Does AI forget things?

More than any other blogpost I have written so far, this post seeks to raise a question which has been bothering me for a while. So, I thought I just shared it with you and sought your views on that.

The main plot

I walk a fair bit on the South Coast and in the South Downs located in South East England and more so since the lockdown started just over a year or so ago, sometimes in the areas I am totally and completely familiar with and sometimes to explore areas I know less about. After the lockdown restrictions were relaxed about a week ago, I was walking a short distance (about 5 miles) from where I live to where I was supposed to meet with a couple of friends for an afternoon coffee outside a local cafe in Hove. I was quite familiar with the route but just decided to use the navigation app on my mobile just to check. I typed in the name of the street which is distinctive enough (and was also suggested by the app assumingly using a semi-AI enabled ML system built in it), used the walking mode to ensure that I am taking the most efficient/fastest route and followed the instruction; easy! To my surprise, I found myself guided to a totally different street in yet another town (Southwick) halfway through to where I was supposed to be (see the map).

This was not the first time it was happening. A few weeks ago, when we, the MAVRiC crew, were going to Seaford Museum to point cloud scan some items of historical values, same thing had happened! The same SatNav application guided us to a point which was 8 mins away from where we were supposed to be! We used three SatNav applications and it happened that the one I was using in my walking journey last week was the only one misleading us! Strange… so this gave me some food for thought…

Memory and forgetfulness

While we may think that forgetfulness is a recurring issue (and not even a problem) in our daily lives, it could be quite serious, dire, and even worrying depending on where, when and how it happens and the timeframe within which it happens. I had the first-hand experience of that which I very well remember even to date. I was much younger and it was a nice, warm weekend afternoon in summer when I parked my car in a parking to join the rest of the family who I dropped off at a local market just a couple of minutes earlier, and for a very short period of time (probably just about 10-20 secs) I forgot where I was or what I was supposed to do; I will never forget the experience, it is deep down there in my long-term memory!

Memory retrieval is not uncommon and probably is what we use rather subconsciously and quite regularly to remember or recall a piece of information from our long-term memory where it has been registered. Forgetting does not necessary mean that the information is totally lost or deleted but it is rather filed or buried somewhere but cannot be retrieved. It is proven that information stored in long-term memory is significantly more stable than what is stored in short-term memory and therefore the retrieval of long-term memory is much more likely and probably easier. Failure to retrieve the memory could result in short or, if persisting, in a more sustained or even permanent forgetfulness. 

Although there are different ways in which memory (and forgetfulness) can be studied – including the biochemistry of memory – Hermann Ebbinghaus (1850-1909) was probably the first scientist who studied the psychology of memory (and forgetfulness). He was the first psychologist to describe the Learning Curve, with which many of us are quite familiar, and definitely the first to theorise the Forgetting Curve, which may be less known to many of us. There are several theories about why we forget including The Interference Theory, The Decay Theory of Forgetting, The Retrieval Failure Theory and The Cue-Dependent Theory of Forgetting, to name but a few. The study of memory loss, Alzheimer’s disease, and dementia* is also very well established and well developed today.

The question

As important as memory and forgetfulness is in and for homo sapiens, this blog is aiming to raise another question that is: can machines (and AI) really forget things? And if that is even possible, could it be planned, selective and even intentional? And can it be pre-empted?

Given the story that I started this blog with, one can suppose that yes… machines can forget; simply put, if and when their power supply is cut. Well… in a way that may look alright although strictly speaking even if the power is cut, the machines’ memory should not be totally and completely lost. It might not be accessible or retrievable for a while, but it does not necessarily mean that it is lost and more to the point lost permanently. But what if the “artificial forgetfulness” or “artificial memory loss” is not due to power cut but rather due to the lack of active/constant use of the data, thereby resulting in a lag in retrieving the necessary data which supports the action requested from the AI. Then the next concern arises as to what if it happens is a time-sensitive activity or task for instance in launching a spacecraft or guiding it to return to atmosphere on its way back from a space mission? What would the consequences of such memory loss be? It will not be like me ending halfway through where I was supposed to be and probably turning out late to an afternoon coffee meeting with friends or to a visit to Seaford Museum… it would be a part of a spacecraft landing – or rather falling – in the middle of New York City instead of safely landing in Cape Canaveral, and well… you can imagine the scale of the total disaster this might cause.   


And I would like to finish with even a more disturbing question. What if this artificial forgetfulness is not purely innocent, arbitrary, or accidental? And what if machines or AIs choose what, where, when and why [“WHY? None of you homo sapiens’ business” the answer by the new Master (Toddler) would probably be] to forget what they decide or choose to forget and what if we have very limited if any control over that or even power to pre-empt, prevent, or reverse their decision, or shut their operation down?

What are YOUR thoughts? Let us know.

* If you are interested in pioneering studies on dementia and the therapeutic impact of nature on it, you may find one of my close friends’, Dr Garuth Chalfont’s works interesting.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s