Our server costs ~$56 per month to run. Please consider donating or becoming a Patron to help keep the site running. Help us gain new members by following us on Twitter and liking our page on Facebook!
Current time: March 29, 2024, 3:07 am

Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Let's measure intelligence in 2016
#11
RE: Let's measure intelligence in 2016
(March 23, 2016 at 11:17 am)Mathilda Wrote:
(March 23, 2016 at 10:38 am)pool the great Wrote: .....should I?

If you want to include the type of intelligence that can catch a ball, plan moves ahead, predict something occurring, recall a tune or learn a dance step then yes.

In terms of AI I'd argue that this is what makes an autonomous agent intelligent rather than a stimulus / response unit.
That'd probably come under pattern analysis..
Reply
#12
RE: Let's measure intelligence in 2016
(March 23, 2016 at 11:17 am)Mathilda Wrote: If you want to include the type of intelligence that can catch a ball, plan moves ahead, predict something occurring, recall a tune or learn a dance step then yes.

In terms of AI I'd argue that this is what makes an autonomous agent intelligent rather than a stimulus / response unit.

I was going to say that counts more as heuristic capability than intelligence; I'd always considered heuristics as a property of intelligence rather than a category. Since you say there are practical examples in AI research, I'd love to hear the reasoning as to why it's a causal measure rather than an emergent property.
Sum ergo sum
Reply
#13
RE: Let's measure intelligence in 2016
I find the more I do puzzles of the sort they use in IQ tests the better I do at the tests. this does not mean that I get smarter just better at those sorts of things. They are more or less meaningless.



You can fix ignorance, you can't fix stupid.

Tinkety Tonk and down with the Nazis.




 








Reply
#14
RE: Let's measure intelligence in 2016
(March 23, 2016 at 11:47 am)pool the great Wrote: That'd probably come under pattern analysis..


Your perspective may change if you ever try creating artificially intelligent temporal sequence learning. It is significantly harder to do than pattern analysis, which is probably why the literature has relatively little of the former and lots about the latter.

And if you believe that strong AI has to be self organising which I definitely do, then self organised temporal sequence learning is fiendishly difficult.

Essentially all the different techniques that we use for pattern analysis are statistical in nature, or equivalent to some other statistical technique. In essence pattern analysis is more like devising a complicated look-up table.

Temporal sequence learning could be considered pattern analysis but with the added dimension of time, but this doesn't work in the case of action selection which does not need to create a model of the world.
Reply
#15
RE: Let's measure intelligence in 2016
(March 23, 2016 at 1:50 pm)Ben Davis Wrote:
(March 23, 2016 at 11:17 am)Mathilda Wrote: If you want to include the type of intelligence that can catch a ball, plan moves ahead, predict something occurring, recall a tune or learn a dance step then yes.

In terms of AI I'd argue that this is what makes an autonomous agent intelligent rather than a stimulus / response unit.

I was going to say that counts more as heuristic capability than intelligence; I'd always considered heuristics as a property of intelligence rather than a category. Since you say there are practical examples in AI research, I'd love to hear the reasoning as to why it's a causal measure rather than an emergent property.

Could you explain what you mean by describing it as a heuristic capability rather than intelligence? Surely any heuristics would either be evolved or learned through intelligence?

I think debating whether it's a causal measure rather than an emergent property sends us down a never ending rabbit hole of definitions and am not sure why the distinction needs to be made.

Most AI research shies away from adapting to sequences because the search space explodes exponentially. Just talking about temporal sequence learning is a very vague term in itself, which is what I tried to get across with all the examples I gave. You do get a few examples in the literature though but they are limited in scope and application.
Reply
#16
RE: Let's measure intelligence in 2016
(March 23, 2016 at 2:38 pm)Mathilda Wrote: Could you explain what you mean by describing it as a heuristic capability rather than intelligence? Surely any heuristics would either be evolved or learned through intelligence?
Only that learning for oneself is a consequence of a number of facets of intelligence (e.g. memory, emotional development, resilience...).

Quote:I think debating whether it's a causal measure rather than an emergent property sends us down a never ending rabbit hole of definitions and am not sure why the distinction needs to be made.
And that's fair enough. In practical terms, one's heuristic capability gives insight into levels of intelligence i.e. one who shows high levels of capability is likely to score highly in base measures of intelligence. It's interesting to understand the hierarchy of the facets of intelligence though I'm not going to get stuck in an argument about what comes before what when the applications are even more interesting!

