1
0
I’ve spent much of the past year getting up to date on all the published academic literature about the impact of technology on the brain. I’ve also been keeping up with the latest developments in generative Artificial Intelligence (genAI) by testing out the latest genAI tools that people are getting excited about from one month to the next, not to mention experimenting with building my own ChatGPTs. I’ve also been working closely with a new start up business called Brain Man VR which is revolutionising learning maths for 12-18 year olds studying for GCSE, Baccalaureate and SATs by immersing learners in a VR pizza restaurant and helping them to get better at quickly calculating ratios. It has an AI maths tutor embedded in the game so that learners can get help and guidance when they need it, so I’ve been commissioned to help with the prompt engineering to ensure the learning experience is a smooth as possible. I’ve also spent a little time getting to know LangChain and HuggingFace with its huge database of open source AI libraries with a view to building my own AI agents from scratch.
By the end of last year I was quite nervous about the prospects of genAI. The then UK Prime Minister Rishi Sunak had very recently assembled the great and good of the AI community at Bletchley Park to discuss the existential threat of rapidly improving genAI algorithms to the whole of mankind. Might sound a bit over the top but the top brains in AI have been thinking for many decades about the long term consequences of an AI that becomes more intelligent than humans and, sadly, whatever we try we’ll almost certainly NOT be able to keep such a technology under humanities control. A sufficiently intelligent AI will always outfox us – is the thrust of most relevant arguments.
I had already listened to Prof Stuart Russell’s (Berkley University) Reith Lecture series on “Living with AI” from a few years back. I was just starting to investigate what Prof Max Tegmark of MIT and FutureOfLife.org had been saying about the pro’s and con’s of a world saturated with human-equivalent levels of intelligence or beyond. But it wasn’t until reading The Coming Wave by Mustafa Suleyman, former co-founder of DeepMind and current head of AI at Microsoft, that my anxiety levels really started to ramp up.
However once I started getting a clearer understand of the current state of genAI, the direction of travel towards Artificial General Intelligence (AGI) and the long term outcomes of various possible approaches to realising its full potential I started feeling much less worried. So I’m hoping that this blog will point people to resources they can use to educate themselves a little bit on AI in order that they too might find themselves slightly less daunted but this spectre looming over us and transforming working practices seemingly overnight.
The message from all the expert sources regarding AGI (as opposed to the AI that corrects your spelling in your text messages or the genAI that enables you to ask ChatGPT or Gemini or Claude or Grok to write you a report, plan a project or draft an email for you) is clear and unanimous in one regard: it’s a matter of when, not if, generative artificial intelligence exceeds human levels of capability across all cognitive capacities. It could be in a week, it could be in a decade, but most experts seem to now measure the time before we are likely to cross this “singularity” in months, rather than years. And it will change human life on Earth forever, irreversibly, irresistibly, whether you choose to engage with it or not.
Nick Bostrom’s 2014 book Superintelligence first got me thinking about the various ways in which super-intelligent AI will always out-manipulate any human who might think they’ve got it suitably chained and shackled. It will inevitably be able to trick us into letting it escape because its intelligence level won’t slowly reach parity with our own capabilities and then stop, giving us a chance to change our tactics, but instead it will surpass our abilities at breath-taking speed and give us as much chance of regaining mastery over our environment as a goldfish in a bowl.
While we probably won’t see the exact date of AI reaching equivalence with human intelligence, we can still take practical steps towards staying vigilant in looking out for signs of manipulation in how technology of all forms guides our actions. It is no secret that Silicon Valley has been steering our attention and decisions using powerful algorithms for decades. A compelling illusion is, of course, created for the vast majority of consumers suggesting to us that we remain fully in control when on social media, for example. The data evidencing the rocketing screen use across the world, however, makes it very clear that we are actually being sucked into endless loop of likes and reposting, whether we like it or not. This diverts energies away from activities that might actually improve wellbeing, which is my major concern for the brain health of those who unquestioningly use technology more and more without doing anything about it.
A few years back this might have sounded like the rantings of an unhinged conspiracy theory nut. But by now, so many whistle-blowers, documentaries and leaked articles / presentations have revealed that the technology behind Facebook, Instagram, Tik Tok and Snapchat has been specifically-engineered with the singular goal of maintaining people’s attention for longer and longer, that these are now broadly accepted and mostly undisputed facts.
The key problem with the technologies cited above is that the business models of those kinds of tech companies revolves entirely around maximising the stickiness of the app. Keeping people using it daily, over long periods of time to entrain habits that change brains to make them addicted (see last month’s blog in the Book Reviews section for three books that explain more) is fundamental to their ability to generate revenue.
It all revolves around the goals that the software engineers are set. If the goal is to maintain people’s attention for as long as possible (to generate more ad revenues), simple A/B testing at scale can generate data with strong statistical power to prove, more or less unequivocally, which of the A or B version of the image, article or video keeps people glued for longest. This process ends up increasing global consumption of screen media year on year (which eats into the time each day that could be spent doing the non-screen based activities that our brains evolved to need to be healthy and happy).
The Trouble with AGI
The problems with artificial general intelligence (AGI) will be similar. They will revolve around the struggle to set goals that won’t end up hurting us. There are ways of setting AGI goals that are safe, described in detail by people who have been thinking about this problem for decades, like Stuart Russell et al but, as usual, the boards and investors of the world’s biggest AI companies are blinded by the desire to outcompete their rivals to increase the value of their firm beyond that of several nations GDP put together, that such prudent considerations fall by the wayside.
The ubiquity of the smartphone changed culture forever, even for those who chose not to use one. This is because it impacted cultural norms, in schools, on high streets and in places of work. History will no doubt repeat itself with genAI. Because, frankly, we never seem to learn from our previous mistakes with allowing tech companies to create apps that are so more-ish the average person struggles to limit their consumption and so our local communities become more and more tolerant of the side effects of these addicted behaviours. Desisting from engagement with genAI will not constitute resistance to the changes it will impose on working practices, social norms and daily human experience. So I’ve decided I might as well inform myself, spot strategies that I can use to mitigate against the inevitable harms and share these insights as widely as possible.
The good news is that our ancestors’ brains evolved to adapt to whatever pressures our immediate environment placed on us. The bad news is that our brains will only adapt our capabilities in a way that benefits us if we think carefully about what goals we want the genAI. A pragmatic approach is to learn what gen AI can do and learn to drive the bus, because pointing genAI in the right direction will become most people’s daily work before you know what hit you. But you to trust that genAI will get it right every time and that you don’t have to bother checking its work is the first step towards catastrophe. We still need humans to use their experience, wisdom and judgement to confirm that the output of the genAI tools is accurate. Otherwise we’ll lose track of the line before truth and falsehood, buried under a sea of genAI generated misinformation.
Watch this space in the coming months for a full suite of practical tips to help you Survive AI.
In addition to these monthly brain blogs, I’m also active on the social media platform formally known as Twitter (www.x.com/drjacklewis), Instagram (www.instagram.com/drjacklewis_neuro) and LinkedIn (www.linkedin.com/in/drjacklewis).
And my previous books are SortYourBrainOut and TheScienceOfSin, which each have their own dedicated website.
