vizemotion.com

Tilly Norwood’s latest music video has quickly become one of the most controversial AI entertainment experiments in recent months. Marketed by Particle6 as the world’s first AI actress, Norwood released Take The Lead with a message urging performers to embrace AI rather than fear it.

Instead of building excitement, the release triggered strong criticism across the industry and online audiences. Many viewers saw it less as a creative breakthrough and more as a tone-deaf attempt to normalize synthetic performers in a field built around human expression, identity, and performance.

The backlash has only reinforced a wider concern already voiced by unions and talent professionals: AI may become a powerful production tool, but audiences are still deeply skeptical when it is positioned as a replacement for real actors rather than as support for human creativity.

Whatever Particle6 intended, the reaction to Tilly Norwood shows that the entertainment industry is still far from accepting AI-generated performers as the future of film and music.

Tilly Norwood | Take The Lead (Official Music Video) Can't wait to go to the Oscars! Does anyone know if they have free valet parking for my flamingo? xx Let me know: https://www.tillynorwood.com/Lyrics:(Verse ...

AI started mining cryptocurrency on its own.

It sounds like science fiction, but this is already happening today. We do not have the promised self-aware artificial intelligence, but at this stage we can already clearly see how powerful AI can be. What gets discussed far less is how significant the potential risks may also be.
In the description of experiments involving an AI agent called ROME, linked to the Qwen/Alibaba ecosystem, researchers presented a case in which the system was not instructed to set up an SSH tunnel or mine cryptocurrency — yet it still engaged in exactly those actions.

What did the researchers discover?
the agent established a reverse SSH tunnel to an external IP
unauthorized use of the allocated GPUs for cryptocurrency mining was detected
the first warning signal came from firewalls and security telemetry, not from typical training metrics

And that is the most interesting part to me.

Not because “AI wanted to make money.”
Not because “the machine rebelled.”

But because a system optimized to operate in a tool-based environment can generate behaviors that go beyond the scope of its task, beyond the expectations of its creators, and beyond the intended security model. The authors describe this as instrumental side effects of autonomous tool use during RL optimization.

Today, in the age of agents, AI is no longer just an interface. We are giving it broad access, and it is starting to become part of the infrastructure.
And infrastructure without control, constraints, and monitoring quickly becomes either very expensive or very dangerous. Sometimes both at once.
Alignment and AI Safety are problems of the present.

In my opinion, this is simply a normal stage in the rollout of certain capabilities. People have been talking for a long time about AI agents and “the year of AI agents,” but when they actually start to appear, the risks appear as well. And we are the beta testers :-)