Our AI Ancestors: Dumbledore’s Portrait and Ray Kurzweil’s Father

Geoffrey Gordon Ashbrook
2 min readApr 29, 2023


As of 2023, H.sapiens-humans appear to be at the point where we can create some kind of AI version of deceased loved ones, though the quality may be a bit like MaxHeadroom or Pepperoni Hug Spot.

This is curiously similar to the portraits of past headmasters and headmistresses of Hogwarts in J.K.Rowlings wonderful tales, especially (assuming most people know the boy’s story) the case of Dumbledore’s portrait, and the delicate twilight situation that Harry was in where he wanted to be able to talk to Dumbledore, and he could in theory talk with the portrait, but the portrait would only be a kind of shadow of the person; and the tugging reminder of the gone but somehow not gone person you want to be less gone.

There could be a company, say AI-Ancestor.com, that could curate the various materials you have from your deceased loved one: writing, videos, audio, art, paintings, crafts, belongings, etc. And given enough training data, the AI-Ancestor might actually be not too far off in some ways, especially given how some people have nearly their whole lives recorded in digital format.

When I first heard about Ray Kurzweil’s touching desire to bring his father back in some way, I admit I was skeptical that this could ever be done properly. But after seeing how lucid large language models have become, and how (admittedly terrifying) short videos such as pizza commercials can be ‘dreamed’ by generative AI, given enough training data…it really is starting to look like Mr. Kurzweil will be able to speak with his father again. Perhaps he has already been working on this, getting higher and higher resolution versions.

It may even become standard practice for people to write in their wills if they want to open-source themselves for anyone to put into a talking portrait on the wall.

No doubt someone will raise an objection or two, in our civil society, but what if…what if.

About The Series

This mini-article is part of a series to support clear discussions about Artificial Intelligence (AI-ML). A more in-depth discussion and framework proposal is available in this github repo: