By Anthony Lewis, Chair of Windsor Humanists and of the South Central England Humanists Network
Anthony claims that we are already using AI in various forms every day. AI is advancing across a broad front from machine learning, data mining, and operational optimisation to sensory perception, autonomous robots, and human augmentation. At its highest level, AI could lead to superintelligence and transhumanism. It could also help us solve the mystery of consciousness and, ultimately, might even save us from ourselves.
Artificial intelligence is already being used in many aspects of our daily lives. For example, I have used AI to help me write this article using both Google and ChatGPT. ChatGPT defines Artificial Intelligence as 'the simulation of human intelligence in machines that are programmed to think and learn like humans'. This can include tasks such as problem-solving, decision-making, perception, and language understanding. Al can be implemented in various forms, such as expert systems, neural networks, and natural language processing. This is very close to how the McKinsey Consultancy defines AI: ‘the ability of a machine to perform cognitive functions typically associated with human minds, such as perceiving, reasoning, learning, interacting with the environment, and problem solving’. Who needs management consultants anymore?
"The Information Revolution of which AI is an integral part... will transform our world as it progresses"
As with all major new technologies, AI will be disruptive. But I believe it will continue to transform our lives for the better. At present, the rapid development of AI is a classic reinforcing cycle of fundamental university scientific research driving technological innovation, leading to both the transformation of existing businesses and the creation of whole new industries. These in turn generate huge amounts of cash and wealth through commercial success, because people buy and use the new products, technologies and services. This new capital is then re-invested almost entirely back into new research and innovation, leading to another wave of accelerated research and innovation. This is a positive, symbiotic process that is very similar to what occurred during the Industrial Revolution several centuries ago. This new Information Revolution, of which AI is an integral part, has been under way for the last fifty years and is equally transforming our world as it progresses.
There are always fears about new technology. AI has been in the headlines a lot recently. For example, in The Times on 26th January 2023, there was an article with the headline ‘Rogue AI could kill everyone’. The present hype around AI is a prime example of ‘Innovation Exuberance’, where researchers and entrepreneurs compete for investment funds, leading to the over-selling of their ideas and proposals. This marketing hype then leads to unrealistic expectations of the benefits of any new technology such as AI and also creates exaggerated fears about its negative impacts. However, it is important to remember that the development of successful technologies always proceeds by trial and error, and is a gradual, rather mundane and incremental process. I remember in the late eighties and mid-nineties having to persuade my reluctant bosses about the benefits of being able to email ‘outside the company’ and for our scientists to have individual internet access to speed up their research. Historically, the adoption of new technologies has almost always led, eventually, to an overall increase in the quantity and quality of employment. This is despite the short term but often permanent disruption to existing jobs and industries. As Kim Povlsen, CEO of Universal Robots, has observed in an article for The Economist: ‘None of the countries that have adopted robots on a large scale has a problem with unemployment’. If the advantages were not, on balance, beneficial to most of us, the funds driving the innovation cycle would very quickly run dry.
A central tenet of humanism is that the flourishing of humanity has been driven in large part by scientific and technological progress; and I believe, as a humanist, that AI has the potential to be far more benevolent than malevolent. However, some of the more malevolent uses of AI could pose an existential threat to humanity on a par with that posed by nuclear weapons and biological warfare. It is reassuring, then, that research into the ethics and dangers of AI is now attracting attention and funding. For example, Stephen Schwarzman, the billionaire CEO and co-founder of the private equity firm Blackstone, recently gave Oxford University the largest private donation it has received since the Renaissance, to invest in the humanities and set up the Institute of Ethics in AI led by Professor John Tasioulas. No doubt humanity will manage the existential dangers in a similar way to those implemented for nuclear and biochemical weapons. I will return to the balance sheet between the potential of AI and its dangers towards the end of this article.
