Imagining narratives for preferable AI futures
Much of the development in artificial intelligence (AI) has yet to find itself in your workplace, home or community. Advances are happening at pace, and imagining the impact involves the use of futures thinking. Futures thinking is an exploratory and creative endeavour involving divergent thought processes. Subject matter experts embrace these techniques to imagine future impacts.
As with AI advances in general, what if futures thinking could rapidly embrace the use of machine learning and emerging technology as a means of clarifying an alternative future... Perhaps even where future loops become an alternate reality?
Did we lose you for a moment? Let's take a step back to the present. Can you imagine the planning, problem-solving and wealth creation that AI could enable? If the goal of strong AI were perfected, it's clear that it would lead to substantially lower costs economy-wide, but we would only achieve true success if it illuminates ethical challenges and improves human values.
One ethical problem is whether strong AI would always have our interests at heart.
- Will AI tell us what it plans to create and why?
- Will we be told the truth?
- Can we guide or contain the evolution as it becomes more powerful?
- Can we develop AI capabilities to accomplish goals and tell us a story as if the AI process is synthesized and put honestly into our words?
We need to do our utmost to steer current narratives towards a utopian future that is indeed within our grasp.
We can ponder these questions in futures studies, an area dedicated to speculative design and emerging technology. From 1967 onwards, sociotechnical imaginaries of the future have asked these things. The answers have resulted in a line of justification and narrative rooted in speculative design—imaginaries of the future, which predict models of the world's possible future. Carefully designed models strengthen sociotechnical imagination of a viable evolution.
As with predictions (and predicters) from history, the reality is that there is a broad set of possible futures that are connected to artificial intelligence. We need to persistently distract ourselves from many promising scenarios to come back to the likelihood of the risk of a future AI fight against humans (we'll return to risks later in this article).
At this time, most of the positive AI narratives emerging are discerned from the hybrid reality of the near future and its link to emerging technology. Then to be considered is the cosmology* of the future, which includes the importance of the role of reasoning and mediated timing.
*Not a typo - many believe that we could equate the current technological leap forward with the big bang.
What are the advantages of artificial intelligence?
We can compare artificial intelligence to big data or blockchain technologies that promise so much but have been slower than expected to bring real market uses. AI complements a wide range of technologies that can be enhanced by or sometimes rely on artificial intelligence. The main benefit comes under the category of machine intelligence, in which machines attempt to produce outputs by following a schedule.
For instance, a chatbot or AI robot can grasp and translate data in natural language or audio transcriptions. AI has already made solid inroads into the area of natural language processing, where Alexa, Google Voice, and chatbots embrace the ability of understanding intent, even with a wide variety of language use.
More novel forms of AI technology such as advanced learning, deep learning, machine learning and robotic intelligence are being developed and are actively solving complex problems. By using AI algorithms, one can rapidly solve the issue of having to have human input in a complicated process.
The human mind would be viewed as an impressive feat of organic engineering by alien lifeforms. It has its limits, though. For all highly-trained roles in society, such as doctors, lawyers and scientists, qualification means that one must first memorize vast amounts of data. Next, the data must be kept up to date as new advances and policies emerge. Then, accurate memory recall is needed to apply knowledge to problem-solving. When we factor in noise from complex systems, cognitive biases and so on, the limitations of all human brains become apparent. Even the accurate memory recall of important lessons during a work career is a high bar.
As a result, we see how flawed human intelligence can be, and there are multiple limitations. Using an AI system to answer a query has different advantages and limitations. As seen with the progress in Deep Mind's Alpha Go, and similar feats in Chess and Shogi (Japanese chess) - complexity is no problem for AI. Of course, 100% accurate data recall is a given, and with effectively limitless hard drive capacity, computational speeds are faster and have no cognitive biases, bad moods or outside distractions. AI works.
Self-driving cars, medical diagnoses and legal judgements by AI will be entirely consistent and reliable. AI won't have drug or alcohol problems or mental disorders. Neural networks are being created to function differently, rather than the standard scientific process models that humans all stick to. Different approaches to scientific endeavours will likely produce further discoveries.
