Life 3.0: Being Human in the Age of Artificial Intelligence

Short for “cosmic apocalypse,” this refers to the ultimate destruction of the entire observable universe. This will likely be tens of billions of years in the future and will bring an end to all life, regardless of advancement or intelligence. In Chapter 6, Tegmark discusses five different cosmocalyptic scenarios.

This refers to any physical material that is constituted by atoms or subatomic parts. Tegmark speculates that, at the upper limits of technologically advanced superintelligence, baryonic matter will be the only resource necessary because technology will be able to rearrange the parts of the atoms at will. All physical material will ultimately be rearrangeable into different material.

AIs are capable of so-called “deep” learning when they are developed with neural networks, a circuitry of arranged neurons, and then exposed to “massive amounts of data.” Deep learning has already been integrated in much of present-day AI, including the self-driving vehicle.

This is a form of deep learning that involves the use of reward and punishment for particular behaviors. Tegmark describes how an AI created by DeepMind developed its capacities because it was continuously given points when it made the right moves in a video game. It treated this as a reward, thereby encouraging it to behave in ways that would maximize this outcome.

Digital Utopians are those who believe that humans will be usurped by more highly intelligent AIs in the near future and that this will ultimately be a good thing. Tegmark cites Larry Page, co-founder of Google, as an example of a Digital Utopian. Digital Utopians are contrasted with Techno-Skeptics, Luddites, and members of the “Beneficial AI Movement,” like Tegmark.

This idea, originally developed by Freeman Dyson, involves the rearrangement of vast amounts of planetary material in a large sphere surrounding a star. Tegmark speculates on the construction of a Dyson Sphere, which could be part of the technological advancement of human and AI society in thousands of years. The Dyson Sphere would provide intelligent beings with an incredible amount of biomass and energy.

Exponential growth, Tegmark writes, occurs when a “power grows at a rate proportional to its current power,” which means that it will continue to double (158). This can cause radical change to occur at faster rates over time in a process known as an explosion. Tegmark thinks it’s possible that AI advancement could be subject to exponential growth, meaning that its development may enter an unprecedented upsurge in the not-too-distant future.

Tegmark emphasizes that there is no consensus on what exactly intelligence is. Therefore, he defines it broadly as “the ability to accomplish complex goals” (50). The term “complex,” he writes, is “deliberately vague” (50). Tegmark’s definition of intelligence does not necessitate consciousness or subjective experience. He also distinguishes between broad and narrow intelligence. Narrow intelligence is the ability to accomplish narrowly defined goals while broad intelligence encompasses a larger range of goals and abilities.

Tegmark writes, “we can think of life as a self-replicating information-processing system whose information (software) determines both its behavior and the blueprints for its hardware” (25). This definition, also deliberately vague, does not require biological material, and is open to the possibility of artificial life.

The titular concept refers to lifeforms that are able to freely transform both their information (software) and the physical material that constitutes the hardware. For Tegmark, the entails a technological evolution that would totally free these lifeforms from their biological/evolutionary inheritance (26). Human beings, according to Tegmark, are members of Life 2.0 because we can transform our software but not our hardware.

Machine learning, closely associated with related concepts like artificial neural networks and deep (reinforcement) learning, is the ability for certain algorithms to self-improve in dialectical response to experience. Machine learning is a very important subfield in AI research, especially for the development of more intelligent creatures.

According to game theory, a field of mathematics, Nash Equilibrium is a state in which all parties are in the best possible position with respect to their own goals and desires if they remain in the current situation. It’s “a situation where any party would be worse off if they altered their strategy” (151). This may mean that power is unequally distributed amongst various parties and is commensurate with the possibility that layers of hierarchy could be in everyone’s optimal interest.

The “physical correlates of consciousness” (PCCs) are simply the patterns of physical material that correspond to conscious events, i.e., subjective experience. NCCs, or neural correlates of consciousness, are the specific patterns of the brain that correspond to biological consciousness. Tegmark believes that NCCs are merely a subset of all possible PCCs, which might mean that it is perfectly reasonable to posit the theoretical possibility of artificially constructed conscious beings.

A word of Tegmark’s invention that classifies any sentient substance, i.e., any subject of experience. Note that this refers to a class of beings that overlaps with but is not necessarily synonymous with living beings.

If information is substrate independent, it is not beholden to or delimited by its relationship to its material form. For Tegmark, this fact is particularly important when thinking through the nature of consciousness. It is possible, on this model of reality, to digitally upload conscious entities onto computer software. Minds may be housed in physical shells of one form or another, but they are not strictly bound to that particular form.