Artificial neural nets and genetic algorithms have been used by the energy industry for nearly thirty years to speed up the interpretation of large 3D seismic datasets. So AI has been around for a long time, and is already embedded into many businesses and in the software tools we are all using daily. Let's explore then a few areas where AI technologies are already being used. The diagram below provides a bird's-eye view of the ‘AI Toolbox’. On the bottom left, AI is driven mainly by data and software such as machine learning, whilst towards the bottom right there is a focus on developing physical hardware such as automated machinery. The vertical axis represents increasing levels of cognition (the process of acquiring and understanding reality through experience) and sentience (the capacity for experiencing both physical sensations and emotional states). At the very top, AI could possibly lead to the emergence of artificial consciousness and even multiple super intelligences. There is too much to cover in a short article like this, so we will focus on five areas: Machine Learning, Operational Optimisation, Autonomous Robots, Transhumanism and Consciousness.
Machine Learning
Browser search engines are examples of ‘Data Mining Tools’ that use artificial intelligence algorithms to quickly find the information that we are looking for on the internet, based on what they already know about us and the information sitting on the web. Data Mining is an area where AI has made the most progress in, for example, the interpretation of large datasets such as 3D seismic mentioned above, and the automatic analysis of medical scans such as MRIs or CT scans which are used to assist diagnosis. As everyone who has used Google Translate or a spell checker knows, many of the current data mining technologies are not perfect but they do enable us to do things that used to be impossible. It's so embedded in our daily lives we already take much of it for granted.
The rapid improvement in Sensor Technologies, for example in digital photography, has led directly to ‘face-ID’ tools that are now routinely installed on our mobile phones to improve both their physical and online security. The widespread deployment of CCTV technology has improved the security of our homes and our public spaces. And the police are able to monitor spaces in real time and search CCTV data using data mining and face-ID tools in crime investigations. As with all new technology, the technologies which can be used positively to fight and prevent crime can also be used by repressive regimes to build surveillance systems to control their populations, as China did during their pandemic lockdowns.
"Virtual ‘lawbots’ will revolutionise our justice systems but they will also disrupt the legal profession..."
Machine Learning is helping us to analyse, interpret and understand the vast amount of data that we have collected. This area is attracting the greatest investment at the moment and no doubt there will be some amazing advances over the next decade. ChatGPT is a recent example of one of these emerging tools which use ‘language engines’ to produce understandable text based on ‘calibration’ databases containing reports, books and writings. At the moment, it is very much an English language tool which operates, in effect, like a ‘prose or software checker’, analogous to how a spell checker operates for spelling. So it's not perfect. A second example, reported by medical researchers, is the discovery that some ovarian cancers can be identified up to a year before a medical diagnosis, based on women’s shopping habits, by running Tesco’s customer database through a machine learning program. A third example is virtual ‘lawbots’ which are being developed to help guide the courts on defence and prosecution strategies and even sentencing, based on the entirety of case law. This has the potential to revolutionise the operation of our justice systems but it will disrupt the current structure of the legal profession. A fourth example is the very likely application of AI pattern recognition algorithms to our growing DNA databases which will lead to improving health care and medical diagnosis.
Operational Optimisation & Simulation
Many of our physical manufacturing assembly lines are already almost completely automated. The whole process of building a car is now done by automated machines, which has reduced costs and increased safety whilst also improving the quality of the finished products. In agriculture, harvesters are now semi-automated mobile factories which contain sensing technology to detect and reject sub-standard plants, vegetables, and fruit before they reach the packing plants, which themselves are also automated. We all use AI on a daily basis to improve our individual efficiency at work. For example, junk filters clear out unwanted emails using AI engines, saving us a lot of time. Also, anti-virus software automatically keeps our computers free from harmful rogue software with little input from ourselves. Our suitcases arrive with us at our destination on the correct belt at the correct airport due, in large part, to automated bar code trackers.
