We can clone you wholesale: Boffins build ML agents that respond like specific people Computer scientists have devised a technique for making AI models behave like specific people. The researchers from Stanford University, Northwestern University, the University of Washington, and Google DeepMind describe their work in a pre-print paper titled "Generative Agent Simulations of 1,000 People." The US-based …
"I'm a personality prototype.
You can tell can't you?"*
OTOH, I do have a brain the size of a planet, so it's not all bad.
I'll get my coat, it's the one with a strong Brownian motion producer in the pocket.
*HHGTTG, Marvin the 'paranoid android'. But it applies to oh so many, many, alleged 'people' I've met.
"We anticipate that within our lifetimes it may become common practice for people to create a custom AI agent to interact with [...] the [...] world after death"
Technological necromancy? What a ghastly, even sacrilegious idea!
Requiescat in pace - less likely if the deceased's AI simulacrum is still abroad in the land of the living.
Would we require the technological equivalent of a wooden stake and interment at crossroads to keep the dead from haunting the living?
I can imagine that the data from pre-employment psychometric tests that are apparently de rigueur in many large companies could be used to construct these models without the explicit consent of the candidate.
Would we require the technological equivalent of a wooden stake and interment at crossroads
Given Google's involvement, I think you can rely on their return to the cemetery in fairly short order.
I don't think there's much harm in someone communicating with facsimiles of the dead if it comforts them, just as long as they accept it's only a facsimile. And, in that regard, I suspect the uncharacteristic tendency of their "loved one" to remind them of the bargains on offer today at various sponsoring merchants will keep them on the right side of reality.
And, in that regard, I suspect the uncharacteristic tendency of their "loved one" to remind them of the bargains on offer today at various sponsoring merchants will keep them on the right side of reality.
I can very easily envisage a deceased wife continuing to order shoes.
Unfortunately, the people who want to speak to simulations of the deceased are also those who most want to believe that it worked, and their long-lost one is back again in some form from beyond the grave. Just look at the number of people who believe in pseudo-science like Astrology or Tarot Card Readings... can you imagine giving them a unique AI "based on the stars?" They'd suck it up like drooling toddlers.
> Would we require the technological equivalent of a wooden stake and interment at crossroads to keep the dead from haunting the living?
Good God, what a horrible thought I just had.
Could you imagine if Donald Trump were to make an AI model of himself and declare that it would be president for ever after his death? Maybe it has already happened
Of course, it would be plugged in to the surveillance networks and have command of the killer robots that patrol the streets.
Obey, citizen. The Donald is Watching You
This post has been deleted by its author
This post has been deleted by its author
ARTHUR: What? Who said that?
VENTILATION SYSTEM: The Ventilation System. You had a go at me yesterday.
ARTHUR: Yes, because you keep filling the air with cheap perfume.
VENTILATION SYSTEM: You like scented air: it’s fresh and invigorating.
ARTHUR: No I do not.
Instead of these tedious old psychometric tests, why not offer a set of dials which allow you to dial up the personality attributes that you find more desirable to create a more appealing avatar without having to stop being a grumpy old git in real life? This will also allow you the option to create avatars for your annoying neighbours and dreadful boss where you can dial up the slavering psychopath and creepy pervert attributes and set them loose on the internet for a good laugh.
The boring old me would rather interact with fictional people (in the pages of a book) than real ones. Traumatic events in my life are when a book I really enjoy finishes and I'll never know what happened to those characters. Real people? Meh, overrated.
Thankfully I have no connections, nobody who would want a digital me. Because as it's for their benefit they'd probably tweak the settings to make me an outgoing extrovert. How utterly ghastly. My inner goth would be appalled.
Let's see these boffins replicate assistant referees at football matches and be at those matches to standby the decisions of the AI assistant referees to defend the result to the football officials and more importantly the football supporters at the match.
At the moment, there is a fair amount of disagreement out there about the decisions by the Video Assistant Referees...
They've managed to create chatbots which quasi-recreate the responses of 1000 privacy-clueless morons willing to answer the dossier-like, open-ended question, "Tell me the story of your life – from your childhood, to education, to family and relationships, and to any major life events you may have had."
It's not what I'd call a win.
Hmm, I recall seeing mention of some AI thingy that participates in a zoom call on your behalf - you provide the voice from a congenial location, and it draws a life-like image of you sat at your desk, lip synced to what your saying. Hook that up to one of these things, give it a realistic sounding voice and you're good to go.
I don't find this particularly creepy, but I can't see it being particularly useful, either. At least, not in its current form - approximately nobody is going to want a virtual me, for example. Maybe, instead of replicating an individual, it could be used to give home assistants a personality - less Star Trek Computer, more Orac. And I could see applications in the virtual girl / boy friend markets, sex robots etc.
All of these things are coming (especially the sex robots), so I guess we're going to have to get used to them.
We will have 1000 AI BOFH in the systems team and another 1000 PFY on the helldesk.
I know I am a grumpy old git who has been in this business too long but AI (or the bits of it that get all the news) is determined to eat itself. I know that teams where I work are hoping to develop LLMs that are super good at spotting pre-cancerous cell development and I can imagine that scanning pictures of the cosmos to find planets will also be well served by these toys but not the crap that MS, Google et al seem to be interested in.
With the talking to dead people bit I think of Rimmer so another poor use case.
