MINDFUL AI: Reflections on Artificial Intelligence

PREFACE

I sit in front of my notebook and try to write a preface. After all the thoughts and quotes I’ve written over the past few years, I’m struggling for words. Honestly, I have yet to think of writing a book about the outpourings of my neurons on the topic of AI. However, the reactions and feedback on social media and the requests of some of my friends were so positive that they finally convinced me to publish it.

Even as a young boy, I had a penchant for snippets of thoughts and quotes. I remember devouring Marcus Aurelius’ Meditations, and they still inspire me more than 20 years later. They make me think and feel like I can penetrate related topics more deeply. To me, quotes are like thought drops that sometimes penetrate deep into our souls and give us insights that a hundred books can’t.

If this book inspires you or makes you look at the subject of artificial intelligence from different angles, then it has fully served its purpose and, at the same time, made me one of the happiest people in this world.

I hope you enjoy reading it and that it gives you a lot of inspiration and insights.

***

Be Mindful with Artificial Intelligence, otherwise, it will overtake You one day.

Murat Durmus –

Frankfurt am Main, October 2022

THOUGHT & QUOTES

ARTIFICIAL INTELLIGENCE

I have done a terrible thing.

I have demystified Artificial Intelligence.

***

Recognizing that two points of data are connected is not enough.

The System must ask

why one point affects another.

***

Artificial Intelligence is not a

new wave of technology.

It is much more like a Tsunami that threatens to flood us if we are not mindful.

***

AI is a branch of philosophy and not of computer science.

AI is not a revolutionary but a transforming technology.

We have only seen & experienced the tip of the iceberg in AI.

AI is an accelerator of evolution.

AI makes us seriously question what it means to be a human.

***

Progress

What worries me a bit is that we don’t know the current state of AI progress. New heights and approaches are reported and published almost daily from all parts of the world. Hardly any area seems to be untouched. Sure, we all want progress, but really at any cost? I think a little more humility in the development of AI would do the whole thing well; otherwise, the whole thing threatens to overwhelm us one day, and we will no longer be able to correct it, let alone control it.

***

Robustness

The foremost question we should always ask ourselves when we get serious about AI:

How can we make robust artificial intelligence systems in the face of lack of knowledge about the world?

***

***

Artificial Intelligence is highly Interdisciplinary. Therefore, let’s approach it in a Multidisciplinary & Holistic way

***

Multidisciplinary

The lack of domain-specific knowledge when designing AI Models:

One of the main reasons many rely on black-box models is not the model’s accuracy and performance but merely the lack of domain-specific knowledge. The problem could be solved by training more experts in a specific domain, such as medicine, biology, psychology, pharmacy, marketing, etc., to Data-Scientists or ML-Engineers. At least to the extent that they can evaluate the whole and adapt it if necessary. There are already some, but it needs to be challenged and promoted much more. We can only benefit most from AI if we approach it in a multidisciplinary way.

***

Robustness of Artificial Intelligence Systems:

Steadily increasing technological advances in artificial intelligence encourage more and more businesses and governments to deploy AI in high-stakes environments, including autonomous driving, diagnostics in medicine, managing the power grid, and controlling autonomous weapons systems. However, for such applications, AI methods must be robust to the known unknowns (the uncertain aspects of the world that the computer can reason about explicitly) and the unknown unknowns (the elements of the world that system models do not capture). Thus, we need to pay more attention to the challenges of dealing with both known and, above all, the unknown unknowns. These issues are essential because they address the fundamental question of how finite systems can survive in a complex and dangerous world and thrive for humankind and nature’s benefit.

***

From pre-existing bias to biased datasets to the emergence of unpredictable correlations; The reasons are manifold. It will still take a while for AI to reach a certain level of maturity.

***

Brute-Force-AI

Did you know that GPT-3¹ has trained on 8 million text documents (data that reflect biases that occur in the real world) scraped from the web, costing $12 million in electricity to train it?

I call such an approach: Irresponsible Brute-Force-AI.

AI, for AI’s sake, is a nearly-guaranteed path to disaster

!***

Uninterpretable (Black-Box) algorithms should only be used for knowledge discovery processes

and not for decision-making.

***

Race of AI

The question of who will win the race and dominate AI in the future is primarily not about who has the leading companies or produces the cutting-edge innovative technology in this sector. It is rather about who has companies that can quickly take over the breakthrough algorithms/technologies and adapt them to processes that create economic (and hopefully sustainable) value.

***

The adoption of Artificial Intelligence must be done holistically and not in fragmented components that are not fully integrated into the Organizations.

***

One should focus on Data, not on Algorithms

***

AI Depression

I am suffering a bit from AI depression.

Let me give you some reasons:

▶ The AI-Control-Problem (still not sufficiently solved)

▶ Still significant concerns about privacy, ethical issues, and security

▶ No uniform/sluggish regulations and laws.

▶ The reality and expectations of AI diverge significantly; AI is overhyped.

▶ Ethical/psychological implications are not yet clear; much is still a theory and not proven in practice.

▶ Philosophical, sociological, and psychological human-science studies/approaches are still not appreciated enough.

▶ The hype is followed by sobriety (AI is still an expert system)

▶ Many AI systems are still inefficient, need a lot of energy, or must be implemented complicatedly. We need different, more efficient approaches.

▶ Data awareness has not yet developed sufficiently across the world.

▶ Still too little usable (real-time) data is generated. The significant expansion of Edge/IoT/5G devices and infrastructures will continue.

▶ AI is dominated by a few Big-Techs (AI is becoming a power instrument increasingly); This could have adverse effects on AI.

***

The field of AI is highly

interdisciplinary & evolutionary.

The more AI penetrates our life and environment, the more comprehensive the points we have to consider and adapt. Technological developments are far ahead ethical & philosophical interpretations.

This fact is disturbing,

The AI-Control-Problem

In my opinion, the AI Control Problem is still underestimated and receives too little attention. Only when we have found a satisfactory solution for the control problem should we allow AI to interfere fully in our lives and