Designing with Extended Intelligence

This week was an intense crash course in the core concepts in the fields of artificial intelligence (AI), machine learning (ML) and subjects relating to extending human intelligence. I have been exploring these themes for several years now, so the course was right down my alley. This has allowed me to explore more speculative and provocative ideas in some of our group discussions. Besides some experiments with Faceswap, I have been mostly focused on the ethical and sociocultural aspects of AI. Therefore I found the technical primer into the underlying mechanisms of machine learning particularly valuable.

Throughout the course, Ramon Sangüesa provided firm philosophical grounding and plenty of food for thought, while Lucas Peña delivered a barrage of in-depth technical concepts and provocations that made us second-guess our initial assumptions. They both worked exceptionally well together and complimented each other to present a comprehensive and engaging overview of the field.

What follows is a sort of a free-form essay related to particular aspects of AI/extended intelligence that I find most salient, intriguing and relevant to my future research.

Intelligence decoupling

Right at the outset of the week, we got into a lively debate about what constitutes intelligence and how do we decide which organisms (or objects) are intelligent. We have come up with various descriptions and partial-definitions, but eventually, we have ended up with a conception so broad and vague that we needed to consider virtually everything around us at intelligent to a higher or lesser degree.

Our conclusions are in line with a techno-optimist strain of Kurzwelian thought that believes that AI will cure all of our ills and that we will ultimately merge with the machines in the grand moment of the singularity. This idea is stemming from a mechanistic view of intelligence as a form of computation based on informational inputs and outputs where the ‘computation is substrate-independent in the same way that information is: it can take on a life of its own, independent of its physical substrate’ ¹.

Thus if intelligence is embedded at a molecular level as a material called computronium - a form of programmable matter – then there is no question whether it is possible to create an inorganic AI that is as (or more) intelligent than humans. It is just a matter of generating a sufficient computational power. Neural networks are directly based on the metaphor of the human brain and simulating connections between neurons.

With that being said, intelligence does not equal consciousness and this is where our scientific theory falls utterly short to the point of sheer ignorance. Men-made AI is improving at an exponential rate but we are nowhere near any sort of artifitical consciousness. The decoupling of intelligence from consciousness presents a major challenge because the most thorny questions facing humanity and its relationship to technology and the spaceship Earth are ethical, moral and cultural – all products of our imagination with no basis in the physical reality.

https://www.nytimes.com/2018/11/05/opinion/artificial-intelligence-machine-learning.html

Unleashing superintelligent entities out in the world without a proper understanding of consciousness is a dangerous prospect. The next section explores some of the more worrisome implications of currents trends in AI.

A false promise of AI

AI has gone through several hype cycles in the past and some sceptics suggest that the current preoccupation with AI is overstated. However, the accelerated progress and convergence in the fields of genetics, biology and computation indicate that this time around the promise (or threat) might be for real this time around.

Fears of AI promoted by popular fiction are sensationalist and deeply misguided. AI does not present a new form of a threat in itself; instead it is set to exacerbate existing problems. According to Yuval Harrari, one of the most eloquent thinkers to emerge in the recent past, AI does not need to be conscious, super-intelligent nor even human-level intelligence to take over the world. It just needs to be good enough to outperform us at specific tasks that are paramount to the way global civilisation operates.² AI is already swaying democratic elections and controlling financial markets and is soon poised to take over our roads, healthcare and virtually every aspect of human existence.

“People worry that computers will get too smart and take over the world, but the real problem is that they’re too stupid and they’ve already taken over the world.”

– Pedro Domingos, The Master Algorithm (2015)

As we inch closer towards developing a true Artifical General Intelligence (AGI), we face risk, not of malice but competence, as illustrated with a paperclip maximizer thought experiment, attributed to Nick Bostrom. In this fictional scenario, an owner of paperclip factory orders an AI to optimise production of paperclips. This seemingly benign task leads the AI to destroy the earth to obtain more material so it can conquer other galaxies, eventually turning the entire known universe into a giant super-computer that produces paperclips. A Frank Lantz from NYU has created a game inspired by this concept – Universal Paperclips – despite being entirely text-based, it is surprisingly entertaining (and informative). Be warned, once you start playing it you will not let go until you finish it which will take a minimum of 4-5 hours.

