It's abundantly clear that the US hyperscalers are unable and unwilling to regulate themselves. Some owners are wealthier than entire EU countries and act like mini-kings, beholden to no one. Many have taken the view that it's more cost-effective to steal the IP of original authors (text, image, video, music, etc.) for use in Gen AI model training and pay to throw legal challenges at any copyright penalties. A $100 million pro-AI super PAC formed in August backed by OpenAI's president has identified its first democrat target to oust. Anti-regulation, deep-pocketed Silicon Valley interests are injecting themselves into local US contests from afar.

The UK/EU has nothing to offer with regard to Gen AI capabilities, it just borrows from the US, but it is right to form AI Legislation to curtail the worst excesses. It has learned from the lack of regulation around Social Media, which has moved from a fun, comms channel to toxic megaphone for the worst humans.

China has less access to high-end AI tech, but its researchers make up for this in academic skills and hard work, witness DeepSeek using ChatGPT to distil its own better, smaller models. I think the saving grace is that US LLMs running on US GPUs have run out of steam. They remain statistical, best-word algorithms, good for wordsmithing but weak in many other areas. There is also likely to be a big, "sub-prime AI" type stock market correction (bubble burst) that will scare investors. So far, the Gen AI hype greatly exceeds its value.

The next innovations may come in neuro-symbolic AI, which is a mix of machine learning and symbolic or rules-driven AI. The US is weak in this area as all the investment is still going to one-trick-pony LLMs, and the current US government is gutting its universities and penalising the entry of foreign students. Basic US literacy is in freefall. The next AI innovations are likely to come from the old-world university ivory towers and start-up labs in UK, France, Germany, Spain etc, plus Beijing.