Wetware Picture Credit: koniku.com
As AI becomes increasingly prevalent in areas as diverse as healthcare, finance, autonomous vehicles and speech recognition, the ‘big five’ tech companies – Apple, Google, Microsoft, Amazon and Facebook – are investing heavily in machine learning to improve decision-focused services for their customers. Machine learning has its limits however, requiring ever-growing computing and storage power coupled with high energy demands and processing costs. In contrast, biological cells generated from stem cells, can be applied to provide the computational functionality of real neurons with extremely low power demands. This development of so-called ‘Wetware’ represents one of the hottest issues in the ongoing advancement of AI-technology.
The Road Ahead for Bio-Computing
In a 2019 scientific paper Pathways to cellular supremacy in biocomputing | Nature Communications, several researchers outlined the potential of synthetic biology which applies engineering concepts and principles to the development of living systems. The promise of synthetic biology lies in its potential to provide new substrates for computation, production and medical diagnosis and to guide the ‘wetware’ of the living cell towards human-defined applications. In addition to providing tangible benefits such as cheaper drug production or more precise bio-sensing, the very process of re-engineering life will require a re-examination of our fundamental understanding of cellular processes. If we accept the notion that the motivation driving synthetic biology should be the search for a ‘killer-application’, we are about to stipulate a mindset of ‘cellular- supremacy’ in which cell-based systems will offer capabilities beyond the reach of existing computers. Simply put, cellular-supremacy will be determined by problems that biocomputers can solve and that traditional microprocessor-based devices cannot. To identify these problems, one must consider the ‘added value’ that living systems offer beyond silicon-based hardware. If we focus on problems that fall within the realm of traditional silicon-based machines and use existing metrics, then cellular computing will inevitably fall short on some of the criteria identified, such as speed and cost. However, the real power of cell-based systems derives from the richness of their internal architectures with a focus on several features of living cells which offer the most potential in terms of cellular-supremacy: reconfigurability of the cell, concurrency enabling massive-parallel communication and coordination within the cell, ability to use non-binary representations for signal-processing and the evolution to seek out novel solutions to problems over time. By harnessing these capabilities, we will eventually obtain bio-computing systems that are self-organising, self-repairing, resilient, distributed and adaptive.
First Steps to Building a Bio-Computing System
The ‘brain on a chip’, also called ‘Neu-Chip’ described in Stem cell AI: Loughborough part of £3m ‘brain on a chip’ project | Media Centre | Loughborough University (lboro.ac.uk), aims to revolutionize computing based on human stem cells to power a new generation of artificial intelligence (AI) devices. The project which started in January 2021 is designed to show how neurons as the brain’s information processors can be used to overcome traditional computers’ ability to learn while dramatically cutting energy use. The research team is embarking on a three-year study to demonstrate how human stem cells grown on a microchip can be taught to solve problems from data, laying the foundation for a ‘paradigm shift’ in machine learning technology. The team will layer networks of stem cells resembling the architecture of the human cortex onto microchips. They will then stimulate the cells by firing changing patterns of light beams at them. Sophisticated 3D computer modelling will allow them to observe any changes the cells undergo, to see how adaptable they are. This mimics the ‘plasticity’ of the human brain, which can rapidly adapt to new information caused by environmental change. The project funded by the European Commission’s Future and Emerging Technologies (FET) programme, with partner institutions from the UK, France, Spain, Switzerland and Israel, is also expected to produce new knowledge about the functionality of the brain which could be used to develop novel stem cell-based treatments for brain health. Professor David Saad, Professor of Mathematics at Aston University, said: “Our aim is to harness the unrivalled computing power of the human brain to dramatically increase the ability of computers to help us solve complex problems. We believe this project has the potential to break through current limitations of processing power and energy consumption of silicon-based computers to bring about a paradigm-shift in machine learning technology.”
‘Wetware’: From Academia to Start-Ups
The human brain is by far the most powerful, energy efficient computer ever created. Hence, what if we could harness the power of the human brain by using actual brain cells to power the next generation of computers? Answering this question, Neuroscientist Osh Agabi’s Start-Up company ‘Koniku’ Koniku® — Intelligence is Natural, originally founded in 2015, has developed a prototype 64-neuron silicon chip to mimic the brain functionality of a bee and its ability to navigate based on the detection and interpretation of smells. Funded by several venture capital companies including the US military, a drone that can smell explosives was their first project to test the concept. Such a drone would be able to smell bombs several kilometres away. It could also be used for surveying farmland, refineries, manufacturing plants – anything where health and safety can be measured by an acute sense of smell. Meanwhile the system has been further improved and is being tested by Airbus to detect explosives at airport luggage deployment areas. Based on induced pluripotent stem cell technology – a method in which cells (from the skin, for example) are genetically reprogrammed into a stem cell – neurons can be assembled. However, like electronic components, live cells need a specific environment to be able to operate. By creating a special shell for each neuron, the Koniku team can control the temperature and pH levels and can also send nutrients to the neurons to keep them alive. They are also able to control how the neurons communicate with each other because an electrode under the shell enables information to be read and written into the neurons. Agabi believes that harnessing the power and efficiency of the human brain is the future of computing. There are no practical limits as to how large we can make our devices or how much we can engineer our neurons. “I believe as intelligent computation goes, biology is the ultimate frontier,” Agabi says. “Biology can be the source of a far more advanced technology than anything ever created by man. We need to replicate biological processes to understand how much we must learn from nature’s billions of years of evolution. Maybe, in a few years, we will be able to put 10 billion neurons on a chip which might be capable of processing our entire knowledge space, including meaning and causation.”
The Ethical Issues
Companies which want to get engaged in wetware-computing do not need brain tissue samples from donors but can simply grow the neurons they need in the lab using stem cell technologies. Scientists can engineer cells from blood samples or skin biopsies into a type of stem cell that can then become any cell type of the human body. This raises ethical questions about donor consent. Do people who provide tissue samples for research and development know that the tissues might be used to make neural computers, and do they need to know this for their consent to be valid? Another key ethical issue regarding neural computers is whether they could develop some form of consciousness or experience pain. Research shows that lab-grown human brain-cells can engage in learning through real-time electro-physiological stimulation and recording. However, as recently discussed in a scientific study, there is no evidence that neurons on a dish have any qualitative or conscious experience, hence, they cannot be distressed and without specific pain receptors they cannot feel pain. Another ethical issue arises when discussing the possibility that organic computers are used for cognitive labour. Is this ethically more problematic than using a horse to pull a cart? Answering all these ethical issues is complex. We are confronted with a radically new phenomenon of science to which humanity has not been exposed so far.
To many of us, ‘wetware’ might stipulate a ‘Frankenstein’ or ‘Matrix’ science fiction phantasy. However, based on the reality of advancements in organic computing, we have moved way beyond any science fiction scenarios. We obviously need strong medical and governmental enforced regulation to govern the development and application of wetware and we need to think about the impact of wetware before things get out of control. On a very basic level, for example, there ought to be a kill switch built into any kind of wetware as the simplest form of constraint and control. Judging its scientific and commercial potential with venture capital as a key driver, we are probably only a few years away from a mayor paradigm-shift regarding the exploitation of human intelligence with wetware. Surprisingly – despite its beneficial as well as destructive potential – attention to this topic seems to lag significantly behind the urgent requirement in defining ethical constraints.
Hochinteressant. Vielen Dank!