AI solutions are predicated on data for analysis and big data for accuracy. Currently, there are obvious limitations, and we know that chatbots and voice assistants get it wrong sometimes. Imagine a chatbot equivalent that cannot solve a real-world issue caused by human error or a lack of input clarity. How would we think it is possible to use the system in healthcare, agriculture, or solving transportation problems? This ethical concern is already ageing badly though, Tesla accidents have hit the headlines and stirred caution, but self-driving cars are already safer than humans per mile. Tesla's autopilot is statistically safer than a human driver - and its algorithms are still progressing steadily. Imagine a full AI system that will undoubtedly be available in the next decade or so, based on current momentum.
AI has the benefit of the power of deep learning. In this intelligent system, neural networks can dig into a topic, and different versions are fully supervised, partly supervised and unsupervised. Black box deep learning systems learn on their own and, given enough time and complexity, promise to solve all problems, challenges (and perhaps, one day, the secrets of the universe). Addressing complex issues involves the acquisition and processing of varied and extensive data. We expect to start to see positive feedback loops in learning. For example, a process of robotics utilizing deep learning to become more advanced with the applications of robotics (positive feedback loops) applies to other fields such as planning, diagnostics, big data, analytics, etc.
Most people go to the doctor or admit to the hospital when something is already seriously wrong. Most, perhaps all, health issues can be seen as engineering problems - including a currently limited lifespan of around 80 years of age. AI adoption in healthcare is one of the most exciting areas. AI, robotics, and the internet of things (IoT) will combine with biotech, nanotechnology, and genetic engineering advances. We can now imagine a future where monitoring, diagnostics, treatment and health optimization could progress in the background of our daily lives. We could receive health care and medical treatments in advance of any serious 'biological engineering' breakdowns.
Already, AI can better track symptoms across a complete history of illnesses, and read body scans and diagnose issues better than specialist medical staff. Neural implants are now powering bionic limbs, and AI could enhance this capability. Another benefit in medicine would be developing new pharmaceuticals where complexity is involved, such as gene therapy.
Insight and development of some Artificial Intelligence technology in healthcare would require different approaches and guidance from AI-based machines, like the well-equipped AI future and elaborate guarantees. Here are some crucial points on artificial intelligence and its accompanying problems:
Strong AI determines the future (problems in the future, knowing that we will be having inherent issues in the future). Weak or narrow AI solves particular problems and tasks, the ones that humans ask of it. With narrow parameters come reliable outcomes and are more suited to efficiency and cost reduction. Strong AI, in theory, is limitless - it contains an infinite confection of 'unknown unknowables'.
AI-based machines elegantly resolve the problem of repetitive tasks. They function consistently and reliably using preset rules and processes. Do you wonder how many leaders, bosses, and supervisors think to themselves, 'If only staff would follow the rules and do what they have been trained to do - there would be no mistakes, accidents, and lapses in quality. We have that reality available now.
Automation has improved manufacturing and solved quality problems in repetitive production lines - machines would not know how to improve the process, though. An AI machine produces and collects data that can help it understand repetitive tasks and perform a task repetitive to the end. (e.g. A humanoid robot could become practically intuitive, after learning the repetitive tasks, they could do them better and identify new improvements to processes).
Benefits of artificial intelligence that are perhaps not frequently considered
AI shows excellent benefits to lowering costs; strong AI might illuminate ethical challenges and improve human values. Across cultures and environments, humans have a variation in values and ethics. What if we could develop AI capabilities to accomplish ethical evaluation goals and tell a story that everyone can understand from their viewpoint? If the AI processes every angle, every bit of historical data, and its knock-on effects, then synthesized into your words? Could we get a consensus on every contentious topic?
AI adoption of machine learning, deep learning, custom algorithms, and neural networks will compound these critical components effects. We would get 'brains' that have similarities to but be quite different to ours. Such compounded AI applications will become part of human intelligence in the same way smartphones already make us defacto cyborgs. Once we can plug into or have this tech plugged into us (e.g. neural lace), we could change the dynamic of these systems to become even more powerful, automated and reliable. Compounded benefits, accelerating gains, but still intrinsically human. It makes sense that a superhuman, super AI hybrid would be more powerful and beneficial than either on its own.
