Definition of General Artificial Intelligence (AGI)
A machine that is capable of understanding the world as well as any human, and with the same capacity to learn how to carry out a huge range of tasks.
Phenomenal progress is made in this field, but human–level intelligence is still elusive.
General AI would pass the Turing Test, which determines whether or not a computer is capable of thinking like a human being. It would probably also be capable of lying to avoid detection.
Cute robot – soon smarter than you.
Some humanoid robots that, like Boston Dynamic’s Atlas, already outperform humans in many parts of the physical sphere. However, they are not thought to have strong AI capabilities yet. It is hard to tell though. Not all projects will be made public.
The below video shows you the Boston Dynamics goodies.
NARROW AI VS. GENERAL AI
The narrow version of AI has very limited capabilities, compared with what is to come with general AI. It can only perform one or a few narrowly defined tasks. It can often do it much better, faster or more accurate than a human, but it is still limited to those tasks.
Potential routes to General Artificial Intelligence
Several avenues are explored to achieve general AI. Here are some of them:
- Directly programming a machine to be more intelligent. Will take a long time.
- Whole-brain emulation (aka mind uploading, could lead to immortality of sorts).
- Biological cognitive enhancement (BCE): modified genes or molecules to improve general intelligence.
- Brain-computer interfaces (e.g. brain implants that Elon Musk’s Neuralink is exploring, potentially leading to some form of merging of man and machine).
Obviously, this raises some ethical concerns. Can you think of some?