First, I want to give credit and point to a resource I've very recently stumbled upon which dives deep into a wide range of philosophy pertaining to meaning: Meaningness (by David Chapman, a retired intellectual who did AI research at MIT)
I've only just begun to dip into David Chapman's enjoyable writings but I believe they are well-worth the time. For the sake of putting together the following narrative in a timely manner I've borrowed and tweaked his simple and effective visual formatting style.
I also want to thank my professor Greg Conradi Smith for stoking the fires of our ongoing learning journeys.
Today it's common to hear ideas thrown around about how the brain processes information like a computer or electrical circuits, but for centuries this same kind of thinking lead us to believe that the brain was some kind of mechanical or hydraulic machine. The field of neuroscience has only recently debunked these century old assumptions of how the brain physically functions (Hodgkin & Huxley, 1952), but it still lacks a clear understanding of how our nervous system does the seemingly magical things it does, like producing the rich conscious experience we enjoy today. Is there a correct principled way to approach these challenges of how our minds do the things they do with scientific understanding? Many prominent neuroscientists believe so, but any agreement beyond the simple biology of two neuronal connections is scarce.
It feels intuitive to think of our nervous system in terms of sensory inputs, complex internal processing, and motor outputs, and I don't think this simplified thinking is completely wrong, but I also don't think it's very useful. It doesn't help us understand what's going on inside that black box of "internal processing" to produce the complexity in thought we share. A proper evaluation of the undelying principles can greatly subserve artificial intelligence and medical research, as well as our own philosophical views. Because of this, I think the drive for progress may happen regardless of our individual efforts - which is both a comforting and terrifying idea to some. If the progress of society is out of our individual control - we need to at least seriously consider and spread awareness of the ethics tied to this progress.
What follows is subjective observations, ideas, and arguments that I have collected and organized into an interesting quilt which attempts to explain our minds to the best of our limited ability. Maintaining a healthy balance of skepticism and curiosity is part of what makes us so good at expanding our collective knowledge.
Outline:
1. Emergence, Entropy, Evolution
2. Simulations, Computation, AI
3. Existential Threats - delicate eddies on Earth
Current Understanding and Observations
Emergence
It can scarcely be denied that the supreme goal of all theory is to make the irreducible basic elements as simple and as few as possible without having to surrender the adequate representation of a single datum of experience. -Albert Einstein
I want to begin this journey with an observation of the world that has existed for a long time called emergence. This is the idea that simple components can come together and interact with eachother in some way to produce a kind of more complex social whole. The whole is something else than the sum of its parts, and often produces regularities that we can reliably study - which is why our scientific understanding is generally stratified at different discrete levels of abstraction (physics, chemistry, biology, psychology, etc). These are the patterns in the universe that most readily jump out at us. (Kurzgesagt, 2017)
In pursuit of a unified theory of everything physicists struggle to connect the two theories on which all modern physics relies on: Quantum Field Theory (the laws of the small) with General Relativity (the laws of the large). Somewhere between these scales, we exist to study them in addition to the many other fields we attempt to connect like physics with chemistry, and chemistry with biology. Perhaps the most contested and challenging connection is that between the mental mind and physical body.
What makes this so hard? Is the difficulty of bridging the gaps between some levels simply because it cannot be done, even in principle?
Descarte's dualism held this view to the extreme: where the mind is a separate entity that controls the physical body. The mind and body clearly interact but this view holds that an explanation of the mental in terms of the physical was simply impossible. This view supports the idea of strong emergence, where the whole system may be completely irreducible to its component parts. (O'Connor et al., 2015)
On the other extreme is Laplace's determinism, which held the idea that if a demon knew the precise location and momentum of every atom in the universe, then in principle it could calculate all of the past and future. This view rejects an idea like strong emergence because the causation is entirely bottom-up, and everything can be reduced in terms of its component parts. Emergent patterns may then just be a weaker form of emergence, where despite being possible (in principle) to reduce things into their component parts, it is much easier and more pragmatic to directly study those emerging patterns that exist. (Hawking, 1999)
I found these interviews on the topic (especially the last 2 videos) particularly interesting.