AI totalitarianism

We tend to see AI as this looming otherworldly entity, epitomised throught Terminator-like imaginary, that appears dettached from our existing socio-economic structures. However, with the current trajectory the more likely scenario is that the progress in AI will further concetrate power and resources in the hands of technocratic elite that have an exclusive say in defining the parameters of AI’s activity.

Ubiquitous biometric surveillance, predictive poicing, digitally-enabled behavioral modification at the scale of an individual, social credit system, autonomous weapons – these are just some of AI-powered Black Mirror-esque realities of day-to-day life in contemporary China that give us a sneak peak of things to come (or already here depending on your perspective) as part of a surveillance capitalism that is taking over the globe ^{3 4 5}.

Reality blending

An area of AI I found particularly interesting is synthetic media ⁷. By letting one neural network train another one using a method called generative adversarial network (GANs), these algorithmic systems are capable of producing photorealistic material that is increasingly difficult to discern by human eye. One of the first widely medialised examples of how synthetic media can be misused were so-called deep fakes that let you use existing material to generate a completely fabricated video of someone saying whatever you put in their mouths using an open-source toolkit.

“Historically, it will turn out that there was this weird time when people just assumed that photography and videography were true. And now that very short little period is fading.”

– Alexei A. Efros (2018)

Fake media is not new. What has changed is the scale and quality of the fabricated content and accessibility of the tools used to produce it. Using standard off-the-shelf computer components anyone with a little patience and a bit of coding skills can synthesise events that didn’t happen by exploiting human’s pechant for visual imaginary. ⁸ As VR and AR technologies get increasingly better, more ubiqutious and immersive, we will experience a total blending of realities throught synthetic ambient media. ⁹

What are the implications of not being able to believe what you see anymore? What happens to a human mind when it cannot discern digital from physical? Are we already living in a simulation?

Ethics of the unknown

Discussions about developing ethical codes and regulating AI are all the rage at the moment. It is an incredibly tricky endeavour and also possibly one of the most critical discussions of our lifetimes. Science cannot help us here, it is an area entirely dependent on the peculiarities and idiosyncrasies of the human mind. It seems that the best we have come up with so far is crowd-sourcing moral judgments through projects such as Moral Machine project from MIT that lets people judge acceptable outcomes of a trolley problem – a thought experiment widely used to explore ethics in the context of self-driving cars.

At first, such an approach might seem adequate and fair. But no matter how many people we survey, such ethical code will inherently be arbitrary and riddled with biases specific to given programmers and the culture they are embedded in. Nevertheless, when it comes to self-driving cars, even a profoundly flawed moral code will still likely result in a dramatic decrease in traffic accidents. Same can probably be said about AI weapons which will likely lead to hyper-localised conflicts and reduction of civilian casualties.

I am in no way advocating for the development of autonomous killing machines. It is just to show that by following a strict rationalist logic, humanity will be better of with AI taking over control of our lives, including in a scenario where it takes over the world and establishes an AI-cracy. Sure, we would live in some sort of perverse hyper-stimulating Huxley-an simulacrum where all of our deepest desires are fulfilled through brain-machine interfaces and bioengineering. But on the flip side, we eradicated hunger, warfare and most forms of human suffering.

Most people would agree that is not the future they want to live in. But actions speak louder than words. We need a spiritual revolution, not an AI one.

Extending consciousness

The problems with AI are not technical but cultural. We need to move away from a scientific dogma that claims that human agency can be reduced to a predefined set of electrical signals in our brains. To do so, we need to critically engage with an age-old question of the meaning of life in the context of Anthropocene and expanding capabilities of biotechnology and genetic engineering. We are probably not going to figure out what is the meaning of life or what is the true nature of consciousness by the means of science any time soon. Everyone needs to tackle those questions by themselves and find out what constitutes content and fulfilled life – and AI can serve as a powerful tool to aid that process of self-realisation.

Our societies overrate Design (Engineering, Optimisation, Control, etc.) and underrate the simple enjoyment of the eternally emerging present moment. pic.twitter.com/FUzTzdXULB

— samim (@samim) November 14, 2018

​

The public discourse about AI at the moment is characterised by fear-mongering and illegibility stemming from a lack of technical understanding in which the technology is seen as something imposed from top-down and detached from the everyday life of citizens. To unleash its humanistic potential, we need to view AI as a commonplace tool available to everyone according to their individual needs. Particularly in the area of food, agriculture and healthcare, AI can help alleviate poverty and bring upon new forms of radical self-sufficiency by empowering everyone to be healthy and well-fed without unnecessary drudgery, the pressure to earn a living or being reliant on central forms of control and allocation of resources.

