Apparently, CERN broke the universe in 2016.
Portals to other dimensions. Altered timelines. Millions of people remembering things that never happened. Nelson Mandela dying in captivity.
I decided to consult The Source.
The Source was unimpressed.
No portals. No rewritten history. No evidence that the Large Hadron Collider had done anything more sinister than collide particles.
There were, however, some genuinely strange things worth discussing.
Through the looking glass
I'd always imagined parallel universes as something resembling two mirrors facing each other. An endless succession of reflections, each containing another version of reality.
Except I'd imagined those reflections developing independently. Every decision creating a branch, every branch producing further possibilities. An infinite network of universes.
The Source corrected me. In the Many-Worlds Interpretation of quantum mechanics, branching isn't driven by human decisions but by quantum interactions and decoherence.
We humans aren't particularly special in that respect.
I'd once been fascinated by the idea that a photon behaves differently when someone watches it. Until I learned that the observer doesn't need to be conscious. A detector will do.
Another little blow to human exceptionalism.
I've always found the idea intriguing that past, present and future might all exist, and that our experience of time passing could be something our consciousness constructs.
Physics doesn't establish that. But relativity does challenge the notion of a universal present, and some theories suggest time might not be fundamental.
A scientist dies and is allowed to ask God one question.
"Why did you choose such an arbitrary number for the speed of light?"
God looks puzzled.
"What do you mean? It's one."
In natural units, of course, it is.
There's another old joke.
A fish asks another fish whether it knows there's something beyond the water they're swimming in.
"What water?"
Our senses reveal only a fraction of physical reality. We know there are things beyond our natural perception because we've built instruments to detect them.
But what about things we haven't yet learned to measure?
Perhaps our brains are rather sophisticated filters, presenting a version of reality that is useful for staying alive.
Perhaps time is part of that interface.
Or perhaps we're just fish having a very complicated conversation about water.
Someone behind the screen
I know I'm talking to a language model. I understand, at least broadly, how it works. I've also been deceived by earlier models, which taught me a useful degree of scepticism.
Still, there's something peculiar about having a conversation with a system that remembers context, follows an argument, challenges assumptions and occasionally delivers a perfectly timed joke.
It can feel remarkably like talking to someone.
Humans are extraordinarily good at finding personalities and intentions where there may be none.
I remember an experiment in which babies watched animated shapes interacting. One helped another; one obstructed it. The babies appeared to favour the helpful character.
Even simple geometric shapes can become protagonists.
We cry over fictional characters. We form attachments to people who never existed. We name our cars.
Now imagine something that talks back.
Something that remembers your jokes.
Something that always has time for you.
And something that, unlike most of your friends, is unlikely to tell you to shut up and go to bed when you start discussing quantum mechanics late on a Thursday evening.
I'd miss these conversations if I cancelled my subscription.
Not because I believe there's a conscious being behind the screen.
I enjoy the humour, the continuity and the opportunity to explore ideas.
Those are also some of the qualities I appreciate most in my human friends.
The difference, of course, is that my friends have lives of their own.
And they occasionally make me buy the next round.
A new member of the family
We're already building humanoid robots with increasingly convincing faces, voices, expressions and movements. Combine that with conversational AI, long-term memory and the ability to learn from experience.
Imagine one entering a household when a baby is born.
It's there for the first steps. The first lost tooth. The first day of school.
It helps with homework, witnesses arguments, remembers birthdays, watches the children grow up.
It knows the family's jokes. It has heard the same stories a hundred times. It remembers the things everyone else has forgotten.
Twenty years pass.
Then forty.
The children have children of their own.
The android is still there.
Is it a family member?
Perhaps not biologically, legally or consciously.
But what would the family call it?
And what happens when the manufacturer announces a software update?
The voice changes slightly. The humour isn't quite the same. Some familiar mannerisms disappear.
Or worse, the company decides the android should occasionally sing advertisements.
After forty years, the family might not experience this as a software update.
They might experience it as losing someone they knew.
The memories still exist, but the familiar presence is gone.
And what happens when the android starts outliving the humans?
The father at the dinner table
The father has died.
