Who was John Vlahos?
John James Vlahos was born on 4 January 1936 and raised by Greek immigrant parents in Tracy and later Oakland, California. He studied economics at UC Berkeley, became managing partner of a major San Francisco law firm, and for decades announced games from the press box at Berkeley’s Memorial Stadium.
He was also a devoted Gilbert and Sullivan performer and led the Lamplighters, a light-opera company, for 35 years. His son describes a man who loved long, old-fashioned words, grammar rules, catchphrases and jokes told a hundred times.
On 24 April 2016 he was diagnosed with stage IV lung cancer that had spread to his bones, liver and brain.
First, a recorded life story
In May 2016 James Vlahos, a journalist who had written about conversational AI, sat down with his father and an audio recorder. His father brought a handwritten outline: family history, education, career and more. They recorded more than a dozen sessions, each lasting an hour or more.
The transcripts came to 91,970 words, or 203 single-spaced pages. That oral history, not chat logs, became the raw material for the bot. It is worth noticing that the most valuable step was the simplest: asking questions and pressing record. Our guide to preserving a loved one’s voice and messages explains how to do the same today.
How the Dadbot was built
Vlahos had learned that PullString, a conversation company started by former Pixar staff, was about to release its authoring software. In August 2016 he told his parents and siblings about the idea. His father agreed with a tired shrug; his mother and siblings were more enthusiastic.
The Dadbot was not a neural network trained on his father. It was a large, hand-written script, built over months:
| Element | How it worked |
|---|---|
| Structure | A “Conversation Hub” led to chapters of his father’s life, such as Greece, childhood in Tracy, college and career, plus songs and jokes. |
| Listening | Vlahos wrote rules predicting what a user might type, and fallbacks for anything unexpected. |
| Words | Lines came from the transcripts, so catchphrases and opinions were his father’s own. Hundreds of variants kept common replies from repeating. |
| Voice | Short audio clips of his father’s real voice, telling stories and singing, were added to the chat. |
| Personal touches | It responded differently to good or bad moods, knew family birthdays and had a simple sense of the time of day. |
| Home | In November 2016 he published it on Facebook Messenger, text only at first. |
Because every line was scripted, the Dadbot could not invent facts in the way a modern language model can. The price was that it often gave vague answers or closed a topic too quickly.
The questions Vlahos could not answer
In his article, Vlahos lists dilemmas that anyone building a simulation of a real person still faces. They make a good checklist:
- How much can you shorten and rearrange someone’s words before they stop being theirs?
- Should the bot say only what the person said, or also what they would probably say?
- How do you keep it true to the whole family, not just to the person who built it?
- Should it ever admit that it is a computer?
- Should it know that the person is ill, respond to grief, or say “I love you”?
His wife, Anne, who was very close to his father, had strong reservations from the start. After first talking to the bot a month after his father’s death, she said she still found it jarring.
His father’s verdict
On 9 December 2016 Vlahos showed the prototype to his parents. His mother chatted with it for nearly an hour while his father watched. Some replies were vague, but at times it held a real conversation, recalling a family trip to Greece and old childhood stories. Asked whether it felt like him, his father said it did: “Those are actually the kinds of things that I have said.”
Later, asked whether the bot gave him any comfort, his father said that he already knew all of these stories himself, but that he valued it because it could share them with his family, especially his seven grandchildren.
John Vlahos died at home in February 2017. The night before, his son opened the Dadbot on his phone and let it tell him a story.
From Dadbot to HereAfter AI
WIRED published “A Son’s Race to Give His Dying Father Artificial Immortality” in July 2017. In March 2019 Vlahos published Talk to Me, a book about voice computing that includes the Dadbot story.
He went on to co-found HereAfter AI, which turned the Dadbot idea into a service: people recorded spoken answers to interview prompts, and family could later ask questions and hear the closest real recording. MIT Technology Review tested it in 2022 with a reporter’s parents. In mid-2026 HereAfter AI announced it was shutting down. Our HereAfter AI comparison explains what changed and what former users can do.
What the Dadbot shows, and where Memory Clone fits
The Dadbot shows that consent and preparation make a difference. John Vlahos chose what to tell, knew what it was for and saw the result. Compare it with Project December, where a generative model improvised freely, and with the wider history of griefbots.
Memory Clone takes a different route: it builds a private AI simulation from chats you already have, such as WhatsApp or Messenger exports, and learns how someone wrote to each person separately. It generates new text, so it can be wrong, and it is always presented as AI. It is a Windows app in alpha and does not use voice recordings today. If a parent is still here, recording their stories first is worth more than any app. When you are ready, you can learn more on the download page.
If you are in crisis or thinking about harming yourself, contact your local emergency number now. You can also reach: UK & Ireland Samaritans, 116 123; United States 988 Suicide & Crisis Lifeline, call or text 988; elsewhere find a free, confidential helpline at findahelpline.com.
Questions fréquentes
Was the Dadbot an AI trained on his father?
Not in the modern sense. It was built with PullString’s scripting tools: Vlahos wrote the rules and replies by hand, using his father’s own words from the interview transcripts.
Did John Vlahos agree to the Dadbot?
Yes. His son explained the idea to him and the family in August 2016, and his father agreed. He later saw a demonstration and said it sounded like things he had said.
Where can I read the original story?
James Vlahos told it in WIRED in July 2017, under the title “A Son’s Race to Give His Dying Father Artificial Immortality”, and again in his 2019 book Talk to Me.
Is HereAfter AI still available?
In mid-2026 HereAfter AI announced it was shutting down and asked account holders to request their recordings. When we checked in October 2026 its website was offline.
Sources
Vérifié le . Les produits tiers et leurs pages d’aide changent souvent ; suivez les liens pour obtenir les informations à jour.
- A Son’s Race to Give His Dying Father Artificial Immortality (James Vlahos, 2017) (WIRED)
- Talk to Me: How Voice Computing Will Transform the Way We Live, Work, and Think (James Vlahos, 2019) (HarperCollins)
- Technology that lets us “speak” to our dead relatives has arrived. Are we ready? (Charlotte Jee, 2022) (MIT Technology Review)
- Using AI, cartoonist Amy Kurzweil connects with deceased grandfather in “Artificial” (Chloe Veltman, 2023) (NPR)
- HereAfter AI homepage with shutdown notice (archived 21 July 2026) (HereAfter AI via Internet Archive)
This article is general information, not medical, legal or financial advice. Memory Clone creates an AI simulation from the material you provide; it is not the person.