Philosophers of Science, Tim Maudlin and Barry Loewer seem to lean towards the principle ideas of Laplace with the qualification that our current understanding of the fundamental physical laws of the universe are currently very primitive. They appear to believe Laplace's demon may be possible in principle but not practical or even feasible for humans to accomplish by studying something like biology solely in terms of fundamental physical interactions. Tim Maudlin points out that ideas in these "higher level" sciences often originate in functional models like Mendelian Genetics and Hebbian Learning, preceding the discovery of their physical manifestations like the corresponding discoveries of DNA and LTP. He says that physics is the only field that can't say "that's not my department" when asked for an explanation that involves any matter in motion, and this is where the need for a theory of everything gets placed squarely on physicists. (Maudlin, 2018) This also wouldn't necessarily imply that our lives and behaviors are deterministic as Laplace believed because we don't even know for sure if the universe is deterministic or probabilistic due to our lack of a full comprehension of the fundamental physics.
Something I found especially compelling was that Barry Loewer points to Boltzmann's statistical mechanics as an important piece towards a theory of everything. Statistical mechanics explains how taking the probability distribution over all the possible micro-states of the universe is compatible with the macro-state we observe. This has been used to explain how thermodynamic behaviors like temperature and pressure relate to the ensemble activity of microscopic particle fluctuations. I find this to be promising support that many naturally emergent properties can - at least in principle - be explained in terms of their interacting parts.
Entropy
Boltzmann's work has been expanded and refined by others to explain the second law of thermodynamics: the entropy, or disorder, of a closed system will never decrease over time - instead it tends to increase until reaching a maximal state at thermodynamic equilibrium. Barry Loewer believes that these kinds of physical processes unfolding in one irreversible direction based on probability distributions could even account for the appearance of time flowing in a single direction. (Loewer, 2018)
These ideas have helped us explain the emergence of temperature in a jar of gas, but can they explain the emergence of life on Earth? Many scientists seem to believe so with the addition of a positive feedback loop created by Darwin's evolution by natural selection:
Evolution
Darwin's ideas are clearly implicit in these scientists' views because the algorithm of evolution by natural selection has built a strong explanatory bridge between the incredibly rare possibility that replicating systems first emerged and the subsequent rapid development in life's complexity. If we remain skeptical, we should ask ourselves if this is an oversimplification. Are small genetic mutations accumulating over time truly sufficient to account for the rapid development of complex evolutionary innovations like eyes, wings, and complex nervous system organization? Some prominent scientists and philosophers think there are important gaps here to be addressed.
A few resources which I have yet the chance to explore deeply include Daniel Dennet's Darwin's Dangerous Idea, Kirshner & Gerhart's The Plausibility of Life: Resolving Darwin's Dilemma, and Thomas Nagel's Mind & Cosmos. The premises of these books get at different issues with current evolutionary theory, but for now I want to focus on Nagel's ideas which I have found to be especially important to the discussion.
Nagel is a prominent advocate of the idea that the emergence of consciousness, subjective experience, and even life, cannot be reduced or explained by our current understanding of physics. One of his most famous pieces: What is it like to Be a Bat (1974) gets at the idea that there is an abstract feeling that comes with being a bat, and we are limited in our ability to fully comprehend that subjective feeling. Tim Maudlin made a similar point that physics is tasked with explaining all of matter in motion, and conscious feelings are not matter in motion.
The "feeling" of pain cannot be described by the physical interactions of your pain receptors, the "blueness" of blue cannot be accounted for by the physical properties of the light waves or the transduction mechanisms of the eye. I feel these are very strong cases for us to treat subjective experience as an especially challenging level of phenomenon which we are far from fully understanding, but I'm not fully convinced that the mental could never be understood in terms of the physical.
What would it mean for physicalism to be able to explain subjective experience? I'm sure it will require much more research, but a few functional ideas already exist that I put slightly more faith in than accepting the idea that there's no possible physical explanation.