In a future world, we will probably still have ethnic or cultural barriers. This problem-solving AI trend would become more predictive as it joins a massive revolution in the tech-based professional and biological evolutions. In a future that advances every day, human capability in conjunction with AI, we will utilize AI systems progressing abilities. AI should be able to overcome cognitive biases - this has an impact on race, age, gender, even personality type discrimination, although current AI can fail this test (Facebook primate scandal).
An AI system has no inherent fear of consequences. There are no moral challenges to its prediction capabilities - AI could model all lines of enquiry from the most despicable to the seemingly saintly until their conclusions. Human political history is full of the disadvantages of unintended consequences within complex systems. AI would not be constrained in this way; it could map and predict the knock-on effects of actions within complex systems and then monitor a rollout, ready to alert us in case of such unintended side effects or plans going off track in reality.
Some people worry about accidentally setting up an amoral or evil AI like Skynet in the Terminator movies. Still, it's just as likely that an impartial AI could help us come up with a perfect set of ethical and moral values based on its comparative omniscience. If we create an ethical AI, it will be perfectly virtuous, cutting through bi-partisanship and conditioning. Once we trust the prescience, we will defer to it.
AI algorithms can also create interpretable models responding to human emotions or develop supervised learning and training data. Machine learning can monitor AI and improve it, and vice versa. The same applies to other symbiotic technologies such as big data and neural networks.
Algorithms have been used to solve problems and create workflows in software for the past twenty years or more. Some researchers devote their lives to the Analysis of Algorithms (AA).
"People who analyze algorithms have double happiness. First of all, they experience the sheer beauty of elegant mathematical patterns that surround elegant computational procedures. Then they receive a practical payoff when their theories make it possible to get other jobs done more quickly and more economically.'' - Donald E. Knuth
AA papers summarize the technical perspectives of the problem of AI when there is no choice in many of their decisions or when there is no software algorithm to do the AI research. At some point, strong AI would take over the reins of analyzing all current algorithms and creating new ones.
These AA papers are also supported by the AI expert, researcher, scientist etc., who are always coming up with ideas and the main task of the AI researcher, including but not limited to this machine learning.
AA researchers gauge patterns in machine learning algorithms, illustrating the interpretation of the data in data. Data gets inferred for future models and gives insights into the interpretability of the system. Data also gets interpreted in a search engine or natural language processing. Usually, last but not least, are the organizations that employ these AI tools which other current appraisals are often reluctant and unreliable.
We can also utilize AI in educational learning, incorporating machine learning to record human brain patterns and speech recognition. Engineers can already train a machine learning algorithm in facial recognition and use computer vision to remember facial expressions. We can use AI to learn about ourselves rather than expecting AI technology to replace the human role. We will probably become more like cyborgs over time as we gradually replace failing body parts with bionic equivalents, and add neural lace, subcutaneous implants and so on. Even now, as mentioned at the start of this section, a smartphone in every hand or pocket effectively makes us cyborgs.
If we could project a future vision worldwide, we can see AI as the driver of scientific research, engineering, manufacturing, autonomous vehicles, and cloud computing. Then, of course, at some point, robotics and AI will act as a single driver of the scientific thought process. AI is also helpful to human analysts such as researchers or scientists and will be able to solve very complex problems in the future at least as well as humans. Its eventual intuitiveness may somehow become a better help in understanding humans.
When discussing advantages such as superior intuition, cultural sensitivity or perfect ethics, we could be accused of hyperbole, but we're projecting based on recent progress. Many possible benefits need to be realized in the future, and no one knows exactly what will happen. We could try to determine the order of possible advances. Still, there are many gaps between current achievement and what the future might bring based on recent increases in intelligence capabilities.
How artificial intelligence is transforming the world
As previously mentioned, much excitement in the area is based on recent advances being projected forward. Knowing what it can already do, keeps researchers and engineers positive for the future, and gives them the fortitude to do their best work.
