Conscious AI will take over the world
An opinionated piece
Artificial Intelligence (AI) has seen a massive surge in about every sector of business. Companies are trying to develop new technologies that people use in their daily life. Instagram, TikTok, and Twitter have all become multibillion-dollar companies by using algorithms to let people browse on their platforms. People use these platforms to be in contact with friends and family, look for entertainment or like-minded people. This is known to a lot of people. What a lot of people don’t know is what these platforms do with the data that the users generate on said platform. There are millions of data-points for every user on these platforms, which the algorithm uses to determine what to show the user. These data-points have been engineered to (almost) perfection, tailoring the user’s experience to each person specifically.
AI in daily life
A prime example of how these data-points can be used in everyday life is with the application of Internet of Things (IoT). This web of intercommunicating devices uses information about people’s routines to tailor their (consumer) products to them. This AI is coming (quite literally) into the home with Amazon’s Alexa, Tesla’s self-driving cars and smart-home applications such as smart-lights and smart-doorbells. An example of this can be turning the thermostat on when somebody is commuting home so that the house is nice and warm when they arrive. These products learn from the routines people have and tailor their experience exactly to their routines. One could say that these implementations have some positive effects on daily life doing the heavy lifting in daily tasks, which gives us more time to focus on more important activities. Nonetheless, we must stay wary of the algorithms taking autonomous decisions, because when it goes wrong it goes severely wrong. These faults can go unnoticed for a long time and have made decisions that had terrible effects on us.
One example of this is the Dutch ‘child benefits affair’. This scandal has impacted Dutch society in many ways. The main problem of the system was that an algorithm was set to decide, without human participation, whether humans were committing fraud in requesting child benefits. The Dutch tax system is held together by benefits to help lower income households, and these child benefits are one of them. Parents can request them at the website of the tax office and have to fill in forms with their information such as age, gender, nationality, income and more. These requests would then be analysed by a machine learning algorithm to allow the data to be passed on to a human or to instantly deny the request. Besides this, the algorithm labelled potential fraudsters. In itself, this is not a horrible idea, given that we want to prevent fraud in society. Rather, it took a wrong turn when the humans that looked at every case completely believed the system and gave it full autonomy. The autonomy of the system led to people losing their livelihood, their house and in some cases even guardianship over their children.
The complete societal disruption purely over the decisions made by an algorithm, with the cherry on top being that mostly minorities were marked as fraudulent, made the system also one that discriminates. To this day, society is still feeling the effects of this crime with low beliefs in the system and politicians.
A deeper dive into consciousness, autonomy and automation
Taking your phone out of your pocket to scroll on your social media of choice, was this really you or did some force in the universe decide that you needed to do this. Was this a conscious or autonomous decision, or was it an automatism that triggered this (and thus was it out of your hands)? These three facets are the main questions that we try to explore in AI, consciousness autonomy and automation, all its own field but very overlapping.
The exact definition of consciousness is a question that humans have been trying to develop an answer to for many hundreds of years, mainly with Locke tossing up “the perception of what passes in a man’s own mind” in 1690. We currently have come to some sort of consensus that consciousness is on a spectrum from low levels to higher levels. Examples of this are dolphins, monkeys, and ravens, which have higher levels of consciousness than ladybugs. This spectrum is an important factor for animal rights in determining the level(s) of suffering, but is also used for other purposes. So where would AI tie into all of this? That is a completely new question that is in rapid development. One algorithm that has made these rapid developments is ChatGPT by OpenAI. This algorithm generates text based on a question given as an input. Where many people before have posed the question whether something is conscious, the question at hand now is if ChatGPT is conscious. Taking into consideration our spectrum, ChatGPT does have some level of consciousness, albeit being low on the spectrum. This is not a crazy idea, especially since it is able to have some level of self reflection. Then comes the question, is this algorithm actually autonomous? Well, this answer lies more complicated.
Autonomy, we are not that doomed
Searle has tossed the term “strong AI” to define AI that, given the right tools, can be said to understand and have different cognitive states. Strong AI can then be used to distinguish between a computer program that is simply executing instructions or if it can actually understand different cognitive states. The former is also defined as “weak AI” which is simply executing instructions. To this, you could pose the question whether or not this is actual “intelligence”. We can refine this by saying that autonomous AI is AI that does not require human intervention in order to execute a task. Strong AI would then be considered as conscious AI and weak AI can be used interchangeably with autonomous AI. Most of the past and current developments can be placed into the second category, but we could place ChatGPT in the middle category in between strong and weak AI, as it exhibits some cognitive abilities, but it is also limited in what it can do.
The best part of this, is that we can ask ChatGPT about its consciousness. When the query “Are you conscious?” was given, it responded with “No, I am not conscious.”. It then described itself as “… an AI language model that does not have the ability to experience consciousness or subjective awareness”. We also asked ChatGPT about the definition of consciousness, to which it responded with “It [red] refers to the subjective experience of awareness, thoughts, feelings and sensations”, which is not something that this language model has.
An even greater reduction in sensory experiences for AI is automation. Automation is the simple execution of tasks or instructions, without a form of self-thinking. A known example of this could be a conveyor belt in a factory. There are no autonomous algorithms working on placing the bolt onto the product, it simply is executing a predefined path. It can then be said that automation is restricted to these predefined paths, where autonomous algorithms can choose to deviate or pick their own chosen path to get out of trouble. Consciousness would be even more capable of experiencing these paths with the inclusion of the subjective experiences.
It is inevitable
So far, we have shared our opinion about historical and current applications of AI with the relation to automation, autonomy and consciousness. We found the growth from automation to the current weak AI, which does not yet contain consciousness but is capable of executing instructions while learning from input. If the future would bring the possibility of creating strong AI, we would reach a hypothetical future point called the technical singularity. This singularity is a point in which “AI will take over”. We believe that with the current growth, it will reach this singularity.
Before even reaching this point, we may in the current day and age have bigger fish to fry. Even if current day AI is weak AI, its impacts are enormous. We see the positive effects of letting AI drive autonomously, take care of our house autonomously by cleaning or controlling the lights and heating, which helps us focus on more important things than these mundane tasks. Nonetheless, this autonomy does have a lot of downsides when mismanaged or when the system is too trusted, where the consequences can be worse than the benefits. The question that we as a society might have to ask is “Are we ready to take responsibility for weak AI or should we hold off on letting them lead our lives and wait for better algorithms to come around?”.
No matter what we decide, what is for sure is that AI is already leading a life of its own, fully autonomous and out of our hands, for good or for the worst.
Written by Mark de Bruijn (2736303) and Bas Maat (2749086)