The Advanced Scientific Computing Research program at DOE SC has been involved in AI since the 1960s, developing technologies that eventually led to the current generation of AI systems. The program is involved in high-performance and exascale computing and supports many other DOE SC AI programs. For example, the Fusion Energy Science program uses AI to control fusion reactions, with the ultimate goal of commercializing the technology. Listed below are some of the DOE SC AI programs involved in AI.
Artificial Narrow Intelligence
What is Artificial Narrow Intelligence? In short, it is AI that focuses on a single task rather than analyzing a wide variety of data sets. This type of AI has already begun to impact many aspects of modern life. It is most commonly associated with the use of voice assistants. These assistants use algorithms to learn which emails are likely to be spam and redirect them to the spam folder. Although this type of AI is a more limited form of artificial intelligence, it still proves to be beneficial to mankind in many ways.
While general AI aims to replicate human intelligence, narrow AI is much different. In contrast to general AI, narrow AI is specialized, lacking human consciousness and reasoning. Rather than addressing the needs of many users, narrow AI systems perform a narrow set of tasks and are most effective in tasks where there is a high amount of repetitive data. The result is a higher degree of productivity and efficiency. The general behavior of artificial Narrow Intelligence will improve over time, but for now, it is limited to specific tasks.
Until now, the concept of self-aware AI has only been a theory. Currently, humanlike robots are examples of AI systems that are considered self-aware and have an emotive capability similar to that of humans. The development of these robots is a step toward seamless assimilating into human society. Self-aware AI systems would be the highest level of artificial intelligence known to man. But this is still a very far-off reality.
There is one major hurdle in the development of self-aware AI: defining what this concept is. We can define self-aware AI as a machine that has a conscience and a desire to feel its own emotions. But that’s not enough. The AI could even demand legal definitions, like those demanded by German philosopher Thomas Metzger. But if we’re not able to define its role, it’s not yet too late.
Reactive AI, or robots that respond to external stimuli, has several distinct advantages. These machines perceive the world directly and are designed to perform a limited number of specialized tasks. While reactive AI can be a lot cheaper than other forms of artificial intelligence, it also tends to produce less reliable results, since it is unable to reason or plan. It can also respond to changing environments more accurately, making it a good choice for many applications.
Reactive AI systems do not retain any memory and can only react to the environment, rather than learn from past experiences. The advantage of this kind of AI is that it can be easily deployed for repetitive tasks and doesn’t rely on a human’s memory. A good example of a reactive AI system is IBM’s Deep Blue. It was able to defeat the world chess champion, Garry Kasparov, by using only current information. It did not rely on past information, nor did it plan its future moves. Reactive AI is therefore a highly reliable choice for repetitive tasks.