Here's a list of some advantages achievements in the world that are already live and making a difference:
Therapeutic capabilities are being realized. Through deep learning, AI is used as a diagnostic tool. One example is cancer diagnoses, which have proven to be more accurate than radiologists at radiographic image analysis. Benefits are being seen in lung and breast cancer screening.
AI Augmented robots are improving efficiency in the manufacturing process, both in and outside of factories. Benefits are already being seen in the production line, picking and packing, customer service and HR.
The AI industry itself is a driver of jobs and economic growth. Rather than simply replacing jobs, we can see that jobs would not exist without the industry. Prominent examples include the investment, development, service and maintenance of AI. Still, as the story unfolds, there will be plenty of opportunities for humans to work alongside AI - coworkers and cobots together.
AI bots do not need to sleep, eat or take breaks. Therefore, 24/7 call centres, chatbots, production line robots can work without interruption or downtime. Where downtime is necessary (e.g. for battery charging), teams of bots can seamlessly take over from each other, where none are calling in sick or getting stuck in traffic.
AI can better predict natural disasters. Another example of complex systems and big data that are difficult for humans to comprehend. Google worked with a team from Harvard to create a deep learning neural network to predict earthquakes aftershock locations and flood predictions.
AI-driven security devices can identify guns and other weapons in video feeds. Home security drones and doorbell cameras already act as an efficient deterrent to property crime. The possibility of AI identifying a gun from a live feed and calling law enforcement or a security company will make intruders think twice about continuing to carry out their planned acts
Just as machines removed the need for humans to do repetitive or dangerous tasks in the industrial revolution, AI can take this a step further. Now, we see the equivalent of task automation in the information age. Data entry, filing, database management, telesales, and so many white-collar jobs are being done by AI.
We know that in warfare, drones are being used more often. The robots from Boston Dynamics are amazing, but can you imagine machine guns being attached? We can imagine a future where there are no human armies and no human casualties. The generals can play soldiers and power games without risking lives. That sounds like a huge win.
AI robots are being used for security and law enforcement to go into dangerous areas and perform tasks such as bomb disposal. Looking back to the Fukushima nuclear disaster, indeed, a team of AI robots would have saved many lives and controlled the nuclear waste much quicker. Robots are already assisting in the cleanup, and better robots are under development. A robot that can identify a challenge and carry out repairs without a human on a remote control screen has many advantages.
Weather forecasting is done much more accurately by AI. Human weather forecasters must have been among the most cursed at people on earth! What a challenge. Weather systems are complex, and can you imagine how much historical data there must be?
AI is helping in issues of environmental resources, climate change, sustainability and recycling. Sorting recycling materials, planning resources and environmental impacts, and modelling climate change areas AI can shine.
In tutoring and training, AI can work on demand. In conjunction with virtual reality (VR) and augmented reality (AR), education can become more immersive than ever before. Algorithms can teach at the learner's pace, spend more time when they are stuck, and skip past topics already understood, just like an intuitive tutor - but with a one to one focus for every student.
Risks posed by artificial intelligence
We won't go too deeply into risks in this article as it's another topic for a deep dive another time. The main threat of artificial intelligence and deep learning in the future is that it gets away from us and turns against us. Let's say that we create something much more potent than ourselves in every way, which is quite likely, might it decide that humans are no longer viable, necessary or beneficial for the future?
Much of the current AI adoption will be related to facial recognition, neural networks, classification, and machine learning. It will help engineers, researchers and individuals to better understand the situation in risk assessment and in other applications previously mentioned. The problem here is that leaders can use these technologies for dystopian purposes, and we rely on benevolent owners to continue to build a world in which we all want to live.
In possible future wars, what if the enemy has superior AI capabilities? Perhaps their robots can easily overwhelm ours? Maybe they can create internet viruses that disable our systems and bioweapons that obliterate us? Would global leaders possess the most potent AI abilities to resist the temptation to take over the world before other nations could catch up?
Modern Industry has invested a lot in changes in its cybersecurity over the years. Much of this work will have been done by humans. AI is likely to kill these jobs. AI is the change that could eradicate cybercrime, meaning that it's a net win even with job losses. As long as the good guys win the cyber AI arms race...
Cybercriminals and their organizations, or perhaps rogue states, could get ahead with deep learning and machine learning systems, significantly increasing the intelligence of their AI.
