Thanks for the Infinite Memory

Everything you are is laid out as stories in your mind. That grey mass sitting motionless inside your skull is a vast library of narratives - scripts of past events and experiences that drive your decisions. The average person keeps over 1,000 Terabytes of information like this in their brains and, FYI - that’s a lot of stories. I describe the information we hold as stories because we interpret events in our own unique ways, building and filing away short movies each day for later use. Two people experiencing the same event will build different stories - each one with a slightly different plot. 

But imagine if all of a sudden your memory of a loved one - a parent, partner or friend - was erased. Your every decision and action with them would be based purely on the here and now. A sentence spoken to you would be all you had to go on in generating a response. You’d have no idea about their likes, passions, motivations or experiences. Having access to a back catalogue of experiences about the people we know enables us to properly communicate with them in the here and now. 

Up to this point, AI language models have lacked this depth of memory. While all are pre-trained on a vast database of information, their understanding of us has always been limited. Some models (e.g. ChatGPT) have been able to store a selection of key details for a while - snippets saved during previous conversations to help provide more targeted answers in future conversations. But we are on the cusp of a major memory breakthrough. AI models will soon be equipped with Infinite Memory - the ability to access and use all of the conversations we have had with them in the past, along with data we have fed them, to better inform future responses. And while this might sound like a mere upgrade to AI’s capabilities, it’s actually one of the most significant developments happening in the evolution of artificial intelligence. More - it’s the beginning of a revolution in how we exist as humans. 

Our Thoughts, Imprisoned

Your thoughts are trapped inside your mind. Your personality and everything you are is held within you, and you alone. Your brain is an exceptionally powerful machine, referencing past experiences at an unfathomable rate. This is happening right now - as you read this article - within your conscious and subconscious mind. Maybe your school experiences were negative, which makes reading educational articles like this less enjoyable. Perhaps your mind has built skewed stories about AI based on the opinions of others or - more likely - sci-fi movies and games you’ve been exposed to. Whatever your particular cognitive circumstance, your brain and everything it knows is unique, and - importantly - walled within you. 

Having our thoughts locked away from the world is obviously a huge advantage in many circumstances - let’s be honest, nobody would want a world where thoughts are sprayed around unabated. There are evolutionary advantages to ‘private thinking’ as it allows strategic planning, social bonding and the smooth operation of society. But there are negative sides to having our thoughts stored on a single server (our brain). 

Unlocking the Cell Door

Although we will always want some of our thoughts to be kept private, having just one biological hard-drive to store our thinking is undeniably limiting. For one, when ideas pop into our heads or important information is discussed, details often slip away. Our brains do not have the capacity to generate thoughts and catalogue information like a computer - the process is far more selective. Secondly, there’s no backup. If we forget something over time, lose some cognitive ability (or worse), information locked away inside us can be lost forever.

AI is already being used to remove some of these limitations. Discussing ideas with LLMs (Large Language Models such as ChatGPT) at the point of inception or recording discussions enables AI to store and analyse information, creating both a backup and raw material for future analysis. Currently, memory-storage in consumer-level AI is basic, but it’s already proving to be useful in a range of contexts. I, for example, get a lot out of ChatGPT’s ability to recall key details of past conversations. When discussing a new idea for an article or brainstorming other ideas, its basic knowledge of me, my aims and intentions make its responses tailored and more useful.  

Getting to Know You 

So, AI’s ability to store memories is directly related to its ability to provide targeted responses. The better AI knows you, the more appropriate and useful its help can be. In the same way that we can provide support, guidance and advice to a loved one because we know them so well, AI is better positioned to aid us if it has a broad back catalogue of personal knowledge. But at present this ability is stifled by the limitations of technology - storing, accessing and computing thousands of conversations is a hardware-intensive process. Doing so at scale is a significant task - ChatGPT serves over 100 million users a day. But remember - we are only at the very early stages of the Intelligence Revolution. The technology that drives AI is evolving at an exponential rate.

At time of writing, ChatGPT has just launched a nascent form of Infinite Memory to paid-subscribers. In short, it can summarise every conversation you have had, and use these concise descriptions to build a better picture of who you are to improve interactions. While not Infinite Memory in totality, it's a move in that direction. As technology progresses, expansions in how AI forms, archives and uses memories will bring greater personalisation, and this will likely be the catalyst for something much bigger.  

