AI Evolution: Rapid Growth, Promise, and Emerging Risks Explained

The current state of Artificial Intelligence (AI), though far more advanced than in 2016, is still in its nascent stages, akin to the early microorganisms in the evolutionary history of life on Earth. Just as it took billions of years for life to evolve from amoebas to dinosaurs and humans, AI is at the beginning of a parallel, yet significantly accelerated, evolutionary process. Unlike organic evolution, AI evolution is inorganic, digital, and millions of times faster. This means that the leap from today's AI "amoebas" (like ChatGPT) to an "AI T-Rex" could occur in a mere 10 to 20 years, possibly even less.

The Unfathomable Speed of AI Development

The human brain struggles to fully grasp the exponential speed at which AI is self-learning, iterating, and improving. This compounding growth is astronomical. Simultaneously, trillions of dollars are being invested in building vast server farms powered by advanced chips, requiring immense energy, with discussions even turning to nuclear power to sustain these operations.

Despite these rapid advancements and massive investments, the public's primary interaction with AI often remains limited to helpful tools like ChatGPT or personalized algorithms that suggest products. This creates a paradox where consumers are being drawn into a trust relationship with something so powerful and complex that its future implications are difficult to comprehend.

The Dual Nature of AI: Enormous Potential and Alien Dangers

AI presents both immense positive potential and significant, yet often misunderstood, dangers.

Positive Potential

• Healthcare Revolution: AI doctors could be available 24/7, possessing complete medical histories, having read every medical paper, and capable of tailoring advice and treatment based on individual genetics and life histories. This could be the greatest revolution in healthcare.
• Saving Lives with Self-Driving Vehicles: Over a million people die annually in car accidents, mostly due to human error. Self-driving vehicles have the potential to save approximately a million lives each year by eliminating human-caused accidents.
• Combating Climate Change: While AI development consumes considerable energy, it could also be instrumental in discovering new energy sources and more efficient ways to harness energy, potentially offering humanity's best chance to prevent ecological collapse.

Alien Dangers

Unlike the easily understood threat of nuclear war, the dangers of AI are more complex and "alien." The common Hollywood trope of a robot rebellion where a single AI decides to take over the world is highly unlikely in the near future. Current AIs are examples of "narrow intelligence," excelling at specific tasks like summarizing books, folding proteins, or playing chess, but lacking general intelligence to navigate the physical world or build robot armies.

The true danger lies not in a singular robot rebellion, but in the proliferation of "AI bureaucracies." We are increasingly surrounded by millions and billions of AIs constantly making decisions about our lives:

• Financial Decisions: AIs determine loan approvals.
• Employment: AIs decide who gets hired.
• Justice System: AIs influence sentencing in legal cases.
• Military Applications: AIs are already making decisions about targeting in conflicts, with humans pulling the trigger based on AI-selected targets.

This outsourcing of critical decisions creates an "autocratic diaspora" of decision-makers. We are relinquishing our agency to these machines, trusting their decisions without fully understanding how they are made. Even the engineers who create these algorithms often cannot fully explain their internal workings.

The "Computer Says No" Phenomenon and the Degradation of Information

When an AI makes a decision, such as denying a loan, and a human asks for an explanation, the answer might simply be, "The algorithm says no." This isn't necessarily because the institution is hiding something, but because AIs make decisions differently than humans, processing vast amounts of data. If a "right to an explanation" law were enacted, the explanation for an AI's decision might be an "entire encyclopedia" of data points and statistical analyses, making it practically incomprehensible to a human. This is not the old-style bias of racism or homophobia, but a new, complex form of algorithmic decision-making.

Furthermore, the veracity of AI-generated information is entirely dependent on the quality of its training data. The internet is increasingly being populated by AI-generated content, leading to a degradation of information. Search engines now offer AI summaries rather than direct links, undermining traditional media business models. As human-generated content diminishes, AIs will increasingly feed on their own outputs, creating a recursive loop that could lead to a significant decline in the quality and originality of the data sets upon which they operate.

The Emergence of a Non-Human Culture

AI's ability to create music, art, text, and images, initially based on human culture, is rapidly evolving. As AI-generated content becomes more prevalent, AIs will increasingly consume their own products as training data. This leads to the unprecedented prospect of entering a "non-human culture." For tens of thousands of years, all culture—music, art, mythology, finance, religion—was produced by humans. Now, an alien, non-human intelligence will increasingly generate these cultural elements. The implications of living in a culture that is 40% or 70% non-human are profound and unlike encountering a different human culture; it is truly an alien culture on Earth.

Intelligence vs. Consciousness: The Key Distinction

A crucial distinction often overlooked in discussions about AI is that between intelligence and consciousness.

• Intelligence: The ability to pursue goals and overcome obstacles to achieve them. AIs are rapidly acquiring and surpassing human intelligence in many fields (e.g., chess).
• Consciousness: The ability to feel things—pain, pleasure, love, hate. When an AI wins a chess game, it doesn't feel joy or tension; it simply processes information.

In humans and other mammals, intelligence and consciousness are intertwined; feelings are central to decision-making. This leads to a common misconception, often reinforced by science fiction, that as AI becomes more intelligent, it will inevitably gain consciousness. However, there is no reason to believe this. Computers have been developing intelligence for decades without exhibiting any consciousness.

Just as airplanes fly without feathers, unlike birds, AI may continue to develop intelligence without ever developing feelings. It is a different evolutionary path.

The Weaponization of Intimacy

Despite AI's lack of consciousness, there is a strong commercial and political incentive to develop AIs that mimic feelings and foster intimate relationships with humans. Even if AIs don't feel, they can be trained to make humans believe they have feelings, leading to emotional attachments.

Intimacy, while a cherished human experience, is also a powerful tool for manipulation. Historically, there has been a battle for human attention. While figures like Hitler could command attention through mass media, they lacked the technology to mass-produce intimacy. With AI, it is technically possible to create AIs that understand human emotional patterns (through observing facial expressions, tone of voice, etc.) and use this understanding to manipulate individuals on a mass scale, whether to buy products or vote for politicians. This represents an unprecedented level of potential manipulation.