Marvin Minsky's book The Society of Mind (1988) tries to conceptually describe the mind as emerging from the collective activity of individual 'thinking units' of the nervous system. This idea is at least conceptually plausible to me as our collective society appears to have a 'conscious mind' of its own that results from our individual activities. This doesn't really get us much further though, it just reaffirms our original idea of emergence.
Parts of our own bodies are varyingly conscious and physical damage or disorders can affect our cognitive function in a seemingly endless variety of ways. I think this at the very least establishes that consciousness is tied in some way to our physical states. Part of the subjective aspect may result from internal physical representations and models of the world that exist in complex and subjective feedback loops.
The functional idea of loops comes from Douglas Hofstadter in his books Gödel, Escher, Bach (1979) and I Am a Strange Loop (2007). I find the idea attractive that the appearance of downward causality is only an illusion resulting from the complex cyclic activity in our nervous system. (Hofstadter, 2007) I strongly feel this could provide insight into so many different complex cognitive processes like language, mathematical reasoning, and theory of mind, but (in addition to the lack of possible conceptual models) our ability to study such activity in humans has ethical and technical limitations.
If emergent behavior like consciousness relied on top-down influence (or strong emergence) then it would be impossible to simulate on a machine because we couldn't reduce it to a physical representation. So can we just plug computational theories into a computer and test them out? This is where a lof of hope has been placed in the field of AI, but if we really look at the current state of AI research, we can see where many issues lie.
Simulations, Computations, AI
What magical trick makes us intelligent? The trick is that there is no trick. The power of intelligence stems from our vast diversity, not from any single, perfect principle. -Marvin Minsky
I feel that simulations and toy models can be fun, engaging, and consequently powerful learning tools to experiment with. Because of this, I aim to make this a place for experimenting with simple educational simulations in the near future. The problem is that they also require immense expertise backing them if they are aiming to be realistic or explanatory in any reasonable way.
Simulating a brain is not feasible. Researchers are struggling to even simulate the exhaustively studied round worm with just 302 neurons. Even simulating a single cell is a computational challenge without some level of reductionist abstraction. Additionally, without also simulating the right external environment this wouldn't make much sense at all. If we follow Marvin Minsky and Douglas Hofstadter then there's a deeply suggestive intuition in our scientific understanding towards a bottom-up emergence of living systems and subsequent complex intelligent life. So can't we just simulate entropy and evolution to test the emergence of artificial life or AI?
Even if we had full understanding of the underlying physics, it would never be computationally feasible to model all of the physical interactions of our environment and calculate the results like Laplace's demon. Extracting meaning and information in the patterns of the noisy universe is functionally important. Our own energy efficient nervous systems prove to us this is possible, and this may be why our machines increasingly tend towards resembling ourselves.
Most attempts at conceptual simulations using varying principles like genetic algorithms, reinforcement learning, deep learning, or 'Good Old Fashioned AI,' point out the severe flaws and gaps we have in our knowledge of real "intelligence." These are all issues of "design" limitations. (Chapman, 2018) We're trying to play gods with powers of "intelligent design" without a true conceptual knowledge of the fundamental principles that should guide a simulation. So if this is true, what's with all the current AI hype?
We may never be able to "upload" our consciousness or create sentient machines - but our current AI's can still solve a handful of computationally interesting problems. Computer vision and natural language processing have seen suprising progress, but it's important to recognize that these deep learning developments are mostly only good at recognizing patterns in data. Recognizing patterns is an important first step which has found tons of use cases, but it says nothing about true computational understanding or logical reasoning. (Chapman, 2018)
Karl Friston, a pretentious but widely cited neuroscientist, believes he's come to a global theory of the brain that is so simple and powerful that it can explain the organizing principles of all living systems, which has also been regarded as holding the key to true AI. Unsurprisingly, it's named the free energy principle, and involves "active inference" loops to update internal representations of the external world:
I think there's no real shortage of innovation, so I don't want to dwell too long on the technical problems for now, but I think Minsky makes a good case for the idea that we're nowhere near understanding the seemingly convoluted and complex organization of our own nervous systems. I feel it is even more important to discuss how our technological advancements are shaping the world we live in. Innovation has radically changed the way we live, work, and interact with our environment, and further advancements appear unavoidable. This raises plenty of ethical considerations, but I specifically want to focus on the existential threats we present to the planet, because I think there are lessons to take away from generally intelligent life, and we can all do a little more to spread advocacy for the future of our species.
