Source text is here
Judges' commentaries (long!) will be uploaded soon.
Winner: E21 Scott Wilson
Runner-up:E07 Robin Griffin
Finalists:
E26 Ayami Kan
E27 Ciaran Doyle
E45 Catherine Xu
E54 Mizuki Kitamoto
*****************
E26 Ayami Kan
In the case of insects, their action appears to be based on a set of behavioral rules they follow in order to survive. These rules dictate an insect’s behavior in response to information from the external environment. Both these behavioral sequences and rules are innate to insects. This suggests that insects can be regarded as autonomous creatures that function in a similar manner to automatic systems; their response is determined according to behavioral rules, informed by their sensory input, such as tactile sense. Yet, even humans are the same of insects in that they share the ultimate purpose of ‘survival’. Generally, we are not conscious of our purpose for ‘survival’ in everyday life. Rather, we are driven to act because of hunger, to finish homework because of our aversion to scoldings, and to make an effort to move because of our desire to meet people. Ultimately though, all these behaviors are for the purpose of ‘survival’. In this sense, insects and humans can be considered analogous systems, continuously selecting what is considered the appropriate behavior for a given situation in order to survive. One difference between humans and insects is that humans possess a wider variety of behaviors that can be activated in response to incoming information from the external environment; the link between this information and the action activated is also extremely complicated for humans. As a system becomes more adept at responding to the external world, it becomes better at adapting to the surrounding environment, thus further increasing the system’s likelihood for survival. Insects can only react according to their innate behavioral rules. In contrast, by making use of what we learned from experience and the knowledge we acquired from others, we are able to both dynamically change the way we respond and create variation in our behaviors. A system can be considered autonomous if it has the ability to test various behaviors and learn new ones in order to achieve a certain goal. Thus, although the Roomba vacuum cleaner and dog robot AIBO appear to act on their own volition, they are in fact automatic robots. Broadly speaking, insects would also fall under this category, whereas animals that can acquire new behavior and knowledge would be considered autonomous systems.
This next generation of AI, namely autonomous AI, plays an incredibly vital role in realizing AI and human coexistence, but is distinct from machine learning. Although lagging in terms of research and development (R&D) on machine learning, especially in comparison to the United States and China, Japan is currently making strides in R&D on autonomous and versatile AIs. Therefore, it is possible that Japan will be a competitive figure in AI R&D in the future.
R&D on autonomous AIs are being actively pursued not only in Japan, but globally as well. There are now concerns for how this technology will lead to Lethal Autonomous Weapon Systems (LAWS), the same way machine learning has given rise to the problem of deepfakes. As the name implies, LAWS are weapons guided by autonomous AI that, based on its own judgement, has the ability to release its own trigger. Unfortunately, several countries have already carried out the development of these weapons, with some even ready to be deployed. Although the United Nations is attempting to create measures that will ban nations from deploying LAWS, preventing the use of deadly autonomous weapons (and war, which should be avoided at all costs in the first place) is truly a situation putting human judgement and intelligence to the test.
E27 Ciaran Doyle
If we examine insects, we shall see that in order to achieve their goal of living (self-preservation), their behavior is apparently entirely determined by a system of rules which selects the course of action to be taken in response to information received from the outside world. This is done by drawing on a previously compiled behavioral system, as the rules are present from birth. This results in what appears to be a highly autonomous organism which responds to sensory input (e.g., from tactile sensors) with rule-based behavior. The insect may even seem to act almost automatically. Humans have exactly the same goal as insects: to continue living. When we perform actions in our daily lives, we do not consciously think of them as directly contributing to this goal. We eat because we are hungry, and we do our homework to avoid getting in trouble. We act in accordance with the people around us. However, if we look to the origin of these motivations, we find that it ultimately comes back to our goal of preservation. In order to keep living, we select what we believe to be the most appropriate response to the environment around us. Just like insects, our “system” continuously chooses the most appropriate behavior. Only two factors separate our two species. First, there is the overwhelming number of different types of behavior that we humans have available to us in response to external stimuli. Second, the specific behavior that is chosen in response to a specific set of input information also differs, to an incredibly complex degree. The more complex the response to the external stimulus, the better adapted to its environment the organism becomes, and the greater its chance of survival. Insects can only respond to stimuli based on innate, predetermined rules. We humans, however, can draw on learned behavior from what we have been taught as well as our own experiences. With this accumulated knowledge and experience, we can dynamically adjust our own reactions and further, dynamically increase the variation of behavior too. Systems with the capability to try various behaviors, and to learn new forms of behavior in order to achieve the goal of preservation – these are exactly what we are talking about when we refer to “autonomous systems”, are they not? Following this line of thinking, while machines such as Roombas and AIBO might appear to exhibit conscious thought, they are still merely automatic machines. Insects might also, in some cases, be included in this category. The result is that we classify only organisms capable of learning as having autonomous systems.