“It won’t matter whether computers will actually be conscious or not. It will matter only what people think about it.”

— Yuval N. Harari, Homo Deus (2017)

Such a shift requires reevaluating our core values and beliefs about how we relate to other living organisms, environment and technology. Concepts such as slow computing ¹⁰, tangible media ¹¹ or screenless interfaces ¹² represent a different relationship towards computing that hint at a possible future in which there is a symbiotic relationship between people and machines that operate on a human scale.

Symbiotic meta-consciousness

Here I would like to briefly explore a more esoteric approach to consciousness that is often dismissed by the mainstream academia as pseudo-science. A growing body of research into biocommunication, plant neurobiology and plant perception is gradually uncovering an immensely intricate, intertwined and complex world of plant intelligence. ¹³ We now know that trees, plants and other macromolecular organisms exchange information via a vast network of links running under the surface of forests. ¹⁴ Fungi and mycelium are increasingly proving to be a sustainable answer to many of the environmental ills facing humanity today. ¹⁵ Or multicellular organisms that defy categorisation, such as slime mould that has a remarkable ability to create intelligent self-organising networks and shows signs of learning.

After years of debates, the general scientific consensus is that animals are sentient – they can feel and perceive things. Is it maybe time we consider plants and other organic beings sentient as well? Can we go a step further and call them conscious as well? Countless indigenous tribes around the world certainly believe so, and many oriental cultures are open to the notion of some form of metaphysical cosmic consciousness permeating everything around us. In the positivist Western cultures, such views are still considered esoteric and borderline crazy.

However, in recent years we have witnessed a resurgence of research into the effects of psychedelics after decades of an academic embargo on the topic. Psychedelic substances are proving to be a potent tool for exploring the hidden depths of the human psyche. Maybe more importantly, they show us how little we actually know of what is going on in our mind. An author and ethnobotanist Terence McKenna went as far as to suggest that human consciousness has developed as a direct consequence of information received through a psychedelic experience induced by psilocybin mushrooms which in his view are conscious beings that come from outer space. ¹⁶ Admittedly, it is a far-fetched theory that is difficult to take seriously. But it is valuable in a sense that it suggests that we could extend our consciousness via a symbiotic connection with other living organisms – similar to how a humanoid species of Na’vi communicate with the spirit of the forest via a direct neural link in the movie Avatar (2009).

“I see the mycelium as the Earth’s natural Internet, a consciousness with which we might be able to communicate. Through cross-species interfacing, we may one day exchange information with these sentient cellular networks. Because these externalized neurological nets sense any impression upon them, from footsteps to falling tree branches, they could relay enormous amounts of data regarding the movements of all organisms through the landscape.”

— Paul Stamets, Mycelium Running (2005)

The metaphysical approach presented here is highly speculative and not compatible with how technology and science are developed today. This is a shame because it offers a much more holistic view of the life of planet Earth – one that is not preoccupied with particularities of the human brain but looks at the world as a tightly integrated symbiotic system based on collaboration between organisms rather than Darwinian dog-eat-dog dogma. I acknowledge its limitations, and I am not suggesting that it contains all the answers. I am only saying that it might serve as a worthy spiritual and moral compass for developing AI and related technology, especially in the context of environmental challenges facing humanity.

What would the world look like if humans merged with plants and machines into a single hyper-conscious meta-organism?

Takeaways

I did not intend for this essay to come across as pessimistic but re-reading some of the passages I see how it may sound like AI is all dark and gloom. I like to believe that I am aware of the untapped potential of AI and biotech to usher humanity into a new era of flourishment in harmony with the planet Earth. But I also believe that the first step towards such a future is clearly acknowledging and challenging the status quo.

I was happy to find out that none in the group identified with the reductionist views of the human psyche. That makes me more optimistic about the emerging futures to come. The course has helped me revisit some of my existing preconceptions of AI which has, in turn, inspired me to start exploring notions of extended intelligence which go beyond the purview of technology and are more in-line with my values and ideals.