 40 years already that they're coming soon.

Fa 40 anys que ens diuen que ja venen.

Very soon they'll be doing everything for us.

Molt aviat faran de tot per nosaltres:

They'll be cooking, cleaning, buying things, shopping, building. But they aren't here.

cuinar, netejar, comprar, construir. Però no són aquí.

Meanwhile, we have illegal immigrants doing all the work,

Mentrestant, tenim immigrants il·legals fent tota aquesta feina,

but we don't have any robots.

però no tenim cap robot.

So what can we do about that? What can we say?

Què podem fer-hi? Què podem dir?

So I want to give a little bit of a different perspective

Vull donar una perspectiva un pèl diferent

of how we can perhaps look at these things in a little bit of a different way.

de com podem mirar-nos-ho d'una altra manera.

And this is an x-ray picture

Això és una imatge de raigs X

of a real beetle, and a Swiss watch, back from '88. You look at that --

d'un escarabat i un rellotge suís del 88.

T'ho mires i el que era veritat encara ho és.

what was true then is certainly true today.

We can still make the pieces. We can make the right pieces.

Encara podem fer les peces. Podem fer les peces exactes.

We can make the circuitry of the right computational power,

Podem fer els circuits del poder computacional exacte,

but we can't actually put them together to make something

però no podem posar-les juntes per a fer alguna cosa

that will actually work and be as adaptive as these systems.

que funcioni i sigui tan adaptativa com aquests sistemes.

So let's try to look at it from a different perspective.

Mirem-ho des d'una altra perspectiva.

Let's summon the best designer, the mother of all designers.

Invoquem el millor dissenyador, la mare de tots els dissenyadors.

Let's see what evolution can do for us.

A veure què hi pot fer l'evolució.

So we threw in -- we created a primordial soup

Hi posem - hem creat un caldo primigeni

with lots of pieces of robots -- with bars, with motors, with neurons.

amb moltes peces de robots- amb barres, motors, neurones.

Put them all together, and put all this under kind of natural selection,

Ho ajuntem tot i ho sotmetem a una certa selecció natural,

under mutation, and rewarded things for how well they can move forward.

a mercè de la mutació, i recompensem la capacitat d'avançar.

A very simple task, and it's interesting to see what kind of things came out of that.

Una tasca senzilla, és interessant veure què en resulta.

Si observeu, podeu veure un munt de màquines diferents

So if you look, you can see a lot of different machines

come out of this. They all move around.

que han sortit d'això. Totes es mouen,

They all crawl in different ways, and you can see on the right,

s'arrosseguen cada una a la seva manera

i podeu veure que n'hem fet un parell

that we actually made a couple of these things,

and they work in reality. These are not very fantastic robots,

que funcionen a la realitat. No són gaire sofisticats

but they evolved to do exactly what we reward them for:

però han evolucionat per fer el que els hem recompensat per fer:

anar endavant. Tot es va fer en una simulació,

for moving forward. So that was all done in simulation,

but we can also do that on a real machine.

però ho podem fer amb màquines reals.

Here's a physical robot that we actually

Aquí hi ha un robot que tenim

una població de cervells

have a population of brains,

competint o evolucionant en la màquina.

competing, or evolving on the machine.

És com un rodeo. Tots passegen la màquina

It's like a rodeo show. They all get a ride on the machine,

i són recompensats per com de ràpid o lluny

and they get rewarded for how fast or how far

they can make the machine move forward.

poden fer que la màquina avanci.

And you can see these robots are not ready

Aquests robots encara no estan llestos

to take over the world yet, but

per dominar el món,

però van aprenent com avançar

they gradually learn how to move forward,

i ho fan de manera autònoma.

and they do this autonomously.

En aquests dos exemples, teníem bàsicament

So in these two examples, we had basically

màquines que han après a caminar en una simulació

machines that learned how to walk in simulation,

and also machines that learned how to walk in reality.

i màquines que han après a caminar en la realitat.

But I want to show you a different approach,

Us vull ensenyar un altre enfocament,

and this is this robot over here, which has four legs.

aquest robot d'aquí,

que té 4 cames i 8 motors

It has eight motors, four on the knees and four on the hip.

4 als genolls i 4 als malucs.

It has also two tilt sensors that tell the machine

També té dos sensors d'inclinació

which way it's tilting.

que li diuen cap a on s'està inclinant.

Però no sap com és ell mateix.

But this machine doesn't know what it looks like.

You look at it and you see it has four legs,

El mireu i veieu que té 4 cames,

the machine doesn't know if it's a snake, if it's a tree,

però ell no sap si és una serp, un arbre,

it doesn't have any idea what it looks like,

no té ni idea de com és,

però intentarà esbrinar-ho.

but it's going to try to find that out.

Al principi fa moviments aleatoris,

Initially, it does some random motion,

intenta esbrinar com pot ser,

and then it tries to figure out what it might look like.

