Artificial Intelligence - Man vs. Machine (Part II)
Image: Soul Machines Auckland Face Simulator via fxguide: Avatars and Agents:BabyX
What will it mean to humankind if machines are better than us in intellect, technical skills, efficiency, and even possibly art? Do we now become unnecessary or obsolete? Well, if we are even asking that question then something essential is being overlooked in the understanding of ourselves. Our value does not rest in how intelligent, attractive, strong, or skilled we are, or how much we can improve our company's bottom line; although many of us may have been "programmed" to think this way. These qualities are valuable nonetheless, but our value as human beings is not confined to these realms. One may reasonably argue that failure to acknowledge our true value is what underlies many of humanity's problems throughout history.
Human beings and other living beings have intrinsic value by being who we are; a unique representation of life itself which has developed over billions of years summarily, and shaped over a lifetime individually. This value is immeasurable and beyond comparison. This does not mean that we shouldn't strive to reach our highest potential, or that we shouldn't celebrate, honor and reward society's great thinkers, scientists, engineers, philanthropists, artists, healers and athletes - those who contribute the most. It means we must acknowledge as fact that it is not only our skills and talents that make us who we are. It is also our history, our ancestry and heritage, our relationships with others, our unique life experiences, challenges, choices, triumphs, and even our vulnerabilities, weaknesses and flaws. Therefore, equality is inherent amongst humans because it is not the things which can be measured that gives each of us our true value. Furthermore, creating advanced intelligent machines which are able to accurately imitate our humanity or surpass our human capabilities, does not lessen us as beings, even if the machine outpaces us in the area we view as our greatest achievement.
Artificial intelligence is being developed in a multitude of forms in laboratories around the world today. We must ask, how should we relate to the emergence of this "new entity" which will eventually outperform us? We should all agree that the ultimate role of AI is only as an extension of humankind. This approach should be taken regardless of why the machine was created, by whom it was created, and regardless of how advanced it is. AI should be programmed to enhance and amplify humanity, not to replace it, and not solely for profit; if we desire to fare well as a species in the near future. We would need to think in terms of our collective benefit, as we are the ones who give meaning and value to the creations, and we remain the gold standard. There should be no concept of human vs. machine, only human and his/her creations.
With this clear approach, we can realize the tremendous benefits which AI can bring to humanity. Auckland's SoulMachines have created incredibly lifelike and consistent digital models of ourselves which bring important insights into how our brain functions and how we learn. Adobe's Sensei AI can draw from vast amounts of data to assist us in creating and editing beautiful works of art and content. Corporations now have the ability to use AI for its super-human capabilities; to intelligently process vast amounts of information and help solve complex problems. We have amplified our analytical power with which we can address global problems of humanity, including resource allocations, natural disasters, climate change, drought and famine.
We need to fully understand the super-human intellectual ability of AI for what it is and understand what it is useful for. Using machine learning, AI can process vast amounts of information and learn via pattern identification, to achieve in a relatively short time, insights which would have otherwise taken years of dedicated human effort. The insights can be used to predict a particular outcome, or determine the best course of action to achieve a particular outcome, with an extremely high degree of accuracy. Google DeepMind's AlphaGoZero taught itself to beat previous versions of itself and human players in the ancient Chinese game Go, without receiving human input, in the span of 3 days. The most skilled human player in this game took years to develop. The power of machine learning is via the speed of the process, and the generous size of the input stream. The algorithms can perform or analyze millions of events within a short period of time in order to "learn" the best way to achieve an outcome or to predict certain outcomes. These insights can be crucial in the field of medicine and already are being applied to the detection of breast cancer at MIT's CSAIL.
Although extremely powerful, there are obvious limitations to this intelligence. If we are dealing with "messy and uncertain" real world problems which don't have a clearly defined set of rules or outcomes, or have seemingly conflicting values, how much can we expect to rely on data analysis alone to find the answers we need? This issue was mentioned in a recent SingularityHub article. To give a crude example, suppose after analyzing all the data we feed into an AI algorithm, or using data it obtains on its own, it learns that the most effective way to reverse climate change is the complete elimination of humanity from earth? Although we may not have the intellectual power to confidently and accurately derive that conclusion, because of our values, we would deem this to be an unacceptable solution, and most likely our desired outcome would change. Perhaps we would decide that instead of reversing climate change, slowing the rate of change is a more useful and achievable outcome, because the cost of achieving the original goal is beyond what we are willing to pay. The algorithm may be super-human at predicting outcomes and solving problems because of its speed and statistical superiority, but these insights only make sense within the constraints and values we give it. If the input it uses to obtain those values comes from observing all of humanity this is obviously much more useful to us, but it does not remove the fact that mathematical and statistical models of reality do not equal reality, and meaningfulness and value is currently only defined in the context of intelligent living beings such as humans. The difference between observation based objective knowledge and consciousness based value driven knowledge is a key factor in this distinction.
We put ourselves, perhaps in the best position as a society or organization, if we can use intelligent machines for what they do best, and allow humans to excel at what we do best in our world. AI should be employed to reduce the time it takes for us to arrive at conclusions from big data, make powerfully useful estimations and predictions, and figure out how to do complex, repetitive or unpleasant tasks which require problem solving, but don't necessarily engage the full spectrum of human capabilities. Ideally human beings, and not AI, should be engaged in relating to other human beings, forming natural relationships, making value judgements, finding meaning in all of our creations and desired outcomes, having a vision for the future goals of our species, and expressing our creativity based on the full spectrum of our humanity. So the question arises, if there is a proper place for machines in relation to man, should we place limitations on AI? Click here to continue reading Part III...
Feel free to check out our other blog posts and leave a comment on the website.