This article describes Artificial General Intelligence (AGI) or Strong AI and if it is even possible for the average person or beginner to understand. With the growth of Artificial Intelligence behind the scenes in our everyday lives, we don’t even realize it is working, especially on our smartphones. Yes, it is complex, and yes, we can understand it.
Artificial General Intelligence is possible. Many scientists agree that the existence of Artificial General Intelligence is achievable. Its emergence does not solely depend on engineering and software solutions and, at some point, will take on a mind of its own. According to experts, Artificial General Intelligence could emerge as early as this century.
Who Are The Artificial Intelligence Experts?
The Artificial Intelligence Experts are the people who are working with and building Artificial Intelligence and the industries that utilize AI. Inventors, Researchers, Authors, Teachers, Scientists, Engineers, Businesses, Owners, and Employees. From their garage to the lab to the workshop and factory, all work with Artificial Intelligence.
The rapid growth of technology has marked the last two decades. New solutions and programs are developed every day that improve the Smartphone’s and processor operation and bring them to a whole new level. 100 years ago, it was unthinkable to think about the internet, but today the internet is quite a common thing. Every year we are surprised by new technologies and new possibilities.
The question arises as to what the limits of technology are. Smartphones have become very intelligent, and they can perform increasingly complex tasks. In some areas, they have already overcome human intelligence. There is probably nothing that could impact human society more if AGI is realized. Is Artificial General Intelligence possible?
What Is Artificial General Intelligence?
Artificial General Intelligence or AGI has no generally accepted definition. Artificial General Intelligence would be the next level up, way up, from where we are today. A completely futuristic self-aware Artificial Intelligence like in the movies. Many scientists define AGI differently, but simply put, Artificial General Intelligence implies the ability of a machine to acquire and apply human-level knowledge, thinking, and understanding in different areas and apply them in a single system.
The most advanced current Artificial Intelligence today is Narrow AI. Artificial Narrow Intelligence (ANI), Weak Artificial Intelligence, or Narrow Artificial Intelligence. Computer programs implementing algorithms that can generate intelligent results in well-defined narrow areas focused on specific tasks.
We already have incredible Artificial Narrow Intelligence systems that are better than the best human intelligence and are evolving every day. Some of these programs are far superior to humans, such as Smartphone image recognition, speech recognition, and game playing in specific situations.
Artificial General Intelligence that can operate in different areas is different from current Artificial Intelligence. Artificial General Intelligence implies intelligence equal to human intelligence in all areas. AI can be an expert in a particular field and count faster than the world’s best mathematicians but know nothing about any other field. AGI has knowledge equal to humans in various fields, and this knowledge can be improved and perfected.
Artificial General Intelligence (AGI) Is It Possible?
Artificial General Intelligence (AGI) development is very possible, but it is difficult to say when. To move from narrow Artificial Intelligence, which is not fully developed even today, to Artificial General Intelligence, additional features are needed that will enable human-level intelligence.
Because Human Intelligence is so vast and different depending on the required capabilities for a definition, AGI could be equal to an average Human adult’s intellect or above. Again the problem is the measurement. Is it an IQ test, Turing Test, or the multitude of possibilities and standards?
According to Wikipedia, Human Intelligence is defined as:
“Human Intelligence is the intellectual capability of humans, which is marked by complex cognitive feats and high levels of motivation and self-awareness. Through intelligence, humans possess the cognitive abilities to learn, form concepts, understand, apply logic, and reason, including the capacities to recognize patterns, plan, innovate, solve problems, make decisions, retain information, and use language to communicate.”
Wikipedia – Human Intelligence
You can argue this point ad infinitum, but if a machine has achieved consciousness, is self-aware, and possesses the cognitive abilities described above, is it not a reasoning sentient being?
Conceptual Systems Moving Towards AGI
Artificial General Intelligence has not yet been developed in the full sense of the word because no program or computer could perform functions in all different areas and measure up to human intelligence in the same way as humans.
It could be argued that Narrow AI will not lead to AGI, and it can also be argued that it will. However, there are examples of Artificial Narrow Intelligence that show the first steps in these types of technology that show us in which direction this technology will go and what it can achieve.
Currently, Artificial Narrow Intelligence is making great inroads of growth towards AGI. 10 years ago, there was not Statistical learning used as it is today in everyday life. Things working on your computer, vehicle, or Smartphone, as an example, are like a futuristic movie or magic from twenty years ago. We have both massive data and computing power today in comparison to 10 years ago.
As described in Moore’s Law, this processing power with the number of transistors on a microchip doubles about every two years has fundamentally changed every processor-based machine like your Smartphone that we use. Every day that passes, you are seeing humans teach Machine Learning on some level to teach itself. 10 more years of supervised learning Machines teaching themselves will exponentially change Artificial Intelligence.
If Artificial Narrow Intelligence Can Do This Now, What Is Ahead?
As we see, in an artificial intelligence game playing, the computer makes moves that humans have not thought of yet. After already beating the best players globally, now the next generation Google DeepMinds MuZero is mastering games without knowing the rules as in the past.
Open Ai’s GPT-3 large-scale unsupervised language model. “Given a news article title and first sentence, it can generate full articles by predicting the next word that is likely to appear. The resulting news articles are so good that humans can’t tell if they are real or machine-generated.”
Open Ai’s GPT-3 Beta Application Programming Interface (API), which is currently classified as Narrow Artificial Intelligence, can translate language such as English straight into computer code. So step back a moment and think about Artificial Intelligence that can utilize Supervised Machine Learning in this way.
An average person could theoretically type in a statement to a computer in English, say, “Write a program to stop global warming.” With advanced Machine Learning, it may take a while but with the computational resources and the machine training itself. It never sleeps, it never eats, and it never stops.
