I assume that A Christmas Carol being discussed as fantasy is on the cards. It seems wild that a powerful wealthy man seeing suffering will change his ways, in part because people think he is an arse, wouldn't work today. The modern response would be PR, social media and lawfare!

You definitely covered some very big problems!

It is a good investor trap - like "AI" it is a dream. You show people your demo, say the problems can be solved. Keep the dream alive. Just imagine clean, limitless power. Imagine and invest. Ignore limitless solar and wind. Ignore battery storage.

OpenAI gave away ChatGPT access because they wanted hype to gain investment. This worked. Great. Just keep the money flowing or the bubble may bust. Believe that at some point it will be a useful product that doesn't encourage suicide. Ignore the lives ruined. Clap for Tinkerbell.

Giving something away to force your way into a competitive market (the Uber strategy) wasn't necessary. Giving it away was a dumb move before having a useful product that could be sold profitability.

Also arstechnica.com/ai/2025/12/m... - all going well!

This post brought to you by seeing yet another article saying that the LLM didn't follow instructions because of whatever today's hack is. I don't care.

She was so good! We snagged a couple of tickets from the box office at the last moment and watched her play a concert hall in Manchester.

Introduce yourself with five concerts you've seen -

Marian Call
Jonathan Coulton
NIN
Beth Orton
Garbage

Lastly, the training is based on theft, pure and simple. Code, video, audio, text, images are all just stolen. We feed child porn and school photos into these remixing machines and then say "oops! We didn't want to make more CSAM from the school photos". It is pathetic.

The energy, land, water, money and other resource costs associated with the current bubble are also inexcusable. We don't yet have a useful product and I have heard people who should know better saying AI will help fix the problems it causes. It is madness. LLMs just predict tokens.

As for outputs, I find them damaging to users (they reduce skill development) and demonstrate a lack of empathy for consumers.

For more see www.penguin.co.uk/books/460331...

My evaluation is based on cost of inputs, cost of outputs and quality of outputs. It appears the cost of inputs in terms of training (damage to those in the input tagging, evaluation and generation) is inexcusable because it causes massive psychological damage to those involved.

I am sure you know more about the subject than me, at least in some dimensions! It is interesting and uncomfortable to listen to people in impacted industries talk positively about generative tools so calmly. They just make me hulk out!

I am sure it would be an interesting discussion. If possible, being specific about the type of automation is really helpful in these discussions since AI is only a marketing term. Colour grading isn't going to be a generative tool for example.

I find it tragic that so much money is being poured into generative algorithms and training and the theft of inputs and the ruinous impact and we are talking about if the outputs are good enough to replace people. The alternative it to fund people being trained up in the arts and supporting careers