This autonomy is crucially important to next generation AI, that is to say, AI which can genuinely coexist alongside humans. However, this AI technology differs from machine learning. In contrast to the field of machine learning, where Japan has the lead over the major research and development work (of the US and China in particular), autonomy and versatility are areas where we must recognize that Japan is lagging behind the others. Japan must act to secure its place at the head of future AI research and development.
Of course, autonomous AI research and development is not unique to Japan. And while research and development of autonomous AI is actively continuing across the globe, it is not without consequences. Just as the issue of deepfakes has emerged in the field of machine learning, AI has its own problem with LAWS (Lethal Autonomous Weapons Systems). As the name suggests, LAWS are weapons systems with built-in AI, which means they can judge when to pull the trigger without any human input. Unfortunately, due to development in multiple countries, there are some instances of practical implementation already. The United Nations are making efforts to ban the development of LAWS, as the absurd horror of ongoing wars will only be further exacerbated if machines begin killing humans. This must be prevented; truly, we are at a point where the wisdom of humanity is being tested.3
E45 Catherine Xu
Consider insects, for instance. In order to achieve (or sustain) their goal of survival, they merely select their actions by following a set of behavioural rules (which are present from birth). These rules define which action or action sequence to activate from their behavioural repertoire in response to sensory input from their environment. Therefore, since the behaviour of these insects are controlled by rules for input data detected by sensors, such as antennae, insects can appear to be autonomous organisms that are extremely close to being automated. In contrast, consider humans, who share the same goal of survival with insects. In our usual daily lives, whenever we perform an action, we do not consciously do it for our survival. Rather, we eat when we are hungry, we do our homework because we do not like to be scolded, and we move somewhere in order to meet with someone. However, it is all, in essence, for the purpose of survival. We are systems which are constantly deciding on an appropriate action to take in response to a situation, in order to achieve our goal of survival. Both insects and humans are alike in this respect. The difference is that humans have an extraordinarily large variety of behaviours with which to respond to information from the external world, and also that the relationship of which behaviours are elicited in response to that information from the external world is extremely complex. The more cleverly we can respond to our environment, the higher our fitness level and greater our probability of survival. While insects are only capable of making responses according to genetically-encoded behavioural rules, humans can not only dynamically change how they respond, but also dynamically add variations to their behaviours by applying the experience and knowledge they have accumulated through the process of learning, whether from being taught or from personal experience. It is perhaps this ability to try out various behaviours and to acquire new ones, in order to achieve a goal, that distinguishes a system as having autonomy. Under this criterion, however conscious AI robots such as Roomba vacuums or Sony’s AIBO dogs may appear to be, they are automatic machines, and it is animals that learn, and roughly speaking, insects too, that fall under the umbrella of autonomous systems.
This autonomy is crucial for the realisation of next-generation AI, in other words, the generation of AI that will be living symbiotically with humans. However, it involves different AI technology to machine learning. Japan may be lagging far behind the US and China in terms of research and development in the field of machine learning, but at present, this is not at all the case in terms of autonomous intelligence and artificial general intelligence. There may be a possibility for Japan to secure a firm position in future AI research and development.
