What is Artificial General Intelligence

What is artificial general intelligence?

Artificial general intelligence is a hypothetical type of intelligent agent that has the potential to accomplish any intellectual task that humans can. In some cases, it outperforms human capabilities in ways beneficial to researchers and companies. 

Companies such as OpenAI are conducting research to advance AGI and what it means for companies, governments, and humanity alike. AGI works by incorporating logic into the AI and machine learning processes instead of just applying an algorithm so that its learning and development mirrors that of humans.

AGI capabilities

AGI is essentially AI that has cognitive computing capability and the ability to gain complete knowledge of multiple subjects the way human brains can. It does not currently exist; it is simply in the process of being researched and experimented with. If it were able to surpass human capabilities, AGI could process data sets at speeds beyond what AI is currently capable of. Some of these could include:

• The ability to think abstractly

• Gathering and drawing from background knowledge of multiple subjects

• Common sense and consciousness

• Causation–a thorough understanding of cause and effect

In practice, this could include capabilities that humans have that AI does not, such as sensory perception. AGI could recognize colors and depth. Along with this are fine motor skills, like how a human reaches into their pocket to take out a wallet or cook a meal without burning their fingers on the stove. AGI could also develop creativity: Rather than generating a Renaissance painting of a cat, it could think of an idea to paint several cats wearing the clothing styles of each ethnic group in China to represent diversity. 

More than just a creative mind, painting cats wearing different Chinese dress patterns requires an understanding of different cultures, symbols, and belief systems. AGI systems would need to handle the subtle nuances of each ethnic group and create a new structure for this task using multiple algorithms at once. 

Examples of artificial general intelligence

While AGI systems are not available just yet, some examples of artificial intelligence are already meeting or exceeding human capabilities. Research and experimentation is currently underway to advance AI into AGI. Some examples of AGI already present in AI systems today include:

• Self-driving cars: These cars are guided by AI to recognize that other cars, people, and things like traffic cones or barriers are in their vicinity. Self-driving cars are designed to know when a car is nearby and to react if it is too close. They are also trained to adhere to driving laws and common rules of the road.

• Language model GPT: AI systems like ChatGPT can be prompted to generate human language that mimics how humans communicate. Both the input and output for tools like ChatGPT can be flawed. They also cannot simulate human emotion the way that AGI hypothetically could.

• Expert systems: An expert system is driven by AI to simulate human judgment. A common example is a health care expert system that prescribes specific types of medicine after reading a patient’s records.

• IBM’s Watson (and other supercomputers): Supercomputers like Watson can calculate faster than the average computer. With the addition of AI, they can carry out tasks such as modeling the birth of the universe.

If AGI were to exist, these examples would be able to surpass human intelligence. At the moment, self-driving cars require humans to be present to make decisions when the car’s AI capability does not allow it to problem solve in ambiguous situations.

The future of artificial general intelligence

In the future, though, AGI could do all this and more. While some researchers question whether it is viable, or even desirable, it is likely that experts will continue working to develop AGI. AI that operates beyond human capabilities is a point of artificial superintelligence that is sometimes called singularity. 

AI creations like ChatGPT and DALL-E were recently released to the general public and embraced by everyday users and professionals looking to generate outlines, sample prototypes, or simple lines of code. No matter how AI continues to develop, it is clear that AGI is on the horizon, and technology will advance to include this type of cognitive computing.

• Compiled by - Aayush Dumka Data Scientist