Quote:Most AI research shies away from adapting to sequences because the search space explodes exponentially. Just talking about temporal sequence learning is a very vague term in itself, which is what I tried to get across with all the examples I gave. You do get a few examples in the literature though but they are limited in scope and application.
And that's curious because the human brain is very good at it without needing masses of processing power. It seems to be the categorisation & organisation of information that's the key although we're only beginning to unlock how the brain does that. It's not surprising that we can't yet translate that in to practical developments in AI.
Sum ergo sum
Reply
#17
Let's measure intelligence in 2016
How about your cultural background IQ? Also known as BITCH.

B. I. T. C. H. Black Intelligence Test of Cultural Homogeneity
http://www.susanohanian.org/show_commentary.php?id=170

Example
Quote:6. Cheap chitlings (not the kind you purchase at a frozen food counter) will taste rubbery unless they are cooked long enough. How soon can you quit cooking them to eat and enjoy them?
a. 45 minutes
b. 2 hours
c. 24 hours
d. 1 week (on a low flame)
e. 1 hour

7. What are the "Dixie Hummingbirds?"
a. part of the KKK
b. a swamp disease
c. a modern gospel group
d. a Mississippi Negro paramilitary group
e. Deacons
Reply
#18
RE: Let's measure intelligence in 2016
(March 23, 2016 at 2:02 pm)downbeatplumb Wrote: I find the more I do puzzles of the sort they use in IQ tests the better I do at the tests. this does not mean that I get smarter just better at those sorts of things. They are more or less meaningless.

Not really. Iq tests are means of testing the capacity of pattern analysis of a person. When you do iq tests regularly you get used to detecting these patterns and thus your result improves.
Reply
#19
RE: Let's measure intelligence in 2016
(March 23, 2016 at 2:30 pm)Mathilda Wrote:
(March 23, 2016 at 11:47 am)pool the great Wrote: That'd probably come under pattern analysis..


Your perspective may change if you ever try creating artificially intelligent temporal sequence learning. It is significantly harder to do than pattern analysis, which is probably why the literature has relatively little of the former and lots about the latter.

And if you believe that strong AI has to be self organising which I definitely do, then self organised temporal sequence learning is fiendishly difficult.

Essentially all the different techniques that we use for pattern analysis are statistical in nature, or equivalent to some other statistical technique. In essence pattern analysis is more like devising a complicated look-up table.

Temporal sequence learning could be considered pattern analysis but with the added dimension of time, but this doesn't work in the case of action selection which does not need to create a model of the world.

Oh if you're talking about computers, then I've found a method to make computers work like human brains.
It's all in my head and quite complicated to communicate across a message board clearly. But I'll tell you my method uses lots and lots of linked lists,sort of like neurons in our brain. And lots and lots of complicated connections between them. Many to Many connection to be exact. I could implement it in a c program but I'm not sure how it'll play out since my beloved laptops only input is his keyboard input and no sensory inputs like my awesome body has. I feel like a lot of learning can be done by computers without actually Hard coding or by using a reward system if we were to give computers some sensors so that sensory inputs can be read. That way a lot of connections across nodes can be done by them itself..
Like fire can be associated with pain.pain can be associated with avoid. Therefore fire can be associated with avoid. Get it? Big Grin
Reply
#20
RE: Let's measure intelligence in 2016
(March 23, 2016 at 11:12 pm)pool the great Wrote:
(March 23, 2016 at 2:02 pm)downbeatplumb Wrote: I find the more I do puzzles of the  sort they use in IQ tests the better I do at the tests. this does not mean that I get smarter just better at those sorts of things. They are more or less meaningless.

Not really. Iq tests are means of testing the capacity of pattern analysis of a person. When you do iq tests regularly you get used to detecting these patterns and thus your result improves.

Indeed, they are meaningless.



You can fix ignorance, you can't fix stupid.

Tinkety Tonk and down with the Nazis.




 








Reply



Possibly Related Threads...
Thread Author Replies Views Last Post
  Human defeated by Artificial Intelligence, from confidence to utter powerlessness causal code 15 2550 October 29, 2017 at 5:16 am
Last Post: I_am_not_mafia
  Supermathematics and Artificial General Intelligence ThoughtCurvature 28 6052 October 23, 2017 at 7:32 pm
Last Post: causal code
  Genius Edward Witten, could he help to intensify artificial intelligence research? ThoughtCurvature 1 994 September 5, 2017 at 5:29 am
Last Post: Edwardo Piet
  Intelligence, Creativity, and a Touch of Madness Rayaan 11 3894 October 25, 2014 at 8:08 am
Last Post: Thumpalumpacus
  Creationist Vs Scientist On Why Human Intelligence Is Declining Gooders1002 0 1156 March 29, 2014 at 1:08 pm
Last Post: Gooders1002
  How to measure consciousness? The_Flying_Skeptic 3 2216 November 27, 2011 at 7:06 am
Last Post: The Grand Nudger



Users browsing this thread: 1 Guest(s)