How close are we to uploading our minds? - Michael S.A. Graziano

View full lesson: https://ed.ted.com/lessons/how-close-are-we-to-uploading-our-minds-michael-s-a-graziano

Imagine a future where nobody dies— instead, our minds are uploaded to a digital world. There they could live on in a realistic, simulated environment with avatar bodies, calling in and contributing to the biological world. Mind-uploading has powerful appeal— but what would it actually take to scan a person’s brain and upload their mind? Michael S. A. Graziano explores the challenges.

Lesson by Michael S. A. Graziano, directed by Lobster Studio.

Imagine a future where nobody dies—

instead, our minds are uploaded
to a digital world.

They might live on in a realistic,
simulated environment with avatar bodies,

and could still call in and contribute
to the biological world.

Mind uploading has powerful appeal—

but what would it actually take to scan a
person’s brain and upload their mind?

The main challenges are scanning a brain
in enough detail to capture the mind

and perfectly recreating
that detail artificially.

But first, we have to know what to scan.

The human brain contains
about 86 billion neurons,

connected by at least a
hundred trillion synapses.

The pattern of connectivity
among the brain’s neurons,

that is, all of the neurons and
all their connections to each other,

is called the connectome.

We haven’t yet mapped
the connectome,

and there’s also a lot more
to neural signaling.

There are hundreds, possibly thousands
of different kinds of connections,

or synapses.

Each functions in a slightly
different way.

Some work faster, some slower.

Some grow or shrink rapidly in
the process of learning;

some are more stable over time.

And beyond the trillions of precise,
1-to-1 connections between neurons,

some neurons also spray out
neurotransmitters

that affect many other neurons at once.

All of these different kinds of
interactions

would need to be mapped in order to
copy a person’s mind.

There are also a lot of influences on
neural signaling

that are poorly understood
or undiscovered.

To name just one example,

patterns of activity between neurons

are likely influenced by a type
of cell called glia.

Glia surround neurons and,
according to some scientists,

may even outnumber them
by as many as ten to one.

Glia were once thought to be purely
for structural support,

and their functions are still
poorly understood,

but at least some of them can generate
their own signals

that influence information processing.

Our understanding of the brain isn’t good
enough to determine

what we’d need to scan in order
to replicate the mind,

but assuming our knowledge does
advance to that point,

how would we scan it?

Currently, we can accurately scan a living
human brain

with resolutions of about half a
millimeter

using our best non-invasive
scanning method, MRI.

To detect a synapse, we’ll need to scan
at a resolution of about a micron—

a thousandth of a millimeter.

To distinguish the kind of synapse and
precisely how strong each synapse is,

we’ll need even better resolution.

MRI depends on powerful magnetic fields.

Scanning at the resolution required

to determine the details of
individual synapses

would requires a field strength high
enough to cook a person’s tissues.

So this kind of leap in resolution

would require fundamentally
new scanning technology.

It would be more feasible to scan a dead
brain using an electron microscope,

but even that technology is nowhere
near good enough–

and requires killing the subject first.

Assuming we eventually understand the
brain well enough to know what to scan

and develop the technology to safely
scan at that resolution,

the next challenge would be to recreate
that information digitally.

The main obstacles to doing so are
computing power and storage space,

both of which are improving every year.

We’re actually much closer to attaining
this technological capacity

than we are to understanding or scanning
our own minds.

Artificial neural networks already run
our internet search engines,

digital assistants, self-driving cars,
Wall Street trading algorithms,

and smart phones.

Nobody has yet built an artificial network
with 86 billion neurons,

but as computing technology improves,

it may be possible to keep track
of such massive data sets.

At every step in the scanning and
uploading process,

we’d have to be certain we were capturing
all the necessary information accurately—

or there’s no telling what ruined
version of a mind might emerge.

While mind uploading is
theoretically possible,

we’re likely hundreds of years away

from the technology
and scientific understanding

that would make it a reality.

And that reality would come with ethical
and philosophical considerations:

who would have access to mind uploading?

What rights would be accorded to
uploaded minds?

How could this technology be abused?

Even if we can eventually upload
our minds,

whether we should remains
an open question.