Don’t _miss

Stay Tuned


What are you looking for?

>Blog >How Far Are We From Achieving Artificial General Intelligence?

How Far Are We From Achieving Artificial General Intelligence?

Artificial general intelligence (AGI) is a machine’s hypothetical intelligence that can have the capacity to learn and understand any task that a human being would be able to do. An AGI machine would be able to understand the world like a human does, with the same scope and ability to learn how to carry out a wide range of tasks. 

Because of its potential, AGI has been also referred to as general intelligent action, full AI, and strong AI. Today’s artificial intelligence is decades away from being AGI as it is limited to software, study, or problem-resolving use instead of trying to perform human cognitive abilities. Even though AGI doesn’t exist, it has been featured in many science-fiction stories for over a century, with its biggest popularization achieved by the film 2001: A Space Odyssey. And starting with 2017, there have been over forty organizations that have been actively researching AGI.

What Could Artificial General Intelligence Do?

In theory, AGI could carry out any task that humans can and those that a human cannot. AGI would be able to combine the flexible thinking and reasoning of human beings with computational advantages. With the help of this intelligence, it would be possible to control robots at least like a person would and even get a new sort of machines that could do any human task. And over time, they would be able to take over any human role.

As Ben Goertzel claims, the next step is going to be the artificial general intelligence revolution, where we will see AI’s not just carrying out highly specific functions but AIs being able to understand the whole coordinate complex system and orchestrate narrow AIs and other narrow automated systems doing highly specific things. 

You can check Ben’s latest talk on artificial general intelligence he gave on Wonderland AI.

Tests for Confirming Human-Level AGI

There are several tests that are considered for confirming human-level AGI:

  • The Turing Test (Turing)
  • The Coffee Test (Wozniak)
  • The Robot College Student Test (Goertzel)
  • The Employment Test (Nilsson)

The Turing Test implies that both a human and a machine converse with a second human who will then evaluate which of them is a machine. If the machine successfully fooled the evaluator, it will pass the test.

The Coffee Test happens when a machine enters an average and tries to make coffee. It needs to find the coffee, the coffee machine, and a mug, add water and brew the coffee with the proper buttons. 

The Robot College Student Test will watch the machine as it enrolls in a university and tries passing the classes and getting a diploma, as a human would, while the Employment Test focuses on a machine working an economically important job and trying to perform as humans would in the same job. 

Should We Talk About AGI Now?

When will artificial general intelligence be invented? The answer depends on who you ask, but the answers range between 11 years and never. Why is it so hard to determine when we will have this technology? One of the reasons is because AGI lacks a clear path. 

Machine-learning systems today can underpin online services, allow computers to understand speech, recognize the language, find faces, and describe different photos and videos. But, the nature of the technologies we currently have is very different from what AGI should perform. Artificial general intelligence can be used for social good, even in the fight against covid. And the biggest problem is that these AIs aren’t necessarily what you would call stepping stones to develop general artificial intelligence.

During Wonderland AI, we got the chance to hear that Ben Goertzel thinks this will happen in the next five years and that we’re going to emerge into the AGI revolution. What does this mean for us? 

Ben Goertzel has pointed out that we use all sorts of AIs today, like the one that recommends music, one that searches for us, or one that gives voice control to our phone. But these have no understanding of what they are doing. But, as it turns out, there’s no overarching intelligent assistant that governs all of our internet interactions. Things like Google Assistant and Siri are very far from that right now. Simply put, this is mostly because we just don’t have the AI algorithms and structures now to understand the context and to bridge gaps between different narrow domains.