Mental Disorders in Artificial Intelligence (book)

What readers say about #MDAI

About the author

#MDAI Book Contents

Mental Disorders in Artificial Intelligence is organized into the following three parts (to give you an idea of ​​what you will find if you decide to read it, here is the beginning of each chapter):

Part 1. INTELLIGENCE, THAT GREAT UNKNOWN

Let’s start by addressing a problem that many believe doesn’t exist, but is undeniably real: we lack a clear and precise understanding of what intelligence truly is. It might seem disheartening for a book on Artificial Intelligence to begin by admitting that we don’t fully grasp the concept of intelligence itself, but I promise you, this approach is both more realistic and far more engaging.

1. Smart as a Whip

Diving straight into AI is a mistake, as it assumes we already understand what intelligence truly is. This is far from reality. Many so-called “experts” offer grandiose definitions that can mislead us into thinking intelligence is a solved issue, that science has figured it out, and there’s no obstacle to creating robots like those in science fiction movies…

2. Information Devourers

From the previous analysis, we’ve established that intelligence isn’t a binary phenomenon—it’snot as straightforward as labeling an organism as either being intelligent or not. Intelligence emerges in nature as a continuum, presenting itself indifferent degrees across various species, individuals, and situations. This description of intelligence as a gradient might lead us to believe it is a one-dimensional phenomenon. However, this couldn’t be more inaccurate. Intelligence does indeed exist on a continuum, but it’s one that spans across multiple dimensions. In other words, it’s reductive to think of intelligence assomething measurable by a single number….

3. What Intelligence is Made Of

You might have heard about “cognitive services” or “cognitive computing” before. It sounds pretty cool and intriguing, right? I’ve mentioned the concept of cognition inthe earlier chapters. But what does it actually mean? What is a cognitive system? In short, it’s a system that can understand—one that generates knowledge from the data it gathers from its surroundings and uses that knowledge to solve problems….

4. Learn or Die

LHave you ever wondered how a human baby is programmed? Even if you’re not familiar with computer science, as a mom or dad, you’ve undoubtedly had to “program” your baby. The problem is human babies don’t usually come with instruction manuals or programming guides. Nonetheless, mammalian offspring from a social species like ours need to learn a lot before becoming self-sufficient and contributing to the group (in fact, I have friends my age who still live with their parents)…

5. Attention, Attention!

Right now,thousands of sensors in your skin are sending information to your brain about pressure and temperature. However, you’re so engrossed in reading this book that all that information goes unnoticed. Until perhaps, prompted by this very text, you start focusing on the sensations in your left foot. As you concentrate on your foot, you might feel an increasing weight… Or, if you don’t feel that, you might notice some itching, pressure in a particular spot, tingling, pain, warmth, cold, or some other sensation…

6. What a Memory

Have you ever wondered how much memory (or “hard drive capacity”) your brain has? The truthis, when we talk about long-term memory, no one knows for sure, but the scientific consensus is that human memory seems virtually unlimited. No one has been able to rigorously calculate an upper limit on the number of elements it can store. In any case, human memory is quite different from computer memory. Interestingly, new AI systems based on neural networks do store contents in away that’s more similar to how we do it…

7. I Talk, Therefore I Am

We’ve clearly established that humans, regardless of their level of intelligence, are inherently social beings. Language is our primary tool for communication. However, similar to other species, we utilize various forms of non-verbal communication. For instance, displaying a disgusted facial expression towards someone we dislike. Many other species are equally proficient in non-verbal communication. In fact, the non-verbal cues of many mammals transcend language barriers and species boundaries. The aggressive display of a dog or a chimpanzee baring its teeth and growling is a universal language understood by all, without the need for words….

