Digital Ghosts: Where Do We Draw the Line With AI-Generated Content?
Digital Ghosts: Where Do We Draw the Line With AI-Generated Content?
Zelda Williams asked for one thing: please stop sending her AI videos of her late father. If you've got any decency, just stop doing this to him and to me, she wrote on Instagram, responding to a flood of Robin Williams deepfakes created with OpenAI's Sora 2.
It's a request that shouldn't need to be made. But it raises a question none of us have answered yet: where do we draw the line with AI-generated content?
Permission changes everything
Look at Star Wars. Grand Moff Tarkin appeared in Rogue One. A younger Luke Skywalker showed up in The Mandalorian. Both were digital recreations of actors who were either deceased (Peter Cushing) or decades older than their character (Mark Hamill).
Cushing's estate gave permission. Hamill was involved. The families and individuals had control over how their likeness was used.
Sora 2 has no permission mechanism for the deceased. OpenAI won't let you generate videos of living people without consent. But the dead? Apparently fair game.
When fake becomes noise
Scroll through X on any given day and you'll see AI-generated nonsense pretending to be real quotes.
I keep seeing ones supposedly from Ryan Giggs about current Manchester United players. They're painfully obvious. Written like a 12-year-old YouTuber. Using phrases no professional footballer would ever say. They're not offensive, just nonsense.
It's like AI has become Jay from The Inbetweeners, constantly making up increasingly ridiculous stories that no one believes but everyone shares anyway.
But people share them. People believe them. And slowly, the real person's voice gets drowned out by AI-generated caricatures.
My hypocrisy on display
For what it's worth, I've completely given up trying to spell or fix my own typos. AI does it for me now. I've also spent an embarrassing amount of time using face-swap apps to make my friends look like Gollum.
So I'm not claiming any moral high ground. I'm neck-deep in this stuff just like everyone else.
Where I land (for now)
If I had to draw a line, it would be around consent and purpose. Using someone's likeness without permission, especially when they can't give it anymore? That feels wrong. Making up quotes and attributing them to real people? Also wrong.
Death used to mean something. It meant your story was complete. What you said and did in life was the final word. AI threatens to turn everyone into ventriloquist dummies long after they're gone.
AI News This Week
AI gets over half of all VC money for the first time
AI companies pulled in $192.7 billion so far this year out of a total $366.8 billion in VC funding. In the most recent quarter, AI accounted for 62.7% of US VC money.
OpenAI hits $500 billion valuation but can't get insurance
OpenAI reached a $500 billion valuation after a secondary share sale. But OpenAI and Anthropic are considering using investor funds to cover potential multibillion-dollar lawsuits because insurers won't provide comprehensive coverage for AI-related risks.
OpenAI wants to turn ChatGPT into an operating system
OpenAI's head of ChatGPT, Nick Turley, says the current ChatGPT is like the command line era of computing. The company wants to transform it into an operating system with third-party apps.
Why coding AI improves faster than everything else
There's a growing reinforcement gap in AI capabilities. Coding tools are improving rapidly while other skills like email writing progress slowly. The reason is simple: code can be automatically tested billions of times.
ChatGPT still dominates but Gemini is catching up
ChatGPT holds 73.8% of generative AI traffic, down from 87.1% a year ago. Google Gemini has doubled its share from 6.5% to 13.7%.
Google's AI agent fixes security bugs autonomously
Google DeepMind deployed CodeMender, an AI agent that finds and fixes security vulnerabilities in code.
Tool Review: Dean's AI Cook-a-Long Workshop
Dean, you can't review your own session.
Yes I can. It's my newsletter and the people need to know!
I've run 11 of these Cook-a-Long workshops now with 65 people across different teams, and the feedback suggests something's working.
What people built
41 functional Gems (custom AI assistants) were created across the 11 sessions. Not practice exercises. Real tools people are using.
One person automated a weekly reporting task they used to hate. Their exact words: I used to hate doing that at the end of the week, but now it's fine.
The NotebookLM revelation
This was the standout moment across nearly every session. Sales teams committed to creating one of these for every customer.
Participant feedback
Superb session, great insight and fantastic to leave with some genuine oven-ready takeaways.
The first AI Cook-a-Long: Learning how to use AI to help us with AI. Such a useful session!
Verdict
11 out of 10. It's mega. You'd be mad not to come.
Think of it like Oasis' Live25 tour. High demand. Delivers what it promises. You'll tell everyone about it afterwards. Except the tickets are free and you get something practical to take home.
Public sessions are every Thursday @ 10.