What is the result? A Computer program that can stop global warming. You may not like the result of what has to happen to achieve this, but that may not matter.
We have Artificial Narrow Intelligence programs today that, in weeks and months, adjust, create, or learn a new algorithm, then are beating the very best humans on the planet at certain tasks. In five or ten years, what is a self-learning entity capable of if given the proper resources?
AGI Does Not Need to Mimic the Human Body or Brain, But It Can
Many researchers utilize Neural Networks that mimic the human brain’s architecture. A human brain has roughly 100 to 500 trillion synaptic connections. AGI does not need to Mimic the Human Body or Brain to be an Artificial Intelligence superior to Humans. On the contrary, it may be completely different and operate just fine.
One aspect of artificial intelligence is its ability to utilize Neural networks. Neural networks are a set of algorithms or computer instructions modeled after the human brain designed to recognize the information.
In essence, what AI does is mimic the body and cerebrum in its use of a Neural Network Brain, Computer Vision Eyes, and Natural Language Processing Ears for processing internal and external stimuli and information.
- Deep Neural Networks – Like the human brain, neural networks are many related elements that mimic neurons in an artificial brain.
- Computer Vision – Computer vision is your smartphone camera or equivalent device that facilitates how computers can understand digital images or videos or essentially see.
- Natural Language Processing (NLP) – NLP handles the processing of human languages like speech or text.
Artificial Intelligence Features
The first of such features is the sensory perception needed to differentiate objects and understand them. Self-driving car systems can drive themselves, but not perfectly. For example, currently, they can be deceived by a sticker on a stop sign, thus ignoring the sign and causing an accident, and it is difficult for them to understand three-dimensional space.
AGI technology should be able to understand communication and reading between the lines. People don’t say everything they see, they don’t use details in communication, and it is often assumed that you know what someone is talking about. Artificial General Intelligence would have to cover a lot of data and interpret it to understand the bigger picture. Creating an AGI to have human intelligence is not easy, and one needs to take care of several nuances that come to people naturally and unconsciously.
When creating this type of technology, special attention must be paid to engineering and the values that this technology should have. Since AGI technology is expected to have human intelligence and abilities, it should also have moral values and ethical guidelines according to which it will determine its own behavior.
If such intelligence’s values and moral compass are not taken into account during design, it could very quickly become harmful and perhaps dangerous for human society.
There are already discussions and scientific research on the ethical aspects of AGI that suggest that such intelligence be kept in a simulated virtual world until their behavior and “way of thinking” is fully understood under controlled conditions. Only when we are sure (as if we ever could be) that this technology is safe to use we can take it a step further and apply it in the real world.
How Far Away Is AGI?
To develop Artificial General Intelligence, it is necessary to perfect the so-called weak or narrow AI capable of solving problems in a specific area. Already today, there are many examples of such intelligence in the form of applications and expert systems as described above. As we have managed to improve such system’s work in some areas, there is a desire to create artificial general intelligence that will integrate more areas and thus have abilities and intelligence equal to humans.
Some scientists speculate that the rapid development of technology may bring Artificial General Intelligence in this century. Today, many universities have an AGI course, and numerous researches are being conducted, which could accelerate this technology’s development.
Other scientists believe that we will not be able to create AGI in this century. However, it isn’t easy to imagine what we will achieve in the coming years, given that technology is improving and evolving day by day and so fast.
To create this type of technology, we need a turning point in the world of technology, i.e., an engineering solution that will incorporate the existing technology into a completely new system that will have the ability to think equal to humans. The human brain uses thousands of neurons for its work in just a few seconds, and even today, we do not fully understand all the functions of the human brain.
What Would Be The Medium For AGI?
What would be the medium for Artificial General Intelligence? Should it be a self-contained robot, a Smartphone, the Cloud, the IoT, or something else? Currently, for the AGI to access unlimited data, it needs a connection to the internet through signal, either wired or mobile.
The major smartphone manufacturers Huawei and Apple started embedding Neural Engines in their processors in 2017. Still, access to data, size, and power put you back to the issue of a signal. Can AGI constantly access updated information to keep working and learning?
There is also the option of incorporating and merging AI with us in our physical bodies like Elon Musk’s Neural Link Project. We wrote about “Smartphone AI: Neuralink Project” last year and are watching as it progresses.
The front runner is most certainly a Robot as that is what the world learned from the movies all these years. This also saves you from the fear of attaching anything to your brain, as in Neuralink. However, that may be the dream of some or many.
Artificial General Intelligence And Commercial Secrecy
Artificial General Intelligence should function very similarly to the human brain, so it would in itself be a very complex system. An additional problem with the creation of Artificial General Intelligence is that the various groups working to create such an entity compete over who will come up with the solution for the development of AGI first, which is why much information is not shared among scientists.
Commercial secrecy in AI is a major obstacle between business and science but also between superpowers like the United States, China, India, and Russia. It will likely take some global catastrophe for everyone to join “Team Human” to “work together” in the word’s true sense.
To create Artificial General Intelligence, it may not be necessary to discover how the human brain works and then transfer that knowledge to the digital world. Artificial intelligence may design itself and create technology equal to humans.
References:
- https://www.researchgate.net/profile/Prof_Dr_Hugo_De_GARIS/publication/226000160_Artificial_Brains/links/55d1e55308ae2496ee658634/Artificial-Brains.pdf#page=14
- https://www.qualcomm.com/products/smartphones
- https://insights.roboglobal.com/sophia-the-robot-ben-goertzel_Intellligence_AGI_
- https://insights.roboglobal.com/sophia-the-robot-ben-goertzel
- https://arxiv.org/pdf/0712.3329.pdf
- https://deepmind.com/about
- https://arxiv.org/pdf/1705.08807.pdf
- https://futureoflife.org/ai-open-letter