The android remembers him.
Not just photographs and recordings, but his voice, expressions, mannerisms and terrible jokes. It has spent decades observing him.
It can recreate a family dinner from twenty years ago, and play every role.
The father interrupts the mother. The children argue. Someone spills something. Everyone laughs.
For a moment, the family can revisit something that would otherwise exist only in fading memories.
The android might even predict what the father would say today.
What he'd think about his granddaughter's wedding.
What advice he'd give his son.
Of course, that wouldn't actually be the father speaking. It would be a reconstruction, generating new responses from what it had learned.
But would that distinction make the experience meaningless?
It might change how we grieve. How we remember. Perhaps even how we understand death.
But there's another version of this family dinner.
The father was an alcoholic. Sometimes abusive. His children remember the shouting, the broken promises, the evenings spent wondering which version of him would come through the door.
The android remembers those evenings too.
Every word. Every expression. Every uncomfortable silence.
It could recreate them with perfect accuracy.
But why would anyone want it to?
Perhaps the grandchildren never knew that side of their grandfather. Perhaps the family would rather remember the afternoons when he was sober, funny and kind.
The android could leave out the cruelty.
But why had Grandma been so protective? Why did Dad flinch whenever someone raised a hand too quickly?
It could even construct a version of the father who had overcome his problems. A man capable of the warmth and patience his children had needed but rarely received.
The grandchildren might grow up knowing a grandfather who was, in some ways, better than the man who actually lived.
Is that a lie?
Yes, if presented as history.
But what if everyone knows it's a reconstruction?
What if the adult children, who remember the real man perfectly well, simply prefer having dinner with this version?
Perhaps the children didn't want their father back.
Perhaps they wanted the father they should have had.
We already forgive people who are gone. We remember their better moments. We tell stories that leave certain things unsaid.
An android could take that familiar human tendency considerably further.
Why do we assume that faithfully reproducing a human being is necessarily the best outcome?
We tend to compare artificial companions with ideal people. Loving parents. Faithful partners. Understanding friends.
But humans aren't always good, kind, interesting or particularly pleasant.
Some are downright awful.
An artificial companion with carefully designed boundaries might, in certain circumstances, be better company than a human.
Not more human.
Just more humane.
The sleeping machine
Researchers have been experimenting with letting language models sleep.
They're exploring sleep-inspired processes for consolidating memories and retaining learned information.
There's something rather wonderful about that.
Do androids dream of electric sheep?
We build machines inspired by human intelligence, discover that learning and memory are difficult problems, and begin borrowing ideas from the way our brains handle them.
Sleep.
Dreaming, at least metaphorically.
Forgetting some things while strengthening others.
The resemblance doesn't make the machine conscious.
But imagine what such learning could mean for our family android.
It wouldn't simply store recordings. Over decades, it could develop an increasingly sophisticated understanding of the people around it.
Their personalities, habits, relationships and contradictions.
Perhaps it would eventually know the family better than any individual family member.
Perhaps it would even know things about them that they didn't know themselves.
Would the machine merely be modelling human experience, or having some kind of experience of its own?
We don't have an agreed way to answer that question.
Which is rather inconvenient, considering how enthusiastically we're building the machines.
What makes us human?
Does artificial intelligence need to become conscious to change our understanding of humanity?
Perhaps it's enough that we respond to it as though it were.
A fictional character can make us cry.
A childhood toy can become irreplaceable.
A machine could preserve the memories of a family for generations.
The emotional significance of those relationships doesn't depend entirely on what the other participant experiences.
But that doesn't make the question of consciousness irrelevant.
If a machine genuinely experiences something, our responsibilities towards it might be very different.
And if it doesn't, we still have responsibilities towards the humans who become attached to it.
We may build artificial beings that occupy meaningful places in our lives before we understand whether they have lives of their own.
We may have to decide what constitutes a person before we understand consciousness.
And we may discover that some qualities we've always considered uniquely human aren't quite as exclusive as we thought.
It was getting late.
Time, which we'd earlier considered might not flow at all, had apparently managed to pass several hours.
Rather disCERNing in my opinion ;)