Black Boxes 

Shining Light

Voice Over Script

A.I. this 

A.I. that

What is Artificial Intelligence ?

Being a Scientist, a Worker or a Government,

 Nowadays everybody is talking about it.

Is it a brave new world ?

Is it the end of the world ?

Is everything changing or is it all the same?

Being an array of cogs or binary code.

 We talked to machines through careful instructions.

If this Then that

And this

Or That
Nor This
Nand That

And so on

And that means that a machine can only be as good as our instructions.

The best thing about computers is that they can do exactly what you tell them to do

The worst thing about computers that they can do exactly what you tell them to do

We are trying to explain to them a world,

we ourselves  barely understand.

But why not explore this world together?

In order to do that we need to ask:

How do we think about the world?

Yet alone why?

Difficult questions.

Maybe we should start with 

how we think we learn.

1. Evolution

a.k.a Reinforcement Learning

With enough time and random mutations 

the possibilities are endless

In other words:

 “Do random things and stick to 

what works at a given time”

A simple instruction 

with cosmic proportions.

It’s truly striking to think that:

 “As far as we know, everything that exists

  -being a cat or a chair- 

emerged out of the near infinite combinations 

of the basic elements of the big bang:

Helium, Hydrogen and a little bit of Lithium”

But what does all that have to do with machines and learning? 

One of the main ways we help machines to learn about the world

 is directly inspired by this phenomenon.

Emerging Complexity

Instead of directly providing the machines with the solution to our problem, 

-in order to solve it faster than us-

we let them find the solution on their own. 

Enabling them to get creative.

In this approach of machine learning

 We define a goal and a framework of action towards it.
The closer it gets to the goal 

the greater the reward.

-or the lesser the punishment-

The machine starts doing random attempts to get to the goal 

and fails miserably.

But with each tiny step closer to the desired outcome,

 it gets to keep this miniscule knowledge for its next attempt. 

The next attempt might be much better or way way worse.

But over many generations 

with every attempt 

knowledge is accumulated 

until the machine comes up with the best fitting solution 

-given the circumstances-. 

Knowledge that lasts a lifetime

But also knowledge fixed within that lifetime

2. Solitary Experience

a.k.a Unsupervised Learning

As evolutionary knowledge traversed through the vastness of time,

It expanded the multiplicity of being that we curiously observe and interact with.

 It also created the circumstances for new kinds of learning. 

A kind of knowledge that doesn’t necessarily depend on

 multiple generations in order to emerge.

When we are born in the world. 

We are exposed to an inconceivable amount of information.

We experience being, in its overwhelming- unfiltered – totality.

Everything all at once.

[Microphone and camera sensitivity play]

In order to make sense of it,

and be able to act in any way.

We start to adjust ourselves by sorting reality in abstract patterns.

Sounds that trigger fear

Things that feel hot.

Actions that lead to pleasure 

Inspired by this existential wonder

We let machines discover patterns in disorganised information.

The catch is that 

with the same information 

We can come up with many different patterns.

And for humans and machines alike 

patterns can either be insightful or narrowing.

There’s always something left out for better or for worse.

A pattern is not certain but probable.

A pattern is a matter of perspective.

3. Collective Experience 

a.k.a Supervised Learning 

We learn through living

We learn through isolating information

We also learn together

Knowledge is passed through generations  

But not only through genetic instructions

Knowledge can be shared

Experience that spans millennia

Beliefs that come and go

Epiphanies and Obscurities 

How do you explain how a cat looks like to a baby 

or a machine ?

Is it its shape?

Its colour maybe?

Does it have to do with where you usually find it?

How do you define a cat ?

You keep showing it examples of what a cat is

And at some point it gets it.

Do you share the same definition of a cat now?

Yes and No

What is a cat can now be successfully communicated

And both parties can now equally redefine it.

?. Entering the Black Box

When we decide to interact with machines 

beyond telling them exactly what to do

Things become even more complicated

We give them agency

We let them learn on their own

And with agency comes unpredictability

In theory of systems

Being our brain, a social structure, an idea 

Or  an Artificial Intelligence Algorithm

A metaphor emerges

The metaphor of the “Black Box”

A System can be considered a black box 

When it is being viewed in terms of its inputs and its outputs

But without any knowledge of  its internal workings

You see what’s getting in 

You see what comes out 

But what is happening inside is obscured

Either  because you cannot understand it 

 or intentionally

Black Boxes training Black Boxes

We dont fully know know how we think or learn

Yet inspired by our limited understanding 

We try to make machines teach themselves 

In order to start learning together

To move beyond step by step instructions

But there is a gap between human and machine learnng

We are constantly exposed to an inconceivable amount of information

Cosmic and cultural

Whereas the trillions of data, witnessed by the machines

-so far-

Are curated by human limits and intentions

When a machine learns

It tries to make sense out of the immense -but limited- information it’s exposed to

To make sense of what it sees

To understand what it hears

To recognize cancer in trillions of cells

To find patterns in the language of whales

When a  machine creates

It tries to make decisions according to the immense-but limited- information  it has learned from

To imagine how something could look or sound like

To predict which word could come after another

To speculate which combination of molecules can be a cure for a disease

To suggest potential targets for an airstrike

And how are “we” anyway?

Humans and machines are not so different after all

They both can be black boxes

We don’t always know what;s happening inside them

And so our perspective can only be temporary and partial

But these unknowns is what enables us to learn together

So far we haven’t found a way 

to know exactly how a human or a learning machine is thinking

We expose our body to a doctor

Even if we don’t know how medicine works

We drive a car that with the slightest wrongdoing can kill us

Without knowing entirely how it functions

Should we all be doctors or engineers?

Should we dismiss them altogether?

What makes us trust a human or a machine ?

And who are “we” anyway ?

Is it us as “humans” or living beings?

Is it the loud western world?

Does  this “we” have a gender?

We live in a world founded on division and oppression

And as long as this is the case

Science and technology will be created on these premises

Forgetting about the silenced

Amplifying the loud ones

And in the end

Who am I even to speak about all these?

I’m just a nerd born in Cyprus with arbinite roots

 talking to humans and machines

In an attempt to figure out our worlds