The Future of Memory

Looking forward, we can build a broad picture of the trajectory we are on. Undoubtedly, greater memory will lead to greater personalisation, and AI models will become more powerful in aiding us as a result. But this transition will also alter the role that AI plays in our lives. LLMs will likely transition into digital PAs who - because they know us inside out - can provide highly personalised assistance. And as this arm of AI’s evolution aligns with developments in autonomy (Agentic AI - AI that can undertake tasks on our behalf) we will start to see a significant departure from the systems and ways of living we know.

The notion of ‘Digital Twins’ describes the creation of a digital counterpart for humans. And it’s no longer a concept walled within the imaginary world of science fiction. AI Agents, armed with an extraordinary knowledge of us, could be the beginning of a move towards simulations of ourselves. Current versions exist and are already capable of acting as digital interpretations of people through the simulation of thoughts, conversation and interaction, emulating the biological original. This has been demonstrated by several high profile figures, including tech evangelist Peter Diamandis who allowed a ‘Peter-bot’ to be trained on his back catalogue of writing, interviews and content in 2023. Even back then, the results were impressive, and I encourage you to watch the video of him interviewing ‘himself’ on YouTube to see how effective they can be. 

All this demonstrates the power of personal data in creating facsimiles of ourselves, and while someone like Diamandis can do this easily due to his status and catalogue of available work, Infinite Memory will soon make the creation of digital counterparts available to all. And when it does, we may start sending our twins out to do the things we currently do - attend meetings, show up at family gatherings, or handle any number of things.

Oh, and I should point out that the clunky smartphone interactions you currently have with ChatGPT are not a reflection on how AI will learn about us in the future. The coming years are likely to see a shift away from smartphones as the dominant interface for AI, towards wearable AR (Augmented Reality) glasses that can hear, see and store most of what we do. When considering this evolution in device technology, it’s easy to see how AI could build a solid picture of us in just a short space of time. 

A Co-pilot and Backup

At the start of this article, I described the memories we hold in our brains as stories. This is important because AI’s version of memory differs from ours. We interpret the world through our own unique lenses, whereas AI does not. Instead, it views information through a broad, clinical, societal lens - unburdened by the bias of an individual’s emotion. This distinction is important as it highlights how AI’s data-driven approach might be the perfect foil for our inherent biases and skewed versions of events.

Because of its approach, AI can already be an excellent co-pilot for life (I personally find it indispensable), but as the human-machine relationship deepens a cascade of light will be shone on its benefits. AI that has access to detailed knowledge of us will be able to better-lead us through the trials and tribulations of life, countering our impulses and guiding us to more sensible decisions. It’s not unreasonable to assume that we might rely on AI to make decisions for us in the future - if not always in our personal lives, then most of the time in work settings. The mantra of ‘what does AI say’ is already becoming commonplace, and memories built up through our interactions will only aid its ability in suggesting the best course of action.

In a previous article - Pollyfilla for the Mind: The Best Use of LLMs - I have delved into AI’s potential in bridging cognitive gaps. This is true without Infinite Memory, but once this is fully integrated the possibilities could be stark. Imagine having access to an AI that understands your mental weaknesses - maths, disorganisation, emotional bias etc. - and offers input to bridge these inadequacies in real-time as you live and work. 

Further, although wisdom grows, our brains deteriorate as we age. Accidents and disease can also impair us cognitively. And while these issues will likely be mitigated to some degree through AI-driven medical breakthroughs, it’s easy to see how a ‘backup’ of our thinking when young or healthy could be used to counteract such impairments. You see, although AI threatens to take things away from us, it also has the potential to give so much. And the levelling-up of our abilities or solutions to dementia might just be the tip of the iceberg. 

Life after death

Although understandably controversial, the expansion of memory storage opens up a clear pathway to at least ‘some form’ of immortality in the digital realm. Even now, when most people die, there is a significant footprint left online. Communications, photos and videos often remain as data long after we have departed from the physical world. But AI, and the memories it can harvest, will take this further by an order of magnitude.

I won’t get into consciousness or what it means to ‘be alive’, because it's a complex, in-depth area and not directly relevant to this piece. And I’m not for a moment suggesting that, in the future, we will ‘live’ in the digital world once we have died. But what I will say is this - if we create an accurate digital twin of ourselves when we are alive, there will be reasons for it to continue once we are gone. Relatives may wish to have a connection with the deceased, and even businesses may want those integral to their organisation to ‘stick around’. I’m sure Apple would have benefitted from Steve Jobs' input in the years since he passed, and there are many examples of why having people exist beyond the grave may be beneficial. While you, like me, may find this daunting, horrifying, alarming and distasteful, the fact remains that it will likely be possible. And when something is possible it invariably comes to pass in some shape or form. 