Organisations can now analyse and model their operations using AI tools to optimise their processes to reduce costs and improve quality. For example, many companies like Amazon, Asda and Tesco all use scheduling software to optimise their delivery schedules, logistics and inventory management (also known as 'Just-in-Time' management), all based on the collection of huge amounts of tracking data made possible by bar code scanning technology. When you order a taxi, a generic algorithm is likely to have chosen the optimum car to pick you up based on the real time tracking of all of their positions using GPS. Such a task would outstrip the ability of most human operators as the size of the fleet grows.
"Computer simulations are like 'virtual realities’ , analogous to our own internal mental models of reality"
Computer simulations incorporating both the science and calibration databases are now used routinely to model many business operations, and complex systems such as the Earth's climate, our economies and human behaviour. Many of these programs are essentially ‘virtual realities’ that are analogous to our own internal mental models of reality. They can be used ‘offline’ to investigate problems, study dependencies and identify potential improvements, without interfering with ongoing operations. In the airline industry, using flight simulators is now routine to help improve pilot training and overall flight safety. The energy industry uses computer simulation models incorporating AI algorithms to help predict demand for oil, gas and petrol, based on short- and long-term weather forecasts and economic data, so that they can make sure the resources are in the right place at the right time. These models are now so reliable that British Gas, for example, during the winter of 2023, were able to offer customer discounts to reduce their electricity consumption at specific times and on specific days. Some of the largest computer simulation models, such as the HadCM3 at the UK Meteorological Office, have been built to study the Earth's climate and have been used by the UN IPCC in their Third and Fourth Assessments.
Autonomous Robots
Controlled Robots or ‘Cobots’ are already all around us. The automatic security and passport control gates at airports are examples, where a series of computer programs link scans of a traveller’s face, boarding pass, and passport to various airline and government databases. The software then ‘decides’ whether a traveller can leave or enter a country by controlling the gate. These gates are Cobots, as their decisions can be operationally over-ridden by human supervisors. Cobots are also often deployed in dangerous situations to improve HSE (Health, Safety and Environment). For example, they have been installed on drilling rigs so that the derrick floor can be unmanned and controlled remotely, significantly improving safety for drillers. Bomb disposal robots are obviously an effective way to deploy Cobots to reduce potential harm to humans. None of us worry (too much!) when an ‘auto-pilot’ program lands our plane safely in bad weather.
Human hands are used to operate a mouse, keyboard or track pads, and at present these are the main ways most of us interface with our computers and AI. There is ongoing research into wearable devices to improve this interface between man and machine, for example by using our eye movements and blinking to control the cursor on a screen. ‘Wearable tech’ which augments our human abilities is already in use, such as night vision goggles and Apple Watches. Neurosurgeons can already connect prosthetic implants to nerve tissue, allowing the patient some direct control of their artificial hands and limbs. And neuroscientists are looking into the direct connection of computers to human brains via neural implants to provide, possibly, the ability to control computers through our thoughts alone.
"Implementing the 'Asimov Robot Rules' is essential, before humanity creates autonomous killing machines..."
The development of Autonomous Robots (AR) is probably the most contentious area of AI research. So far, no one has been idiotic enough to connect their ring security camera’s motion sensors to a gun, for the obvious reason that doing this could kill the postman or the neighbour's pet! However, the US Defense Advanced Research Projects Agency (DARPA) is one of the largest funders of Autonomous Robots research, much of it with the chilling objective of ‘efficient killing’. The military drones in use at present in Ukraine and the Middle East have human operators, so are not yet autonomous. The difficulties car developers have encountered in deploying ‘self driving’ cars in the real world demonstrate the dangers of fully automated systems which are deployed without human override. Reality can often be a lot more complicated than expected. Implementing the 'Asimov Robot Rules' or laws (suggested by the science fiction writer Isaac Asimov in 1942) may be an essential first step before humanity considers creating automated autonomous killing machines. They should be fairly obvious and unambiguous:
A robot may not injure a human being or, through inaction, allow a human being to come to harm;
A robot must obey the orders given it by human beings except where such orders would conflict with the First Law;
A robot must protect its own existence as long as such protection does not conflict with the First or Second Law; and
A robot may not harm humanity, or, by inaction, allow humanity to come to harm.