Of course, Japan is certainly not the only country with their attention on the research and development of autonomous AI systems. Just as machine learning has its problem of deepfake technology, autonomous AI technology has its own issue of lethal autonomous weapons systems (LAWS). In order to address this challenge, research into autonomous AI systems is being engaged with globally. As they sound, LAWS are weapons systems with autonomous AI which can independently select and engage targets and make the final decision to pull the trigger. Regrettably, the development of LAWS is occurring in several countries, and there already exists weapons which have reached the stage of deployment. Efforts are underway to ban LAWS at the United Nations, but to ensure that the senseless possibility of machines taking lives does not become a reality in already senseless wars, it is precisely the wisdom of humanity that is now being tested.
E54 Mizuki Kitamoto
For example, insects choose actions only based on inherent rules, which determine which of the innate behavioral sequences should be put into action in response to the information from the outside world, in order to achieve their goal of living, or remaining alive. Their behaviors are chosen according to action rules for different inputs obtained through sensors such as touch receptors, and it means that the autonomy level of insects is quite close to automation. And humans are the same as insects in that their goals are "to live." In everyday life, we do not choose certain actions thinking it is necessary "to live"; we eat simply because we are hungry, we do our homework because we do not want to get in trouble, and we go somewhere in order to meet someone. However, we are the same as insects in a sense that every action is for living and that both of our systems keep selecting the best action for each situation in order to achieve the goal. Differences lie where humans have a far wider variety of behaviors in response to information from outside and where the correlation between such information and selected behavior is much more complex than that of insects. The more ingenious reactions to the outside world get, the better animals can adapt to the environment, which results in higher survival probability. While insects can only react by innate action rule, we humans can learn by being taught or experiencing ourselves, and by utilizing the accumulated knowledge and experience, we can dynamically alter the way we react and also increase the variety of behaviors. I think that what we can call autonomous systems are ones with abilities to test various behaviors and acquire new behaviors in order to achieve the goal. Based on this definition, although Roomba and AIBO look as if they can really think, they are automated machines, and roughly speaking, insects are one of them. By contrast, we can say that animals which can learn have autonomous systems.
The autonomy is imperative in realizing the AI of the next generation, which can truly live in harmony with humans, and it's important to note that this AI technology is different from machine learning. Although Japan lags far behind the U.S. and China in research and development of machine learning, I don't think we are behind in developing autonomous and versatile AI technologies at this point. I think there is a possibility that Japan could secure a firm position in the field of future AI research and development.
Of course, Japan is not alone in giving attention to autonomous AI. Lethal Autonomous Weapons Systems (LAWS) is causing controversy in the field of autonomous AI, as deepfake is in that of machine learning, and it shows that autonomous AI is actively being researched and developed around the world. Weapons with autonomous AI literally means that their AI can make a final decision to pull the deadly trigger. The sad thing is that several countries have been developing such weapons, and some weapons have already reached a practical level. Although the UN is also working on banning LAWS, whether or not we can avoid absolutely unacceptable situations where "machinery kills humankind," during a war which is abominable itself, really depends on the wisdom of humanity.
Runner-up:E07 Robin Griffin
Consider an insect, for example. An insect’s goal is to live (or to keep living). To achieve this goal, the insect is equipped with a series of action patterns to respond to external data. Determining which action the insect takes is simply a matter of following a set of rules that are innate to that insect. Its behavior, then, is the product of rules that take tactile and other sensory information as its input. The insect can thus be seen as an autonomous, quite nearly automatic living organism. In that respect, a human’s goal of living is similar to that of an insect. While our actions in our daily routine may not be consciously motivated by a goal to live, we are compelled to perform actions such as eating when we are hungry, doing our homework to avoid being scolded, and traveling to meet with others. Indeed, the human and insect are similar in that their actions can be boiled down to the desire to live, and in order to do so are equipped with a perpetual system that chooses what is considered to be the most apt course of action for a given situation. Their difference lies in the fact that we as humans have an overwhelming array of potential actions with which to respond to external information, and determining which course of action to take can be incredibly complicated. The more adept we are in reacting to the outside world, the greater our adaptability and survivability in a given environment becomes. While the insect is limited to responses based on an innate set of rules and actions, we are able to learn from instruction and experience; we then use our accumulated experience and knowledge to not only dynamically change our response system, but to also expand our repertoire of potential behavioral responses. Such a system that is capable of trying and acquiring different or new behavioral responses is what we could call autonomous. Based on this, one might consider Roomba vacuums or AIBO robotic dogs to have a will of their own. These automated machines, however, might be roughly categorized closer to the insect rather than autonomous animals that are capable of learning.