And you're seeing a lot of things passing through its minds,

i veiem un munt de coses que passen per les "seves ments",

a lot of self-models that try to explain the relationship

models d'ells mateixos que intenten explicar la relació

between actuation and sensing. It then tries to do

entre les accions i les percepcions.

Llavors intenta executar una segona acció

a second action that creates the most disagreement

que generi la major discrepància

entre les prediccions d'aquests models alternatius,

among predictions of these alternative models,

like a scientist in a lab. Then it does that

com un científic en un laboratori.

and tries to explain that, and prune out its self-models.

Ho fa i ho intenta explicar i reduïr els models.

This is the last cycle, and you can see it's pretty much

Aquest és l'últim cicle i podeu veure que pràcticament s'ha adonat com és.

figured out what its self looks like. And once it has a self-model,

Quan ja té el model de si mateix

it can use that to derive a pattern of locomotion.

pot fer-lo servir per derivar-ne un patró de locomoció.

So what you're seeing here are a couple of machines --

El que veieu aquí és un parell de màquines,

a pattern of locomotion.

un patró de locomoció.

We were hoping that it wass going to have a kind of evil, spidery walk,

Esperàvem que tingués un caminar maligne, aràcnid,

but instead it created this pretty lame way of moving forward.

però en canvi, ha creat aquesta manera patètica d'avançar.

But when you look at that, you have to remember

Quan mireu això heu de tenir present

that this machine did not do any physical trials on how to move forward,

que no havia fet cap assaig físic de com avançar

ni tenia un model de si mateixa.

nor did it have a model of itself.

Va esbrinar per si mateixa com era i com avançar,

It kind of figured out what it looks like, and how to move forward,

i ho va posar en pràctica.

and then actually tried that out.

(Aplaudiments)

(Applause)

So, we'll move forward to a different idea.

Avancem cap a una altra idea.

So that was what happened when we had a couple of --

Això és el que va passar amb un parell de...

that's what happened when you had a couple of -- OK, OK, OK --

Això és el que va passar amb un parell de...

D'acord, d'acord, d'acord.

(Laughter)

(Rialles)

-- they don't like each other. So

No s'agraden.

there's a different robot.

Aquí tenim un altre robot.

Això era el que va passar quan els robots

That's what happened when the robots actually

are rewarded for doing something.

van ser verdaderament recompensats pel que feien.

What happens if you don't reward them for anything, you just throw them in?

Què passa si no els recompensem per res, si els deixem estar?

So we have these cubes, like the diagram showed here.

Tenim aquests cubs, com es veu al diagrama.

The cube can swivel, or flip on its side,

El cub pot rotar o donar la volta sobre un costat.

Vam afegir 1000 d'aquests cubs al caldo, això és en una simulació,

and we just throw 1,000 of these cubes into a soup --

this is in simulation --and don't reward them for anything,

i no els vam recompensar per res.

we just let them flip. We pump energy into this

Els vam deixar donant voltes.

Si li insuflem energia, veiem que es donen un parell de mutacions.

and see what happens in a couple of mutations.

Al principi no passa res, només donen voltes.

So, initially nothing happens, they're just flipping around there.

Però al cap de ben poc, podeu veure que aquestes coses blaves

But after a very short while, you can see these blue things

on the right there begin to take over.

a la dreta comencen a dominar.

They begin to self-replicate. So in absence of any reward,

Es comencen a auto-replicar.

En absència de recompensa,

la recompensa intrínseca és l'auto-rèplica.

the intrinsic reward is self-replication.

N'hem construït un parell de veritat,

And we've actually built a couple of these,

and this is part of a larger robot made out of these cubes.

aquest és part d'un robot més gros fet d'aquests cubs.

It's an accelerated view, where you can see the robot actually

En aquest vídeo accelerat, podeu veure al robot

carrying out some of its replication process.

realitzant el procés de replicar-se.

So you're feeding it with more material -- cubes in this case --

Si l'alimentem amb més material, cubs,

and more energy, and it can make another robot.

i energia, pot fer un altre robot.

So of course, this is a very crude machine,

Per suposat, és una màquina rudimentària,

però estem treballant en una versió micro,

but we're working on a micro-scale version of these,

and hopefully the cubes will be like a powder that you pour in.

i esperem que algun dia els cubs siguin com un polsim.

OK, so what can we learn? These robots are of course

Així doncs, què n'aprenem?

Aquests robots no són gaire útils per si mateixos,

not very useful in themselves, but they might teach us something

però ens poden ensenyar com podem construir-ne de millors

about how we can build better robots,

and perhaps how humans, animals, create self-models and learn.

i potser com els humans o animals aprenen i creen automodels.

And one of the things that I think is important

Una de les coses que penso que és important,

is that we have to get away from this idea

és deixar la idea de dissenyar màquines artesanalment.

of designing the machines manually,

Ans al contrari, hem de deixar-les evolucionar i aprendre com els infants,

but actually let them evolve and learn, like children,

and perhaps that's the way we'll get there. Thank you.

i potser d'aquesta manera hi arribarem.

(Applause)

Gràcies. (Aplaudiments)