Transhumanism
Almost all of us use 'nootropics' (based on the Greek word 'noos' meaning mind). These are any natural or synthetic substances which may have a positive impact on mental skills, such as the caffeine in a cup of coffee. The US Airforce has admitted in court to the routine use of medical grade amphetamines and other cognitive enhancers to improve the performance of their fighter pilots during tactical operations in the Middle East. The prescription nootropic modafinil is claimed to improve performance in exams and in the workplace where enhanced cognitive function is beneficial. It is likely that ongoing research into brain-related medical conditions, such as dementia and motor neurone disease, will also lead to the development of more powerful nootropics, leading directly to the ability to boost significantly our natural mental and cognitive abilities through Biochemical Enhancement. Those experimenting with these emerging psychoactive chemicals very much see themselves as ‘psychonauts’ exploring the limits of our existing cognitive abilities.
"Some predict that humanity is potentially reaching a series of ‘singularities’ where AI becomes more capable than humans..."
Advances in nanotechnology are also proceeding at an accelerated pace with the development of artificial eyes and electronic neurones. These have already been implanted into volunteers to investigate the potential of linking our brains directly into computers and, also, the possibility of downloading software directly into our brains. Mark O'Connell’s book To Be a Machine: Adventures Among Cyborgs, Utopians, Hackers, and the Futurists Solving the Modest Problem of Death (2017) is an entertaining exploration of transhumanism. Adherents seek to use technology and biochemical advances, to fundamentally change the human condition through the transformation of our ‘wetware' (brains) and our ‘meat’ (bodies) to create ‘super humans’. At the outer reaches of this movement there are individuals who want to be the first person to have their mind downloaded completely into a computer. Some adherents, such as the Head of Google Technologies Ray Kurzweil, or the billionaire entrepreneur Peter Thiel, predict that humanity is potentially reaching a series of ‘singularities’ where either humans succeed in vanquishing death through biochemical engineering, or AI becomes more capable than humans, or both.
Consciousness
The nature of consciousness and the nature of intelligence are often controversially referred to as the ‘hard problems of science’ as neither has been clearly defined. The neuroscientist and professor of psychiatry Giulio Tononi in his book Phi: a Voyage from the Brain to the Soul (2012) has developed an ‘integrated information theory of consciousness’ based on his work with patients with degenerative mental conditions. He argues that human consciousness is an emergent property of our brains, which consist of over 85 billion neurons, with over 100 trillion synapses, which are connected into a complex hierarchy of information in a ‘highly structured, differentiated and integrated’ manner. He is sceptical that similar types of artificial consciousness will be possible, given the very different and less integrated internal architecture of microchip technology. Tononi does not preclude the emergence of multiple artificial consciousnesses, but argues that they will be very different from ours, given that the hardware architectures are fundamentally very different. Given that we cannot define consciousness, there is the possibility that we may not even recognise an artificial consciousness if and when it emerges.
"If multiple artificial consciousnesses do emerge, the effect on those who still cling to the ‘exceptionalism’ of humans will be severe..."
What is clear is that current AI developments are the start of an exciting new exploration into the nature of consciousness. In some ways, robots and virtual AI bots can be considered to be like computer simulations of ourselves, that can be used to study and experiment with different types of consciousness and intelligence. What are its limits? How does it emerge? Are there different types of consciousness, as argued by Thomas Malone, founding director of the MIT Centre for Collective Intelligence, in his book Superminds: How Hyperconnectivity is Changing the Way We Solve Problems (2018)? If so, what are they and what controls the variations? Is Tononi right that our consciousness will not be easily recreated using microchips as our brains are the result of billions of years of evolution? If multiple artificial consciousnesses do emerge, the effect on those who still cling to the ‘exceptionalism’ of humans will be severe, and it will certainly have a profound impact on our understanding of the nature of reality and our place in the universe.