This principle of autonomy is vital in the next generation of AI development — that is, for AI that can truly coexist with human beings. Such AI technology is distinct from machine learning, an area of research and development in which Japan has fallen behind that of the US and China. However, when it comes to autonomy and versatility in AI, rather than lagging behind, Japan has the potential to stand at the forefront of future AI research and development.
Of course, autonomous AI research and development is not just garnering attention in Japan, but is actively progressing worldwide. Similar to discussions surrounding the problematic use of machine learning in producing deepfakes, many are denouncing the use of autonomous AI technology in lethal autonomous weapon systems (LAWS). As the name suggests, these are weapons equipped with autonomous AI that determines when to pull the trigger on its own. Unfortunately, several countries are currently developing such weapons, and some have already reached the level of practical use. Though there are ongoing efforts in the UN to ban LAWS, whether these inexcusable machines capable of ending human lives become a reality in an already senseless war will surely be a test of humanity’s intelligence.
Winner: E21 Scott Wilson
Take insects for example. They accomplish (and maintain) their goal of surviving by simply selecting behaviors from a pool of innate behavioral sequences, following rules that activate certain sequences in response to input from external stimuli. (These rules themselves also being something that the insects are born with.) For that reason, due to their actions being based on behavioral rules that respond to input from their tactile sensors, we can see them as an organism whose level of autonomy is extremely close to automatic. Now, humans also have the same goal of "survival" as insects, but whenever we do something in our daily lives, we're not consciously thinking that we're doing it to survive. Rather, we're taking steps to eat when we're hungry, or doing homework to not get yelled at, or relocating somewhere to meet up with someone. However, the base impetus behind all of these behaviors is survival. Both humans and insects alike accomplish their goal of surviving through a system of responding to their surroundings and choosing appropriate behaviors. The difference lies in the fact that humans have an overwhelming amount of possible actions that can be taken in response to external stimuli, and the links between those actions and which are activated in response are extremely complex. The better an organism responds to stimuli, the more its environmental fitness and chances of survival increase, and while insects can only respond through instinctive behavioral rules, we humans are capable of not only dynamically changing how we respond, but also dynamically increasing the variation of possible actions themselves. This is done by applying experience and knowledge that we've accumulated from being taught by others, or that we've discovered on our own. We could even say that this is the heart of an autonomous system: one with the ability to test out a variety of actions and acquire new ones in order to accomplish goals. Based on this line of thinking, while Roomba vacuum cleaners and AIBO robot dogs may appear to have a consciousness of their own, they are automated machines. Broadly speaking, insects also fall under that definition, but any being that is capable of learning can be said to have a system of autonomy.
This autonomy is extremely important in the implementation of the next generation of AI — AI that will actually coexist alongside humans — but is also AI technology that is separate from machine learning. While Japan is particularly lagging greatly behind the U.S. and China in terms of research and development in that field, at present I have identified that there is no such lag when it comes to autonomy and versatility, and I believe that it is possible for Japan to secure a solid position in future AI research and development.
Naturally, Japan is not the only country with its eye on the research and development of autonomous AI. Progress is actively being made all over the world in this field, and as we've seen before with the issue of deep fakes arising along with machine learning, such has the issue of LAWS (Lethal Autonomous Weapon Systems) been identified along with autonomous AI technology. As the name implies, this is the development of weapons equipped with autonomous AI that uses its own judgment when deciding to pull the final trigger. Unfortunately LAWS are currently being developed in many countries, and some weapons have already reached the stage of implementation. Initiatives to ban LAWS are being taken at the United Nations, but preventing a future where war — already monstrous enough — becomes even more monstrous through machines being the ones that decide who lives and dies, rests on humanity's judgment.