Embracing Machine Intelligence: No Need for Anthropomorphism

In the rapidly evolving field of artificial intelligence (AI), there's a common assumption that machines must ultimately mimic human intelligence to be truly valuable or revolutionary. However, such an assumption may limit our understanding and potential use of AI. This article posits that we shouldn't necessarily expect or desire machine intelligence to mirror human intelligence. The reason is simple: machines are not human, and this is not a disadvantage, but rather an asset. Let's start by revisiting the fundamental principles of AI.

The purpose of AI is to solve problems, make predictions, and perform tasks with efficiency and accuracy that could potentially exceed human capacity. Notably, the aim is not to create machines that think, feel, or behave exactly like humans. To understand this better, consider a simple analogy: the goal of designing an airplane was to enable humans to fly. But engineers did not model airplanes on birds. Instead, they leveraged aerodynamics, materials science, and propulsion systems to create a machine that could fly much higher, faster, and further than any bird. By not anthropomorphizing or rather, ornithomorphizing, airplane design, we reaped far greater benefits. Similarly, when engineers designed cars, they did not model them after horses, even though horses were the primary mode of transportation at the time. Cars don't eat grass, and they don't gallop. They were designed to be faster, more efficient, and less labor-intensive than a horse. Again, breaking free from the model of what was familiar led to a more beneficial outcome.

AI can process vast amounts of data at speeds incomprehensible to the human brain, find patterns in complex data sets, and make predictions with remarkable accuracy. These capabilities are inherently non-human, yet they offer significant advantages. Just as we don't need airplanes to chirp or cars to neigh, we don't need AI to mimic human cognition to be useful, efficient, or innovative.

Moreover, machines do not get tired, stressed, or emotionally compromised. They do not have biases or preconceived notions. They can work continuously, consistently, and objectively, offering a different, often more reliable, form of intelligence. Expecting them to mirror human intelligence may not only be unnecessary but could also limit the unique capabilities they can offer.

GPS is a fitting example of a technology that doesn't need to mimic biology to offer immense benefits. By combining precise timing, triangulation from multiple satellites, and complex algorithms, GPS devices can pinpoint our exact location anywhere on the globe.

This capability is far beyond what any human, or for that matter, any biological organism, can achieve. Human navigational skills, which rely on landmarks, learned routes, and sometimes the position of the sun and stars, are extremely limited in comparison. Even the impressive homing instincts of certain birds, which rely on geomagnetic fields and visual cues, can't match the near-universal coverage and pinpoint accuracy of GPS.

GPS doesn't mimic human navigational skills; it transcends them. Its design did not try to emulate the cognitive mapping or landmark recognition that humans and animals use. Instead, it introduced an entirely new principle of operation, completely alien to our biology but astoundingly effective.

This technology highlights how we can take advantage of the unique capabilities that machines and algorithms offer, achieving feats that biology cannot. GPS is effectively a superpower, an ability beyond human reach, yet readily available to us thanks to the non-anthropomorphic design of technology. It underscores that machines don't need to mirror human capabilities to be beneficial; indeed, it is often their most non-human characteristics that make them most useful. What we should strive for is machine intelligence that serves its intended function optimally. For instance, if an AI's purpose is to predict weather patterns, it should be designed to analyze meteorological data with utmost accuracy. Whether it 'understands' what rain feels like or 'experiences' the heat of the sun is irrelevant. In this case, anthropomorphizing the AI doesn't enhance its functional capabilities and may lead to misplaced design efforts.

Similarly, an AI developed to diagnose diseases should be optimally designed to analyze patient data, medical histories, and scientific research. It does not need to empathize with patients or 'experience' human biology to perform its role effectively.

There's often a fear that if we do not instill human-like moral reasoning or empathy in AI, we might end up creating machines that can act unethically or harmfully. However, ethical considerations in AI design should not be about making machines 'understand' human morals but about incorporating checks and balances that prevent misuse. Accountability in AI should lie primarily with the humans who design, deploy, and oversee these systems, not with the machines themselves.

We should embrace the uniqueness of machine intelligence, much like we did with airplanes and cars. We should leverage AI's strengths – its speed, accuracy, consistency, and ability to process vast amounts of data – rather than strive to make it emulate human cognition. Designing AI with a clear purpose in mind, focusing on optimizing its functional capabilities, and ensuring responsible use, will lead to more beneficial and innovative outcomes than anthropomorphizing machine intelligence.

Predicting future

copyright 2022
simone di somma