Time for a catch-up!

Things continue to develop at great speed in the AI World, with new models being released what feels like daily. The market turmoil following Trump’s on/off tariff saga continue to play into the overall narrative, underlining not only the vast sums of money involved, but also the competition between the USA and China.

So, after a break of a few weeks, like dive in and catch up on the key bits …

Big Sharks vs. small fish

There was another flurry of software releases with OpenAI announcing yet more models – I won’t even bother to list them as it will probably have changed again by the time you’re read this!! However, OpenAI are not the only ones releasing models, with xAI finally releasing Grok-3, Google’s Gemini 2.5 Flash and Anthropic’s Claude 3.7 Sonnet.

 However, the bigger news in my view was the continued opening of various features and models to “free users”, as token prices reduce dramatically in the face of competition from Chinese competitors. Recent changes in computing capabilities also means that the playing field is levelling in terms of what the models can do, meaning that there is likely to be a greater focus on price moving forward. As an example Gemini 2.5 Turbo’s tokens cost about 1/6^th of OpenAI’s, which is great for consumers, but is it sustainable?

For now, the income of the major players is clearly not being impacted, with both Amazon and Meta posting healthy income, much of it derived from or linked to AI usage. However, storm clouds are potentially gathering, with questions being raised around the accuracy of how the models are scored, as a new study from top universities, including MIT and Stanford, says LMArena, a popular AI model ranking site, might be giving unfair advantages to big tech companies like Google, Meta, and OpenAI. There was also controversy with the recent Meta Llama releases, when it turned out the testing was undergoing a degree of “manipulation”, with allegedly only the better working parts of some models being tested.

All of this has not slowed down the way of investment, from computing infrastructure to power supply, with a recent claim suggesting AI in the future could require 97% of the World’s electricity to run itself. I can honestly envision a point at which the big players eventually start building themselves their own nuclear power stations, as the power requirements increase, and the sustainable power sources become fully utilised.

Legislation, policy and other news

On the legislative front, things were relatively quiet, although there has been a flurry of local legislation around deepfakes. Policy and in particular ethics came into sharp focus, with increasing news about both the USA and China developing military applications of AI, including in terms of humanoid robots – did these people never watch the Terminator films!!??

Microsoft also announced that they had helped prevent $4 billion USD of AI-assisted fraud in the last year, whilst it was revealed that there has been a dramatic drop in the use of malware, whilst AI fake audio and video are increasingly enabling new adaptions of traditional frauds.

Despite the potentially scary nature of AI, these last few weeks have again highlighted the immense potential benefits it could bring to mankind, with indications of rapid development of medical treatments, from diagnostic capability through to synthetising new drugs.

What is clear however, is the impact that AI will have on businesses and jobs. On the one hand we had companies like Duolingo whose CEO stated that he intends to replace as many jobs as possible with AI, whilst Wikipedia took a more human-centric position, stating that it had no intention of sacking any of its researchers, journalists or editors, instead utilising AI to support those staff to be more efficient and effective.

Some sectors will clearly be impacted more than others, as underlined by Microsoft’s announcement that 30% of its coding is done by AI. So, now is the time to adapt and be ahead of the curve, before it’s too late.

Stay informed. Stay critical. And wherever possible—stay ahead.

Regards

Tom Carter