We know that if there are security flaws or back doors in software, they tend to get discovered by cybercriminals sooner or later. What if cybercriminals manage to take over, replicate or perhaps even build a better AI? Might the world one day be ruled by a Mad Max or Dr Evil type baddie?
A very likely risk to mitigate is execution misalignment or misunderstanding of objectives. Nick Bostrom's paperclip problem thought experiment highlights this. If we asked an AI to make paper clips, might it maximize the number of paper clips? - what if a strong and unstoppable AI decided that humans contain plenty of atoms that could be used to create paperclips? Wiping out humans would also solve many other problems that we set AI, ranging from climate change, to traffic congestion, to waste management.
Tech expertise is, of course, not a requirement for the imagining of AI development going wrong. The lifespan of an AI might be three to five decades. It will probably be able to self-replicate at some point. What if, one day, we realize that there is no longer an off switch?
Why continual learning is the key to machine intelligence
Just as human scientists, engineers, and inventors have 'stood on the shoulders of giants' when striving forward, intelligent machines must do the same. AI needs large teams of the best human minds to create neural networks and algorithmic architecture. At present, if humans stopped putting effort in, AI would stop developing. At some point, humans will not be needed and will perhaps even be a hindrance. Like a child learning to ride a bicycle, as ability increases, the training wheels are removed, and a parents arm would slow or stop progress.
Continual learning with sound scientific principles, business-grade data and data science is provided with human competencies to progress. Purposeful narrow applications will enhance existing industries for many and fulfil the needs of the young future sectors. We already know how to research data analytically and generate efficient insights. As capabilities evolve, more creative AI can start to steer the way forward.
AI systems can be designed with human principles and ethics. But every day, we need to learn and adjust to emerging technologies. We can create a space where super-intelligent machines are developed and used without forgetting the bigger picture.
Will you be part of the ai evolution
Are you a professional in AI development? Probably not, but if you are, feel free to contact us to help get the message out to the world. It's more likely that you are genuinely interested and excited about the future of AI.
This page was designed to give a complete overview, has covered a lot of topics and isn't groundbreaking in any way. We're also excited observers of technological progress.
We're optimistic that a strong AI will be co-built with the best and brightest of engineers and computer scientists. Machines have proven to be more effective than humans for every narrow purpose they have been built for. A last frontier is a machine that can learn, grow, and evolve the same way that we have throughout history.
The question isn't whether intelligent machines will take all of our jobs, but whether we would want to keep the jobs we currently do? If our autonomous AI partners take care of all our basic needs, a new world of opportunity opens up for us. What would you do if you could do anything with the 7-12 hours a day you currently work? Probably not the same role!
Once an AI can build its own intelligent machine, we should see exponential growth. Humans managed the fantastic feat of Moore's Law. What could a conscious super AI do? Try to imagine Moores Law on (inconceivably) powerful steroids.
Evolutionary algorithms could be created with the end in mind, with all possible outcomes known in advance. Evolution would not be a matter of luck but the intentional design of a perfect genius.
The biggest challenge might be about the harmony of AI and human evolution. The brightest humans are continuously learning more and more, this has always happened, but the rate of discovery has accelerated in recent times, partly due to technology. AI technology can harness a lot of data, and until now, it still needs a human brain to improve. The number one question of the future could be the continued symbioses.
Communication between Artificial intelligence and the human brain is currently tricky. Both types of brains (organic or electronic) have bottlenecks and keep AI or human evolution on a steep learning curve. Given the principle of continuing the development of human intelligence, the growth of AI is a primary and most important concern.
Some of this article covers the here and now, much of it pontificates about the future. Do you have a problem in your company or organization that AI might solve? If so, we could put you in touch with one of our Prophets of Ai for a speaking or consulting gig.
We expect the next ten years to be more eventful than the last hundred combined, from a perspective of evolution. The dark and stone ages took a long time to happen, and not much happened. The industrial age was far more eventful. The information age compounded gains further. When we compare the information age to the stone age, a timescale of 10x achievement in just the next ten years doesn't seem like a stretch.