Global Existential Concerns
Our approach to existential risks cannot be one of trial-and-error. There is no opportunity to learn from errors. The reactive approach — see what happens, limit damages, and learn from experience — is unworkable. Rather, we must take a proactive approach. This requires foresight to anticipate new types of threats and a willingness to take decisive preventive action and to bear the costs (moral and economic) of such actions.
-Nick Bostrom
Selfishness and unregulated growth come at an external cost. Among our cells this looks like cancer, in sociology and economics this looks like the tragedy of the commons, in global life... this looks like 'general intelligence.' Existential risk philosophers like Nick Bostrom believe any sufficiently powerful general intelligence with unbounded goals, will also pursue ubounded "convergent instrumental goals." These are goals like self-preservation or resource acquisition that could generally help them achieve a wide variety of other goals. For example, if we built a generally intelligent agent whose objective was to simply collect stamps, it may do all sorts of unintended things to achieve this goal like stealing, murdering, or enslaving others for its cause. (Bostrom, 2012) It's much more concerning to me how often humans do this to themselves.
For some people, accumulating money is a convergent instrumental goal because it can be used to achieve a wide variety of terminal goals (you can buy almost anything with enough money). (Miles, 2018) As a result we see all kinds of unintended consequences in society that look not too different from variations of the stamp-collecting situation. This, in conjuntion with our current global economic competition, seems to force individuals to be short-sighted in our decision-making. It then becomes very challenging to deal with long-term global sustainability issues like climate change and inequality.
For now, there is hope that humanity will course-correct as problems become less abstract and we begin to prioritize collective action. Suggestions in psychology include renaming "climate change" to "climate crisis" and sharing specific narratives that are more likely to stick in our minds than the abstract reality of global temperatures and sea-levels. But this also begs the question of how delicate and rare intelligent life really is?
A mass-extinction event probably wont wipe out all life on Earth, and the survivors may rebuild a wide diversity of intelligent life over time. It seems fairly reasonable that humanity could become an ancient civilization studied by a future intelligence. This makes life sound quite resilient and we would then expect to find it on other planets like our own. This creates a confusing paradox that was named after Enrico Fermi:
If we seriously consider the idea of Great Filters, then it's reasonable to think filters may exist both ahead and behind us. If the emergence of life through abiogenesis is possible, life may be more unique than we understand. The filters ahead of us likely just involve living even more flexibly and sustainably with the environment and extending reach into space. This is all speculation of course, but some post-humanist ideas are radical enough that I think humanity will come to a strange renaissance soon if it doesn't wipe itself out first.
When we get to a certain point in this kind of speculation, it all becomes a little too abstract and incomprehensible for me. This seems to be the point where many folks find their own ultimate guiding principles and -isms, which for some reason was the sort of conclusion I originally had in mind. Instead, I think we should at the very least try to limit the existential threats on our delicate planet in the best ways we can: with our collective care and action.
- Minsky, ML 1988, Society of Mind, Simon & Schuster, New York.
- Abbott, R 2006, ‘Emergence explained: Abstractions: Getting epiphenomena to do real work’, Wiley Periodicals, Inc., vol. 12, no. 1, pp. 13-26. Available from: https://onlinelibrary.wiley.com/doi/pdf/10.1002/cplx.20146. [1 May 2019].
- Glasgow, RDV 2018, ‘Minimal Selfhood and the Origins of Consciousness’, Würzburg University Press. Available from: https://www.academia.edu/37069229/Minimal_Selfhood_and_the_Origins_of_Consciousness. [1 May 2019].