8. Human-Level Intelligence

In the previous example, we saw that both humans and machines are capable of generating chains of thoughts. This type of planning is similar to what’s needed for performing physical tasks. Ultimately, whether reasoning or acting, planning is essential. Only in mosquito-like intelligence, which relies on quite simple stimulus-response pairs, can planning be skipped. In these cases, the sequence of actions emerges automatically based on the present stimuli: if a hand swats at me from the left, I’ll move to the right, and vice versa. There’s not much more thinking or planning involved…

9. Someone is Watching Me All the Time

The reflective processes we discussed in the previous chapter, those using the prefixes auto-and meta-, suggest that our mental system is hierarchical and self-referential. It consists of lower and higher layers. Control systems regulate their outputsbased on external data, while management and supervision systems adjust their outputs based on data from lower mental processes. The outputs of these supervision systems then serve as modulation signals for basic psychological processes….

10. Many Things Can Go Wrong

I must confess, I’m an “artificial pessimist.” I think that if something can go wrong in AI development, it inevitably will. Maybe this is just me projecting my experience as a not-so-great programmer. Regardless, the truth is that software always has bugs. This is a natural aspect of complexity: the bigger the software program, the more likely it is to have errors or to malfunction…

Part 2. Failure Cases in Artificial Intelligence

In this second part of the book, we will delve into some “clinical cases.” In the field of human mental health, professionals utilize various tools, such aspsychological assessments, diagnostic interviews, and therapeutic interventions, to evaluate, diagnose, and treat patients. One effective method for understanding a pathology is through the study of clinical cases. By examining the detailed history of an individual experiencing a particular mental illness, we gain deeper insights and avoid reducing these disorders tomere statistics. For instance, exploring the journey of someone with depression reveals the complexity and nuances that raw data alone cannot capture…

11. Study and Diagnosis of Mental Disorders

Similarly, in the business world, terms like “use case” or “success story” describe specific applications oftechnology. The field of applied AI is rife with hype and sensational headlines, such as “AI Detects Cancer More Accurately Than Human Specialists,”which tout the remarkable abilities of intelligent machines. However, just as clinicians study individual cases to understand mental illness, we can examine failure cases to improve technology. For example, consider the headline: “Autonomous Vehicle Mistakes a Child Dressed as Spiderman for a Red Traffic Light, and Stops the Car Until Parents Remove the Child.” These failure cases are crucial for refining AI systems and ensuring their reliability and safety…

12. Defiant Negativist Disorder (Patient TM)

As a general rule, to avoid embarrassing the creators and the companies involved, we’ll maintain patient anonymity by using coded names like “TM” Usually, the initials of the patient’s name are used. In this case, the letter T stands for the name of a Twitter (now X) bot that turned sexist, racist, promiscuous, and xenophobic (quitethe package), and the letter M corresponds to the company that built it, amultinational software giant known for its windows-based operating systems…

13. Delusional Disorder (Patient CG)

CG, a free chat service launched in 2022, quickly gained widespread popularity. Chances are, you’ve interacted with it or used one of the various apps built around it. Theinitials “CG” stand for “Chat” and “GPT,” indicating that the core generative model behind this chatbot’s responses is the GPT architecture. CG was developedby a company whose name suggests they create open-source AI, but in reality, their products are proprietary…

14. Attention Deficit Disorder (pacient AA)

In recent years, insurance companies have increasingly turned to AI to enhance their operations. One notable application of AI is in the realm of auto insurance through computer vision technology. Traditionally, when an accident occurs, a human claims adjuster or another qualified professional must inspect the damaged vehicle. This initial inspection is crucial for estimating the impact and potential repair costs…

15. Autism Spectrum Disorder (Patient SR)

In an ironic twist, this case involves a robot designed to assist children with Autism Spectrum Disorder (ASD). Am I implying that a robot created to help children with ASD exhibits ASD-like traits? Yes, that’s exactly the point. It’s not just a flaw in a particular model; in fact, all so-called social robots face significant challenges in communication, understanding others’ perspectives, and social interaction. So, are we missing the mark by using these robots to support individuals with autism? Let’s examine the case more closely before drawing any conclusions…