The Dark Side

Although great potential lies in a future of harvested memories and AI assistance, there are - inevitably - drawbacks and considerations that we must reflect on. As biological creatures developed over millions of years, a sudden upgrade to our cognitive abilities opens the door to a range of possible, unintended consequences. Our current approach to forming and using memories is - through the process of evolution - perfect in many ways. Forgetting things isn’t always a flaw - it can be a strength. Over time, our experiences are triaged - analysed, filtered and filed - based on their importance and relevance. This builds us into who we are, with no machine assistance required in the process. ‘Bolting-on’ an augmentation such as AI with Infinite Memory risks altering us on an intrinsic level. If you’re of a religious disposition you may view this as playing God, and despite my own agnostic beliefs, I’d be inclined to agree.

When zoomed-out, what is life but a collection of memories built through our individual experiences? Orson Welles is often attributed as having said ‘We are born alone, live alone and die alone’, but in an age of digital twins that may no longer be true. Creating a detailed tracing of our minds in the digital space will surely fracture us, removing at least some part of our individuality and magic. And while bridging our cognitive gaps has clear benefits, even this changes the paradigm of what it means to be human. That’s because we are defined by our inabilities as much as we are by our abilities. To level everyone out would be to dull-down the vibrancy of humanity - making a one-size-fits-all, vanilla ice cream of society. The messiness of human thought is a key ingredient of our uniqueness, and the enhancement of thinking may not be the cure-all many believe it will be. 

Finally, there’s the question of security and privacy - a not insignificant dilemma if this all comes to pass. Right now, you own 99% of what makes you, you. But in a future where most of your thoughts have been passed on to a centralised tech company, that may not be the case. And make no mistake - the main driver pushing Big AI to steam towards Infinite Memory is the financial advantages it could offer. Once you begin on the journey of cataloguing your thinking at the hands of a single company you may find yourself trapped in the ecosystem of a single tech giant. And then, who even owns your memories? One solution to this is decentralised AI driven by Web3 (the Internet of Blockchains), but even then we are still duplicating our thinking, and once that happens we put ourselves at risk of losing control of our identity.

And so, yes - there’s a lot to consider as we drift into the world of Infinite Memory. Many reading this will be deeply troubled by what this could mean for society and us as individuals - I know I am. But the best way we can tackle what’s coming is through intelligent reflection and an understanding of the risks. All of this is likely to occur in the coming years, and so it’s important to begin discussing the implications of a technology that will progress regardless of any personal misgivings.  

A New Mind

The technological pathway suggests we are heading to a place where the ways we think, act and operate could be markedly different. How exactly AI + memory will evolve is uncertain, but it seems plausible that the future will be one of combined thinking - the human brain augmented by AI and everything it can learn about us. This symbiosis will alter the way we live and has the potential to reform what it means to be a person. No longer restricted by the limitations of a single brain, we will be able to reach a new position of intellectual power - one that outsources our uniqueness and levels up our cognition through the magic of machines.

Despite how this sounds, this change won’t occur from a starting point of zero. Machines already play a significant role in all our lives. We’ve come to rely on technology so deeply that - when you stop and think about it - the progression toward augmenting our minds feels not just possible, but inevitable. The Industrial Revolution shifted manual tasks to the might of technology, and the future of AI will do the same for our cognition. 

Alongside a broad range of key benefits, there will be drawbacks. This is the area I am most interested in because it is of paramount importance that we protect our humanity in the face of such changes. Human instinct dictates that we innovate, build and deploy with little to no consideration for the impact of our actions. Technology is a runaway train of progress that humans are forced to ride. But we have it within us to identify our biological strengths and preserve them through the storm that’s coming.

Infinite Memory will make AI better - of that I have no doubt. In spite of my reservations concerning what this may one day lead to, I am excited for its potential. My personal experience of LLMs has taught me that AI can be incredibly beneficial to our lives - offering reasoned reflection, strategic planning and data-driven guidance. Enhanced memory will act like an adrenaline shot to AI’s potential, evolving it from a highly useful tool into an intelligence we can team up with. But we must proceed with caution, and ensure that what we gain from this transition doesn’t remove vital aspects of who we are.