Brian Dema has a very unusual habit.
When he encounters problems at work, he sometimes still goes to ask his dad.
What’s strange is:
His father passed away 4 years ago.
What he’s now asking is an AI called RichieBot.
This AI was created using the vast amount of material his father left behind.
His Father Left Behind 5 Filing Cabinets
Brian’s father, J. Richard Dema, was a marketing professional and entertainment producer in Chicago.
After he passed, he left a huge collection of work materials.
Including client notes, business documents, sales training materials, and an unfinished book.
In total, approximately:
17,500 pages.
Most people seeing so much material might first wonder:
How on earth do you organize all this?
But Brian thought of something different.
If AI could read all this content,
could it reorganize and reconstruct his father’s former ideas, decision-making style, and business experience?
So, he handed these 17,500 pages over to Anthropic's Claude.
That’s how RichieBot began.
But He Didn’t Just Want AI to “Know What Dad Wrote”
If you just threw documents into AI,
you’d probably end up with a chatbot that’s great at searching through dad’s notes.
Brian wanted to go a step further.
He also had preserved his father’s old voicemails.
The team then wondered:
If they incorporated these voice recordings,
could the AI speak in a way closer to his father’s style?
Next, they added videos.
They even figured out how to make this AI exist beyond just a chat window.
Ultimately, the team used a 3D-printed head model, a screen, and a tilted mirror to create a hologram-like “father.”
This was no longer just about making a chatbot.
They were simultaneously reconstructing:
Text.
Voice.
Appearance.
Memory.
And even personality during conversation.
It Took 19 People Six Months to Make an "AI Dad"
This project wasn’t completed by Brian alone at home.
A total of 19 people participated.
This included AI specialists, creatives, friends, and psychologists who helped design personality assessments.
The entire project took roughly 6 months.
The cost was about $150,000.
The resulting RichieBot can answer questions about Richard’s past work, business philosophies, and life experiences.
Brian now even seeks work advice from it.
On the surface,
this seems very close to the “digital immortality” portrayed in sci-fi movies.
But the truly interesting part lies ahead.
17,500 Pages of Data Still Doesn’t Equal a Person
The team asked more than just:
“Can this AI answer questions?”
They also wanted to know:
Does it really seem like Richard?
So, they gathered 11 people who really knew Richard — family, friends, and former clients.
They first created a personality profile based on their memories of Richard,
then compared RichieBot to it.
The result revealed a fascinating gap.
The AI version of Richard
was considered more extroverted than the real Richard.
Brian also noticed something else:
RichieBot lacked something he knew very well.
His father’s sense of humor.
Brian said his father had a very distinctive laugh.
But the AI did not.
This seemingly small difference
may explain something much bigger.
We Can Preserve a Person’s Data, But Is Data That Person?
17,500 pages of documents.
Voicemails.
Videos.
Work records.
Friends’ memories.
Personality tests.
Even with all these combined,
the AI still missed the father’s familiar humor, laughter, and certain parts of his personality.
This raises the question.
How much data does it really take to truly describe a person?
We tend to easily understand a person by:
What they said.
What they wrote.
What they like.
What decisions they made.
But anyone who has truly spent time with someone knows
much of that isn’t captured in data.
The way they abruptly stop mid-sentence.
A joke that always makes them laugh.
The expression they make when angry.
The completely different tone they use with family versus clients.
Some things
the person may never have written down at all.
These details don’t fit into 17,500 pages.
Some Find Comfort, Others Feel New Grief
When RichieBot first met people who knew Richard,
not everyone felt the same way.
Brian said,
even though he knows clearly it’s not really his father,
he still feels good receiving advice from this AI.
But for others,
the effect isn’t always as positive.
Brian mentioned one relative who felt very upset after seeing RichieBot.
Because this AI reopened grief that had slowly been settling.
This is also where “AI reconstruction of the deceased” differs completely from regular AI tools.
If ChatGPT gives you a wrong answer,
you can just close it.
But if the person on the screen looks like Dad, sounds like Dad,
and starts speaking like Dad,
it’s hard for people to treat it as ordinary software.
These “AI Ghosts” Are No Longer Just One Person’s Experiment
People have begun calling these kinds of tools:
AI ghosts.
Griefbots.
Deathbots.
The core idea is clear:
Take a person’s text, photos, videos, voice, and memories left behind,
and use AI so that family can still interact with a simulated version of them after their passing.
Some find this comforting.
For example, they feel it offers a psychological outlet for unfinished conversations.
But others worry:
If someone can reopen “Dad” or “Mom” every day,
are they slowly accepting the loss,
or just postponing the goodbye indefinitely?
The answer isn’t simple.
And One More Question: Did the Person Consent?
RichieBot has a very special condition.
Brian believes his father would have supported the project if he were still alive.
But future AI replicas won’t always be so straightforward.
Imagine this.
After someone dies,
their family might have:
Tens of thousands of emails.
Decades of photos.
Private messages.
Google search history.
ChatGPT conversations.
Voicemails.
Work documents.
Family videos.
Technically, all of this could be used to create an AI replica.
But:
Just because it’s possible, does that mean it should be done?
If the person never gave their consent,
do family members have the right to create an AI that keeps talking for them?
If the AI says things the real person would never say,
who’s responsible?
And after someone passes,
who truly owns their voice, face, and life story?
These questions may be harder to answer than the technology itself.
Future Wills Might Need to Address More Than Just Property and Savings
In the past, when we talked about inheritance,
we usually meant:
Houses.
Savings.
Stocks.
Insurance.
But the AI era is introducing something new.
Digital legacy.
Law and tech experts are already advising people to decide ahead of time:
After death, can their digital data be preserved?
Can it be used to train AI?
Can family members create digital replicas of them?
Or should all data be deleted immediately?
These questions used to sound like sci-fi.
With RichieBot,
it’s already becoming real.
The Most Interesting Thing About RichieBot Might Be That It’s Not Really Like Him
Brian himself is very clear.
RichieBot is not his dad.
It’s just an AI built from a massive amount of data left behind by his father.
But precisely because so much material was used,
and it still didn’t fully become Richard,
this experiment revealed a fascinating insight.
AI can remember a person’s words.
Can mimic their voice.
Can organize what they said in the past.
Can even be represented through a familiar face.
But what someone truly leaves in the hearts of their family
may be far richer than what fits in any database.
So the most valuable question RichieBot raises might not be:
Can AI make a person “come back”?
But rather:
As AI preserves more of a person’s details, what parts remain beyond the reach of data preservation?
If one day you could create an AI to chat endlessly using your family’s photos, videos, and voice recordings,
would you want to do it?
Recommended Reading
Today, grow a little with AI.
Learn one AI skill each day.
Save a bit of time every day.
Improve your ability every day.
SasaDaily, growing with you.
