The AI we created

Am I bored of AI and bored with AI?

It seems that everything that mere humans could write about AI is already written. From now on, everything that we can write will be written with AI. As AI increasingly becomes our everyday language, nothing that has not been created already can be said.

In 1999, AI was hard but optimistic. Every day was a failure, and the rare, few, bold steps forward were overly celebrated and publicised. A few knew that a revolution in global affairs based on the internet and completed with AI was taking place, although the components still needed to be in the correct scale or order.

At the time, the positive intent was palpable.

In 2010, we could see a society with the right components at the correct scale with global connectivity, internet scale data, and cloud computing capacity but a need for more appetite globally to finance those components. The world was still reeling from the recent market crash that restricted the appetite for risk or investment, and society was ill-prepared to adopt new technologies whilst still staggering.

The spectre of the dot-com bubble crash in 2000 reinforced a view that technology investment would inevitably fail.

By 2015, global ownership decisions around AI adoption had been resolved. Google had recently purchased Deepmind for merely $650m, Microsoft began to move into AI under Satya Nadella, and ownership of AI gravitated into the hands of a few: the Chinese government, South Korean communication, and US global technology firms.

These organisations had pulled away to the extent that any new AI entrant would require one or more to succeed.

A decade later, consider ownership of OpenAI, a similar groundbreaking startup that offered to revolutionise AI and, in turn, society. Funded initially by "tech bros" and investors, it now seems inevitable that a US global technology would acquire it in some format. Not just because those organisations had surplus cash. Microsoft's 2023 $10bn investment into OpenAI is a fraction of what they spent acquiring game studios, having simultaneously paid $70bn for Activision as a single purchase.

It is the infrastructure elements that make it inevitable. OpenAI needed cloud computing, global connectivity, and internet-scale data to grow and expand. Only US technology firms could offer that investment; a large proportion of Microsoft's $10bn investment uses these tangible assets rather than financial ones.

Any new AI organisation seems unlikely to rival the existing ownership collective, reinforced as governments and international bodies begin to regulate and control AI. Their increased requirements for governance and openness, with, in turn, increased regulatory costs, make any challenge even more unlikely.

2023 has seen the completion of regulation, at least in terms of intent. China established government control and direction in the early 2010s. The US adopted its market-led process simultaneously, creating freedom of investment for expansion. The recent EU AI regulations complete the major regional approaches.

The EU approach will make an EU emergent AI organisation that could rival China, South Korea, or the US even less likely. A simple test would be whether the EU could establish an AI organisation that could scale to rival the existing ones and comply with their new regulations competitively and productively. The simple answer is that they could not, as they need the essential infrastructure, even if they offered the financial investment.

The result is bleak.

My children use AI to write better essays for school; my wife uses AI to improve her work, and my 84-year-old mother uses AI to write longer Christmas card messages. AI has entered our society and become as present as the internet.

Companies have adopted AI to increase productivity. Today, we measure progress with minor, daily improvements, a little time saved here and an extra bit of work avoided there. Yet, those companies that have adopted AI are slowly and surely pulling away from their competitors, who are still discussing how or when to adopt it. Each day, the gap widens, and those yet to adopt are inevitably falling further behind to the extent that later adoption will never help them catch up.

Governments, especially militaries, are racing to adopt AI, often with good intentions. We have primarily resolved the concerns from the 2010s that unethical AI will run rampant through legal and societal controls and boundaries. It seems that an Ethical AI Conference occurs weekly now, and even governments, after a decade of campaigning, are beginning to understand the risks of unrestricted AI.

Increased regulation has created discrepancies and irregularities between nations. The EU seeks to restrict facial recognition technology whilst the UK expands its reach by enabling police access to passport and driver licence facial images. The US wants to limit Chinese access to AI processing units, whilst South Korea increasingly employs Taiwanese-manufactured processing units. Comprehensive regulation is embedding the current norms and states.

It feels that we are now in the equivalent of a chess or Go endgame. The significant plays on the board are complete, and players are now establishing their control zones. For society, one side may secure a stalemate at best while the other attempts to develop an inevitable checkmate.

AI is everywhere, truly becoming our new language. Regardless of our activity, AI will become as present as the internet, mobile phones, and computers. It will take part in all of our interactions.

Like many others, I have become exhausted with AI. A stalemate is not the best outcome for our society, yet an endgame that further empowers AI as a common language is hardly better.

Quantum computing may renew the debate; although being so dependent on AI to develop its code, we may still see software control being more critical than hardware ingenuity. Quantum may merely replicate its computing potential and accelerate the endgame exponentially.

Perhaps populist politicians will turn against technology and AI, removing the US technology organisations from the game. Even as populism seems more likely, driven by the wealth and power inequalities between the technologist few and many in the US, it is unlikely that this would be a better outcome for society, especially Western society.

In 2024, we have a society reliant upon the language of AI, controlled by a few organisations, dependent upon the existing technology infrastructure, regulated in different ways depending upon geography, and driven by productivity gains over societal improvement.

The inevitable conclusion is to accept the outcome we have decided on and at least set my lands in order. We have the internet we created, if not the one we wanted. We now have AI that we have built, if not the one that we should prefer.