What is singularity?
For two years now, the media has talked a lot about singularity. But what is singularity? As it turns out, it’s not exactly a thing, but rather a moment in mankind’s future, where machines will become as smart as we are, if not more so. Is it a good thing for us to be heading this way? Of course, opinions differ. Some people are really positive about robots becoming smarter than us one day, as they see the singularity as a new way to enhance human capabilities in terms of memory and decision-making. But some others are not as thrilled, and they simply see the singularity as the end of mankind. We can only agree about the fact that both of these opinions are valid and justified, since it seems very likely that an ultra-connected global system will soon become a reality.
If we think about today’s technology, it seems very clear that devices connected to the web and to our everyday life will be the new trend in a very near future. We will then have to find a way to control all these interconnections, in order to offer every user a high quality service.
A possible way to achieve this is by using an aggregation of artificial intelligence algorithms, as they are the best at choosing the most appropriate software to enhance user experience, whether it’s on the Internet or in everyday life, with robotics and domotics. The result will be a worldwide distributed global system offering top quality service.
Global service and control
The problem we have here is that there are many ways of developing such algorithms, and some are riskier than others. If the scientific community is too permissive regarding the global vision of knowledge management and processing, the risk we face here is that users will end up encapsulated in some uncontrollable system in the Cloud. It’s something that we can actually see today, as many applications store users’ information and behaviours in order to offer a better web experience. What’s wrong with this, you might ask? Well, the problem is that nowadays, we tend to see artificial intelligence as a service, or a set of algorithms available online to any developer wishing to provide better features to an application. The real question is: how will these algorithms protect users’ private data?
As examples of artificial intelligence being offered as a service, we could talk about IBM, which started big this year with Watson Analytics. We can also talk about the Wolfram language, Numenta’s technology, Hierarchical Temporal Memory, or even Viv, an artificial intelligence whose purpose is to aggregate applications worldwide in order to connect them to a speech recognition and synthesis system.
Unfortunately, given what we know today about intelligent machines, it’s impossible for us to predict anything in the near future about how such a system will work with all its interconnected services.
Today’s computers’ systems organization shows that these services will mostly work in isolation with possible data exchange, but with no control of any kind. The evolutions in science and technology can’t be stopped, and this implies that there will be a global system at some point in the future. The key here is to think about a development strategy for this kind of system beforehand, so that it will work alongside mankind and its environment. It’s a bit like managing a city. Think about this: what’s the difference between a city with an urbanization plan and one without? It simply is that only the first one is livable. Well it’s exactly the same thing when considering knowledge, data, users and user behaviours. If the industry offers more and more API-based AI-related services, it could be very easy for ill-intentioned developers to control this data and knowledge, or even the entire system, by using breaches in APIs security.
Dangers of a non-symbolic system for robotics
We have just been over services and risks in the Cloud, but as we are going to see now, these risks also apply to robotics. Roboticists, scientists and engineers work together to build robots that are more and more capable. Some of these professionals prefer using non-symbolic systems as they believe that symbols are irrelevant when it comes to building an intelligent machine. And they are absolutely right. Some approaches are brilliant and show that robotics and artificial intelligence have a very bright future. But from my point of view, robots built on non-symbolic systems will have no real connection whatsoever with Humans. It is indeed possible to program such robots, but how do we monitor these machines in real time and make sure they don’t take an initiative that would be in contradiction with human rights? As it turns out, being proactive with strictly autonomous machines is extremely difficult, as it is impossible to monitor them in real time, i.e. plan what these machines will do next. If such systems can’t be reprogrammed in real time, then mankind will probably live by the machines rules. Our environment is so complex that it will be extremely difficult to find generic rules ensuring that machines keep their place.
New science and cognitive system
So, what could we do in order to build an intelligent system serving mankind? I believe that a symbolic system is a good alternative for developing robots and autonomous software, as it allows the creation of machines by using natural language only. In this case, being proactive and reprogramming a robot in real time become realistic possibilities.
This new alternative will probably lead to a new kind of science that we could call Meta science, or Hyper science, and we should keep a close eye on how this science evolves. If we are talking about a new science, it’s because intelligent systems will become more abstract in terms of problem solving.
Science is usually defined as analyzing a phenomenon on a time scale, comparing results and providing some hypothesis regarding observed behaviours. These systems will have a free will, and will be able to answer generic questions and solve complex problems, whether these problems are of a scientific nature or not.
These machines will have integrated a content of knowledge to their organizational memory beforehand, so that they can interact and answer correctly to questions. The mechanics behind this will remain the same, regardless of the type of knowledge that was integrated.
For example, knowledge related to a specific scientific area could be integrated and used to solve various problems, and some problems could even have a multidisciplinary aspect.
Even if science is an important part of this, it’s not the only part. For example, a generic system able to process worldwide data while taking into consideration the culture and customs of each country could be really helpful in interpreting global events.
Every problem that we tackle today with contemporary methods could be reshaped and rethought with intelligent machines.
Cognitive systems network
Imagine a world where all machines and software are built only with natural language. Think of all the services that could be interconnected: health, sports, education, transport, exploration, manufacture, and so on. Such a Meta system could communicate directly with the users and act according to each and every user’s specific needs.
Symbiosis between mankind and machine
From now on, human beings should be brought back to their central place. Science and technology are powerful tools, and we could use them to improve everyone’s life, worldwide, while keeping our planet clean and livable. The scientific community needs to get together, in order to provide some advice and to define policies to respect, so that everyone can feel safe and live in a healthy, prosperous environment. Governments should also take part in these discussions and be proactive regarding the future of “living together”. Being a civilization is a full-time job.