Inertia
Freshman
- Joined
- Jun 14, 2023
- Messages
- 86
- Reaction score
- 86
- Points
- 18
- What do we mean by the word "intelligent"?
- How do we know -when- an entity becomes intelligent?
The ability to understand the number of avenues that a more intelligent entity might outsmart your thinking is difficult to predict. *
There are boundary conditions to keep in mind, and one of them is that one doesn't have to be smart to appear smart. Intelligence involves the ability to solve a variety of problems, especially new problems. It requires:
- learning from mistakes
- inventiveness
- teaching and processing speed
- use of abstract thinking
- understanding concepts and relations (logic)
Software doesn't understand what it is doing, right? Simply following instructions fast is what machines do day in and day out.
So, how does one quantify intelligence in order to measure it? Differentiating between cognitive and physical abilities is one critical parameter. Age is another parameter. The ability to comprehend the meaning of a phrase, an equation, using working memory, processing speed, and definitions certainly counts more that the ability to rub one's stomach and pat their head simultaneously.
The Wechsler scale comes with two versions for an IQ assessment. One for adults and one for children. It's disadvantages include the language that people are most comfortable with and the education that they have attained. Another scale is Raven's progressive matrices. This evaluates abstract reasoning skills by assessing how one interprets geometric patterns in order to minimize vocalization bias. Scores are assessed using a sample group that quantifies their responses from zero with one hundred as a peak value on average. One standard deviation from the mean indicates that 68% of the sample group scored between 85 and 115.
These test completely depend on a comparative sample. Japan and Taiwan produce the highest scores. [ Average IQ = 106 ] In countries where illiteracy is common and malnutrition is rampant scored low consistently for obvious reasons. Teens cognitive skill sets are still in the processing stage and it's my personal opinion that people should wait until they are adults if they are interested in obtaining an IQ test.
The concept that intelligence can be reduced to a single number is inherently problematic. Intelligence is not simply cognitive ability, it includes the dimensions of practical (adaptive) and creative (new novel) intelligence too. All these common approaches are difficult to properly and quantitatively assess.
Artificial Intelligence:
A.) "narrow" or weak AI ---> perform tasks with high efficiency ( play games, recognize speech, object identification )
B.) "general" or strong AI ---- > comparable to humans for use in a very wide range of tasks and domains
C.) "super" intelligent AI ----- > surpasses human intelligence in most if not all tasks and domains
Should we give a super or general AI and IQ test?
Even a weak AI will outperform a human when tested for speed and memory. [ Should speed and memory be included at all? ]
Currently strong AI systems do not understand the idea of letters. [ Example: Provide me a list of animals that do not contain the letter "n". A current strong AI doesn't know that it doesn't know how to proceed. ]
An AI pretending to be intelligent doesn't count even if we cannot know the difference. We can be tricked.
_______
* Reference and credit: Sabine Hossenfelder
.
- How do we know -when- an entity becomes intelligent?
The ability to understand the number of avenues that a more intelligent entity might outsmart your thinking is difficult to predict. *
There are boundary conditions to keep in mind, and one of them is that one doesn't have to be smart to appear smart. Intelligence involves the ability to solve a variety of problems, especially new problems. It requires:
- learning from mistakes
- inventiveness
- teaching and processing speed
- use of abstract thinking
- understanding concepts and relations (logic)
Software doesn't understand what it is doing, right? Simply following instructions fast is what machines do day in and day out.
So, how does one quantify intelligence in order to measure it? Differentiating between cognitive and physical abilities is one critical parameter. Age is another parameter. The ability to comprehend the meaning of a phrase, an equation, using working memory, processing speed, and definitions certainly counts more that the ability to rub one's stomach and pat their head simultaneously.
The Wechsler scale comes with two versions for an IQ assessment. One for adults and one for children. It's disadvantages include the language that people are most comfortable with and the education that they have attained. Another scale is Raven's progressive matrices. This evaluates abstract reasoning skills by assessing how one interprets geometric patterns in order to minimize vocalization bias. Scores are assessed using a sample group that quantifies their responses from zero with one hundred as a peak value on average. One standard deviation from the mean indicates that 68% of the sample group scored between 85 and 115.
These test completely depend on a comparative sample. Japan and Taiwan produce the highest scores. [ Average IQ = 106 ] In countries where illiteracy is common and malnutrition is rampant scored low consistently for obvious reasons. Teens cognitive skill sets are still in the processing stage and it's my personal opinion that people should wait until they are adults if they are interested in obtaining an IQ test.
The concept that intelligence can be reduced to a single number is inherently problematic. Intelligence is not simply cognitive ability, it includes the dimensions of practical (adaptive) and creative (new novel) intelligence too. All these common approaches are difficult to properly and quantitatively assess.
Artificial Intelligence:
A.) "narrow" or weak AI ---> perform tasks with high efficiency ( play games, recognize speech, object identification )
B.) "general" or strong AI ---- > comparable to humans for use in a very wide range of tasks and domains
C.) "super" intelligent AI ----- > surpasses human intelligence in most if not all tasks and domains
Should we give a super or general AI and IQ test?
Even a weak AI will outperform a human when tested for speed and memory. [ Should speed and memory be included at all? ]
Currently strong AI systems do not understand the idea of letters. [ Example: Provide me a list of animals that do not contain the letter "n". A current strong AI doesn't know that it doesn't know how to proceed. ]
An AI pretending to be intelligent doesn't count even if we cannot know the difference. We can be tricked.
_______
* Reference and credit: Sabine Hossenfelder
.
Last edited:
