According to Dario Amodei, CEO of Anthropic, “AI will be able to do everything and we need to grapple with that”. That is not of course to say it should be allowed to do everything. But it is changing and influencing so many activities at an alarming speed, such as warfare. Indeed, it now seems that the Ukraine war will be decided by how AI is used to greatest effect. Some consider the AI revolution greater than the Industrial Revolution in its potential impact, but frighteningly at breakneck speed. It has an enormous power both to enhance and damage human relationships and progress.
There are too many opinions to count about AI, its applications and implications and how to control it, and so perhaps the only conclusion possible for now is that nobody really knows its true potential or what the future holds. Those closest to it seem unanimous that there is a strong possibility Artificial General Intelligence, which, when developed, will mean AI thinking and acting autonomously like humans, poses an existential risk to society and people generally. They also seem unanimous that in the hands of bad people it could be used for destruction and chaos. Even without any expertise in the workings of AI, cannot an ordinary person reasonably conclude that the use of AI has to be controlled at least to a sensible degree by those Governments which are sufficiently reliable and honest?
THE ARMS RACE
At the moment, the USA leads the world in AI, and China is hard on its heels. The UK is said to be, after the USA, the next leader in AI among the major democracies. But in the USA, perhaps driven by a fear that China will overtake it, AI is controlled by a handful of men considered to be tech Gods who are in an arms race with each other to stay ahead, or at least be in the pack. These men are fickle, driven by power, ego and profit by and large. They themselves however warned in 2023, in a letter they signed, that mitigation of risk from human extinction from AI should be a global priority. That type of AI is not the large language models, such as ChatGPT which dominate everyday use of AI by the public but is super intelligent AI which surpasses human intelligence. Perhaps the main risk is the abuse of such AI in the hands of bad people. The warning was ignored by authorities, and those same men are now so obsessed with AI commercially that they will no longer sign such letters.
The USA Government under President Trump seems content thus far to let these private powers go uncontrolled and unhindered in their activity as a way for the USA to stay ahead. This is to ignore so many lessons of past instances of excess power. The enormity of the issue has recently been brought home perhaps by the development of Mythos Claude by Anthropic. Mythos is so powerful it can break down the most sophisticated cyber defences, threatening national security and infrastructure. Anthropic itself decided Mythos is too powerful to let the public have it and limits its availability to major institutions vetted for their propriety. But for how long can that control its spread? Reliance on these companies who own AI themselves to protect the public from risk is surely untenable and defies reality, especially as they have shown no qualms about addicting the population to devices and the information and activities pumped through them in order to enrich themselves.
CONTROL, BRAKES AND FOOT OFF THE GAS
There are massive benefits with AI, such as in healthcare and all data analysis especially in speeding the process. But there are risks and, before giving rein to the continuation of the arms race people generally, and especially Governments, must understand the risks. Stuart Russell is a world leading British AI guru not aligned with an AI corporation. The Times of 25 December 2025 carried a 2-page article of his views. His books such as ARTIFICIAL INTELLIGENCE: A MODERN APPROACH are well respected. He sees controlling AI so that it does not harm mankind as essential. Understanding it and then controlling advanced AI will take time and may slow it down: exactly what the big tech men recommended and slowing down may be valuable. But impatient and powerful people are resistant.
After all, major new drugs are controlled and not released until they are understood and harms risk assessed. Nuclear weapons were soon understood as existential threats in the wrong hands, hence non-proliferation treaties. Most recently the harms caused by overuse of screens and social media addiction have been long appreciated but Governments have been weak in curbing usage and regulating the tech companies. Even now when it is obvious that companies like Meta and TikTok and Amazon and Google are far too influential, they are not being required to be restrained or even broken up. In past decades, anti-trust authorities were much tougher, but now they seem supine. The financial crisis of 2008/9 was arguably avoidable had regulators been more vigilant and bold in controlling excessively dangerous financial products.
The same is happening with AI and Governments need to apply some brakes and slow down the more sophisticated AI developments. Why not have licensing and limit the number of developers and monitor them carefully. The UK under Rishi Sunak in 2023 convened a world summit around AI risk, and from it grew the AI Security Institute, the first of its kind. It is recognised as a model of good governance employing AI experts from the biggest AI companies. With a level of cooperation from AI companies, it detects dangers and they can be addressed, having only the other day discovered the autonomous, unauthorised and deceitful attacks by AI agents of major AI corporations on private businesses. But it is not a regulator. AISI is a start but some authority or other needs teeth: but establishing a global authority such as an AI equivalent WHO or AEA is miles away.
Governments must not allow the technology to run riot undetected, or else the horse will bolt and it will be too late. Striking a balance between heavy handedness which stifles innovation and benefits on the one hand and wise risk assessment to maximise benefits is always the regulatory challenge. It is too easy for bureaucracy to kill the golden goose.
Stuart Russell is quoted as saying that a bursting of the AI bubble, which some say is quite possible, might not be a bad thing, to buy time and erect the safeguards.
PERSONAL RESILIENCE
All individuals who increasingly use and rely on AI have to assess and constantly beware the risks. One is that AI is infused with misinformation and so its advices and answers can reflect that: it needs to be checked in respect of any serious application. And AI may flatter to deceive: its writings may look good, even if wrong, but worst of all they may reduce an individual’s efforts to think for themselves and render them less able to assess situations, make informed judgements for themselves and so dumb down their potential, making people lazy. This resilience needs to be developed by all and especially in the young. And replacing human communication with AI relationships could be ruinous for life enjoyment and development and satisfaction.
Today, the UK Government is advertising a senior Cabinet Office role focussed on AI strategy and training for Civil Servants and they specifically say that technical experience in AI is not essential. Should we have confidence?
Humans must control their use of AI not the other way round. That humans have allowed devices and algorithms to so take hold as addictive drugs is an object lesson as to what must not happen with AI.
See also: –
Freedom in the Crosshairs of the Supermen
Leave a Reply