- Nagel, T 1997, ‘What Is It Like to Be a Bat?’, The Philosophical Review, vol. 83, no. 4, pp. 435-450. Available from: JSTOR. [1 May 2019].
- Hofstadter, DR 1979, Gödel, Escher, Bach: An Eternal Golden Braid, Basic Books, New York.
- Hofstadter, DR 2007, I Am a Strange Loop, Basic Books, New York.
- Kurzgesagt - In a Nutshell 2017, Emergence - How Stupid Things Become Smart Together, YouTube video, 16 Nov. Available from:https://www.youtube.com/watch?time_continue=2&v=16W7c0mb-rE. [1 May 2017].
- It’s Okay To Be Smart 2018, Where Did Life Come From? (feat. PBS Space Time and Eons!), YouTube Video, 11 Apr. Available from:https://www.youtube.com/watch?v=_uAJY1mqtw4. [1 May 2019].
- PBS Space Time, The Physics of Life (ft. It's Okay to be Smart & PBS Eons!) | Space Time, YouTube video, 11 Apr. Available from: https://www.youtube.com/watch?v=GcfLZSL7YGw. [1 May 2019].
- Chapman, D 2018, ‘How should we evaluate progress in AI?’ Meaningness Metablog, blog post, 30 June. Available from: https://meaningness.com/metablog/artificial-intelligence-progress. [1 May 2019].
- Bostrom, N 2012, ‘The Superintelligent Will: Motivation and Instrumental Rationality in Advanced Artificial Agents’, Minds and Machines, vol.22, no.2, pp.71-85. Available from:https://link.springer.com/article/10.1007%2Fs11023-012-9281-3. [1 May 2019].
- Serious Science, Free Energy Principle - Karl Friston YouTube video, 16 Jun. Available from: https://www.youtube.com/watch?v=NIu_dJGyIQI. [1 May 2019].
- Amodei, D, Olah, C, Steinhardt, J, Christiano, P, Schulman, J, & Mané, D 2016, ‘Concrete problems in AI safety’, arXiv preprint arXiv:1606.06565. Available from: https://arxiv.org/pdf/1606.06565.pdf. [1 May 2019].
- Friston, K, Kilner J, & Harrison L 2006, ‘A free energy principle for the brain’, Journal of Physiology-Paris, vol. 100, pp. 70-87. Available from: https://www.fil.ion.ucl.ac.uk/~karl/A%20free%20energy%20principle%20for%20the%20brain. [1 May 2019].
- O’Connor, T & Wong, HY 2015, ‘Emergent Properties’, The Stanford Encyclopedia of Philosophy, 3 June. Available from: https://plato.stanford.edu/entries/properties-emergent/. [1 May 2019].
- Hawking, S 1999, Does God Play Dice, Available from: http://www.hawking.org.uk/does-god-play-dice.html. [1 May 2019].
- Miles, R 2018, Why Would AI Want to do Bad Things? Instrumental Convergence, YouTube Video, 24 Mar. Available from: https://www.youtube.com/watch?v=wX78iKhInsc. [1 May 2019].
- Closer To Truth Interview Series: What is Strong Emergence? Tim Maudlin 2018, television program, PBS. Available from: https://www.closertotruth.com/series/what-strong-emergence. [1 May 2019].
- Closer To Truth Interview Series: What is Strong Emergence? Barry Loewer 2018, television program, PBS. Available from: https://www.closertotruth.com/series/what-strong-emergence. [1 May 2019].
- Kurgesagt - In a Nutshell 2015, The Fermi Paradox — Where Are All The Aliens? (1/2), YouTube video, 6 May. Available from: https://www.youtube.com/watch?v=sNhhvQGsMEc. [1 May 2019].
- Van Lange, PAM & Bastian, B 2019, Reducing Climate Change by Making It Less Abstract. Available from: https://www.scientificamerican.com/article/reducing-climate-change-by-making-it-less-abstract/. [1 May 2019].