As a scientist and a humanist, I am really excited by the prospect of what we will discover as the current research and developments into AI proceed across all fronts. Of course, this needs to take into account the obvious dangers and ethical considerations. But in some ways, having these debates, about how best to engineer appropriate safeguards and how to design emergency ‘abort’ capabilities, must be a fundamental part of the ongoing research and exploration of consciousness itself. I suspect that things will turn out to be very much harder and more complex than at present envisaged. If we get it right, I think it is very likely that any superintelligence that does emerge will be wiser, more benevolent, and kinder than most of us organic humans! Finally, I am very much on the side of the transhuman advocates and ‘psychonauts’ as they bravely explore the limits to human cognition, even though sometimes they do sound a bit like ‘psychonuts’.
Links to Some Sources
The Future is now: Unlocking the Promise of AI in industrials McKinsey & Company 2022 https://www.mckinsey.com/industries/automotive-and-assembly/our-insights/the-future-is-now-unlocking-the-promise-of-ai-in-industrials
Innovative Exuberance by Tobias Huber of Newco Shift online tech magazine https://shift.newco.co/2017/07/03/innovative-exuberance/
Economist articles in The World Ahead 2023 by Kim Povlsen from Universal Robots says we should welcome robots, not fear them, from https://www.economist.com/the-world-ahead/2022/11/18/kim-povlsen-says-we-should-welcome-robots-not-fear-them
The Future of Jobs The World Economic Forum 2020 https://www.weforum.org/reports/the-future-of-jobs-report-2020/
The Sunday Times reviews the work of the new Institute of Ethics at Oxford University: How Do We Stop the Robot Takeover? https://www.thetimes.co.uk/article/how-do-we-stop-the-robot-takeover-oxford-dons-have-a-plan-c39dzkkfw and https://www.oxford-aiethics.ox.ac.uk/
The UK Met Office Climate Prediction Model HadCM3 https://www.metoffice.gov.uk/research/approach/modelling-systems/unified-model/climate-models/hadcm3
Asimov's Court Robot Laws in Britannica https://www.britannica.com/topic/Three-Laws-of-Robotics
Mark O’Connell To Be a Machine - Adventures Among Cyborgs, Utopians, Hackers, and the Futurists Solving the Modest Problem of Death (2017) reviewed in The Guardian by Paul Laity https://www.theguardian.com/books/2017/mar/23/to-be-a-machine-by-mark-oconnell-review
The Psychonauts’ World of Cognitive Enhancers Napoletano et al 2020 https://www.frontiersin.org/articles/10.3389/fpsyt.2020.546796/full
Giulio Tononi PHI - A Voyage from the Brain to the Soul (2012) also reviewed in a BBC Future article by David Robson in 2019 - https://www.bbc.com/future/article/20190326-are-we-close-to-solving-the-puzzle-of-consciousness
Superminds: How Hyperconnecivity is Changing the Way We Solve Problems (2018) by Thomas Malone, Director of the MIT Center for Collective Intelligence, reviewed in Kirkus https://www.kirkusreviews.com/book-reviews/thomas-w-malone/superminds/ and the MIT Centre for Collective Intelligence https://cci.mit.edu/
How to Use ChatGPT in Effective Ways for your Career in Emeritus website https://emeritus.org/blog/ai-and-ml-how-to-use-chatgpt/
10 Best Examples of How AI is Already used in Everyday Life by Bernard Marr in Forbes (2019) https://www.forbes.com/sites/bernardmarr/2019/12/16/the-10-best-examples-of-how-ai-is-already-used-in-our-everyday-life/?sh=5ba1e2511171Forbes
Wikipedia on Consciousness, Intelligence and Machine Learning https://en.wikipedia.org/wiki/Consciousness, https://en.wikipedia.org/wiki/Intelligence, https://en.wikipedia.org/wiki/Machine_learning
Further references are available on request.
Comentários