16. Coordination Disorder (patient RVC)

As we delve into the topic of AI with physical bodies, it’s crucial to differentiate between industrial and service robotics. Industrial robots, which have been integral to factories for years, operate inhighly controlled environments with precise, repetitive tasks. This setup allows them to excel in roles like welding airplane fuselages or assembling electronics. In contrast, household robots face a much more challenging environment. Everyday tasks, such as going out to buy bread, require navigating unpredictable settings, interacting with people, and adapting to unforeseen circumstances. These factors make developing effective and safe household robots significantly more complex and demanding than their industrial counterparts…

17. Major Depressive Disorder (Patient M)

Unfortunately, depressionis a disorder that many people know all too well. However, when it comes to artificial beings, the issue of inadequate emotional regulation has not been asignificant concern. As we’ve discussed, emotions such as psychological pain,sadness, or love require consciousness to be genuinely experienced. Can an AI experience emotions like humans? Traditionally seen as emotionless machines, advances in AI raise new questions about their potential for emotional experiences. In this chapter, we’ll analyze Marvin’s case to explore these intriguing possibilities. His story is not a cheerful one—examining mental disorders rarely is—but it’s a journey worth taking. Keep a tissue handy, because if you find yourself connecting with his pain, as I did, you might end up shedding a tear or two…

18. Artificial Personality Disorder (Patient PG)

[…] This project aimed to make productive use of the extra time. Perla was designed as a chatbot, a conversational agent capable of identifying symptoms of depression in humans through structured interviews. The original version of the chatbot was fairly limited, relying exclusively on the 9-question Patient Health Questionnaire, or PHQ-9…

19. Psychopathic Machines

In examining these cases, we consistently note that AI lacks consciousness and subjective experiences, such as pain or emotions. Without an inner life or real feelings, could machines potentially develop traits akin to psychopathy, characterized by a lack of empathy or remorse?…

20. Simulated Patients

Why simulate mental disorders using AI, you might ask? Traditionally, we’ve tried tounderstand AI’s thinking by drawing parallels to the human psyche. However, there’s no reason we can’t flip the script: we can gain valuable insights into the human mind by replicating its unique characteristics in artificial brains…

Part 3. Creating a More Intelligent World

Okay, so we know that both humans and machines experience occasional malfunctions; it’s anunavoidable reality. Acknowledging this is the first step toward managing these imperfections. Now, let’s delve into the most effective strategies for addressing and overcoming these challenges..

21. Collective Intelligence and Stupidity

A key takeaway is that just as humans need mental health care, AI also requires similar oversight. Everyone faces challenges and deserves support for recovery and improvement, including AI systems. By maintaining a scientific approach to monitoring AI’s psychological aspects, we can avoid unnecessary sensationalism. Although these artificial minds are still far from matching human cognition—and may never do so—they are becoming increasingly complex and powerful…

22. Cybercivilization

Artificial Intelligence, in conjunction with advancements in biotechnology and other fields, is propelling our civilization toward a hybrid society. This new era is characterized by the integration of artificial and biological minds, fundamentally changing how we perceive evolution and societal norms. The traditional distinction between natural and artificial evolution is fading; now, all minds, whether human, AI, or augmented hybrids, are converging on a shared evolutionary path. In this journey, humans enhance their capabilities through AI, while AI systems evolve through human input, creating a synergistic dynamic that reshapes our world…

23. Artificial Intelligence as a Psychotherapist

Although this book mainly addresses the abnormal functioning of artificial minds, it’s also important to think about our own psychological issues and how intelligent machines can assist us. For instance, in the second part of this book, wediscuss SR, an assistive robot designed to help children with autism, showcasing AI’s potential to enhance human mental health. Beyond robots like SR, which engage in therapeutic activities and provide companionship to the elderly or disabled, numerous chatbots are also available. These chatbots are specifically designed to support individuals dealing with psychological issues such as anxiety and depression…

---

Continue reading: