Life 3.0

In his great book Tegmark explores the foundations, impact, meaning and possibilities for AI and ML in human history and beyond.

Importance of AI

Sketching out the history of our Universe, Tegmark then argues that it all -however physically interesting- is quite dull until a complex pattern able to maintain and replicate itself ie. life appeared. In chapter 1 Tegmark then talks about three stages of life:

• 1.0 simple biological - can survive and replicate
• 2.0 cultural - can “design its software”
• 3.0 technological - can “design its hardware”

Life 1.0 relies on its DNA for information exchange, whereas Life 2.0 in our cas uses synapses. ~100Tb in brain (10Gb electrically, 100Tb chemicallybiologically), ~1.6Gb in DNA. This classification is obviously a fuzzy one.

Next Tegmark goes on to talk about when AGI might appear and some common misconceptions regarding the real dangers and problems.

What is intelligence?

• Tegmarks defines intelligence as ability to accomplish complex goals.
• Hans Moravec’s “landscape of human competence” is discussed. It’s being flooded by computer/the sea

Memory

• Memory devices can be in many long-lived states
• Gold can be etched, water cannot; it takes more energy to change gold than water
• The simplest such device is a binary one: having two states
• Information can take on a life of its own, independent of its physical substrate
• Smallest known biological memory device is the genome of Candidatus Carsonella ruddii; 40kb
• John Hopfield: you can squeeze in 138 different memories for every thousand neurons without major confusion

Computation

• A transformation of memory states
• Intelligent matter must exibit complex dynamics
• NAND gates are universal
• computronium: any substance capable of performing arbitrary computations
• computation is substrate-independent

Learning

• Learning matter can rearrange itself to get better at computing its function
• Neural nets are universal - Maxwell Stinchcombe, White 1989

Cosmic endowment

• Advanced future life needs resources: bariyonic matter (quarks and electrons)
• Sunlight hitting 0.5% of Sahara covers our global energy needs

Energy, matter and computation

• Dyson spheres around sun
• O’Neill cylinders orbiting sun
• Digesting candy bar is not efficient, see table below
• Computing speed is limited by energy and maximum memory of matter - Seth Lloyd 2000

Mehod	Efficiency
Digesting candy bar	$10^{-7}\%$
Burning coal	$3\cdot10^{-7}\%$
Burning gasoline	$3\cdot10^{-7} \ %$
Fission of U-235	$0.08 \%$
Using Dyson sphere until Sun dies	$0.08 \%$
Fusion of H → He	$0.7 \%$
Spinning black hole engine	$29 \%$
Dyson sphere around quasar	$42 \%$
Sphalerizer	$50 \%$
Black hole evaporation	$90 \% ?$

Colonisation

• Available matter in our Universe depends on distance
• R. Hansons Great Filter: Technological/evolutionary roadblock for space-settling life

End of Universe

• Big Chill - Universe expands forever, cooling and darkening everything
• Big Crunch - reverse Big Bang
• Big Rip - Infinite expansion rate rips matter apart
• Big Snap - Big Rip assuming quantum gravity space granularity limits
• Death Bubbles - Vacuum reaching different energy level (phase of space) expands

Goals

• dissipation-driven adaptation: random groups of particles strive to organise themselves as so as to extract energy from their environment as efficiently as possible
• This is nature’s built in goal: producing self-oranising systems of increasing complexity
• Living systems maintains or reduces its entropy by increasing entropy arount it - “What’s Life?” Schrödinger 1944
• Its hard to align an AI’s goals with ours
• Humans are solution to evolutionary optimisation problem

Consciousness

• Important to know since we don’t want to leave the future of humanity in the hands of unconscious zombies
• Tegmark defines conciousness as subjective experience
• “If any scienteist wants to argue that subjective experiences are irrelevant, their challenge is to explain why torture or rape are wrong without reference to any subjective experience” - “Homo Deus” Yuval Noah Harari
• Particle perspective: Humans are food rearranged. Food is unconscious, humans are conscious. Why?
• qualia: individual instances of subjective experience
• Consciousness problems
  • 0 Easy: How does the brain process information? How does intelligence work?
  • 1 Pretty hard: What physical properties distinguish conscious & unconscious systems?
  • 2 Even harder: How does physical properties determine qualia?
  • 3 Really hard: Why is anything conscious?
• Scientific theory of consciousness, problem 2 above is testable

Human consciousness

• Humans do many (most) things unconsciously: we’re aware of what we do but not how we do it
• Of the $\approx10^7$ bits/sec of information entering our brain we’re aware of only a tiny fraction (10-50 bits)
• Only certain parts of the brain is conscious
• hemineglect: condition of missing half of the visual field without being aware
• the half billion neurons in the gut are not conscious
• neither is the cerebellum
• thalamus and rear cortex are suspects for the home of consciousness
• we get conscious of visual stimuli after about 250ms - Christof Koch

Integrated Information Theory $\Phi$ (IIT)

• Giulio Tononi’s IIT can be said to be the most successful of the consciousness theories out there
• $\Phi$ measures inability to split a process into independent parts
• Tegmark suggests the following necessary conditions for consciousness of a system:
  • Information principle: has substantial information-storage capacity
  • Dynamics principle: substantial information-processing capacity
  • Independence principle: substantial independence from the rest of the world
  • Integration principle: cannot consist of nearly independent parts

Large AI

• Earth-sized AI: 10 thoughts / sec
• Galaxy-sized AI: 1 thought / 100'000 years
• Large AI’s are thus incentivised to delegate tasks to subsystems
• One can even imagine a consciousness hierarchy

Free will

• Free-will is how a computation feels from the inside
• Tegmark argues that any conscious system will feel itself having free-will

Extracts

1. When a bacterium makes a copy of its DNA, no new atoms are created, buta new set of atoms are arranged in the same pattern as the original, therby copying the information. In other words, we can think of life as a self-replicating information-processing system whose information (software) determines both its behavious and the blueprints for its hardware.

   p. 24

2. Just as a blue whale is rearranged krill and krill is rearranged plankton, our entire Solar System is simply hydrogen rearranged during 13.8 billion years of cosmic evolution

   p. 218

3. … now we’re talking about the entire future of life in our Universe, limited by nothing but the (still not fully known) laws of physics, so defining a goal is daunting! Quantum effects asiem a truly well-defined goal would specify how all particles in our Universe should be arranged at the end of time.

   p. 277

4. As we’ve explored above, the only reason that we humans have any preferences at all may be that we’re the solution to an evolutionary optimisation problem. Thus all normative words in our human language, such as “delicious”, “fragrant”, “beautiful”, “comfortable”, “interesting”, “sexy”, “meanintful”, “happy” and “good”, trace their origin to this evolutionary optimisatoin: there is therefore no guarantee that a superintelligent AI would find them rigrously definable.

   p. 278

5. Some people tell me that they find causality degrading, that it makes their though processes meaningless and that it renders them “mere” machines. I find such negativity absurd and unwarranted. First of all, there’s nothing mere about human brains which as far as I’m concerned, are the most amazingly sophisticated physical objects in our known Universe. Second, what alternative would they prefer? Don’t they want it to be their own thought processes (the computations performed by their brains) that make their decisions? Their subjective experiece of free will is simply how their computations feel from inside: they don’t know the outcome of a computation until they’ve finished it. That’s what it means to say that the computation is the decision.

   On causality and free-will p. 313