all this talk about AI/ML...
my 2 bits (could be my ignorance/lack of knowledge,understanding)
but i reckon there is no such thing as AI in the 'literal sense'.
the coder/programmer is the god.
the machine just cannot perform even the simplest task unless programmed to perform that.
without the software/code written, even the most sophisticated/advanced hardware cannot think on its own.
all the zettabytes of data are inert unless coded to perform any given task. computers/machines just cannot think on it's own.
AM I WAY OFF?
if i am, would someone dumb AI down...
thanks
Interesting discussion !
I'd like both you and Ella to continue digging deeper ...
-Mel
Well, this article:
https://www.nature.com/articles/d41586-019-00083-3
is trying to explain to non-mathematicians, that the "learnability problem" is undecidable, so we are free to build a world (axiom system) where it's true, or another world, where it's false.
Currently libraries like OpenCV are using raster images - these are identifying contours by processing 2D pixel arrays.
On the Internet I've found only some research papers about shape-based image recognition, which would use images stored as vectors and curves - there is a long way to go until getting really smart and reliable robots.
On Monday, August 2, 2021 at 9:48:34 AM UTC-6, Ella Stern wrote:
Well, this article:
https://www.nature.com/articles/d41586-019-00083-3
is trying to explain to non-mathematicians, that the "learnability problem" is undecidable, so we are free to build a world (axiom system) where it's true, or another world, where it's false.
Currently libraries like OpenCV are using raster images - these are identifying contours by processing 2D pixel arrays.My M.Sc thesis was entitled "On the Detection of Edges in Pictures" University of Alberta 1973. So your notes above lit a small memory fire in my soul :)
On the Internet I've found only some research papers about shape-based image recognition, which would use images stored as vectors and curves - there is a long way to go until getting really smart and reliable robots.
And I remember using/analyzing extent edge detection techniques, and then developing my own --- many years ago.
Thanks for your note that sparks more memories.
- Mel
Data Science includes Artificial Intelligence, and AI includesMachine Learning. ML has applications in natural language processing
Interesting discussion !
I'd like both you and Ella to continue digging deeper ...
-Mel
On Monday, August 2, 2021 at 7:53:56 PM UTC+5:30, meds...@gmail.com wrote:
Interesting discussion !actually mel,
I'd like both you and Ella to continue digging deeper ...
-Mel
ella (and most others) are just way ahead than me in their understanding/knowledge of tech/computers/science... and their functioning...
i am just not in their league...
but, am very keen to augment my knowledge by interacting/studying...
this may seem over-simplistic:
in gmail, if you type PFA in the subject/matter,
and you hit 'send', forgetting/without the attachment, gmail prompts to add the attachment.
is this sort of AI/ML or just plain coding.
.
if a family(with children) is streaming a movie on a smart tv,
and some 'X' rated content is about to be streamed on the screen...
would that smart TV (networked), sense presence of children (from the tv camera) and blank out those type of scenes.
(like a driverless car, sensing in it's path a wall/pole/obstuction...swerves)
On Tuesday, August 3, 2021 at 11:52:23 PM UTC-6, timec...@gmail.com wrote:
On Monday, August 2, 2021 at 7:53:56 PM UTC+5:30, meds...@gmail.com wrote:
Interesting discussion !actually mel,
I'd like both you and Ella to continue digging deeper ...
-Mel
ella (and most others) are just way ahead than me in their understanding/knowledge of tech/computers/science... and their functioning...
i am just not in their league...
but, am very keen to augment my knowledge by interacting/studying...
this may seem over-simplistic:
in gmail, if you type PFA in the subject/matter,I just 'sent' an email with PFA as the 'Subject', and nothing else. So, I guess my gmail is stupider than yours :))
and you hit 'send', forgetting/without the attachment, gmail prompts to add the attachment.
is this sort of AI/ML or just plain coding.
.So, I join you in not being sophisticated in AI.
if a family(with children) is streaming a movie on a smart tv,
and some 'X' rated content is about to be streamed on the screen...
would that smart TV (networked), sense presence of children (from the tv camera) and blank out those type of scenes.
(like a driverless car, sensing in it's path a wall/pole/obstuction...swerves)
Maybe we should get some help from Ella and see how she would use the Harbour language to create a pseudo-intelligent sample process.
As humans, we 'wonder' and are 'curious'. So, how can one build a proggy (in xHarbour) that wonders. Maybe,
Function Wonder(lStarttime, cSubject, etc, etc)
// do magical stuff here
Return cConjecture
-Mel
I think Reinforcement Learning is the area that might be of your interest.
Il 04/08/2021 23:18, Ella Stern ha scritto:
...
One more thing.
I answered you, but the post is intended for the OP. I credit you of a
vaste knowledge of IT, so I didn't want to teach you anyhting... maybe
I'd have something to learn from you, instead :-)
Dan
On Friday, August 6, 2021 at 5:20:29 PM UTC-6, Dan wrote:Hi Dan:
Il 04/08/2021 23:18, Ella Stern ha scritto:
...
One more thing.
I answered you, but the post is intended for the OP. I credit you of a vaste knowledge of IT, so I didn't want to teach you anyhting... maybe
I'd have something to learn from you, instead :-)
DanHi Dan:
I was never able to 'Save' the Rules.txt. It was always empty. i.e., the size of rules.txt == 0
I also note that 'saving' requires an 'S' rather than a Y for Yes.
Anyway thank you for the intro to machine learning in the Harbour language. I also quickly compiled your proggie under the Harbour fork too (using hbmk2 and MinGW11.2.0 and Harbour Version 3.2)
In xHarbour and Harbour, the program reactions were identical. But, in both case I got a zero-sized rules.txt
Thank you !
-Mel
On Friday, August 6, 2021 at 5:55:12 PM UTC-6, Mel Smith wrote:intro to ML. !
On Friday, August 6, 2021 at 5:20:29 PM UTC-6, Dan wrote:Hi Dan:
Il 04/08/2021 23:18, Ella Stern ha scritto:Hi Dan:
...
One more thing.
I answered you, but the post is intended for the OP. I credit you of a
vaste knowledge of IT, so I didn't want to teach you anyhting... maybe
I'd have something to learn from you, instead :-)
Dan
I was never able to 'Save' the Rules.txt. It was always empty. i.e., the size of rules.txt == 0
I also note that 'saving' requires an 'S' rather than a Y for Yes.
Anyway thank you for the intro to machine learning in the Harbour language. >> I also quickly compiled your proggie under the Harbour fork too (using hbmk2 and MinGW11.2.0 and Harbour Version 3.2)
In xHarbour and Harbour, the program reactions were identical. But, in both case I got a zero-sized rules.txt
Thank you !
-Mel
It appears that your use of 'Handles' has a small flaw in the closing statements of saving the 'rules.txt' file. i.e., lines 98 and 99 where you use fcreate() and fopen(). I fixed this for myself, and it works great. Anyway, I again appreciate this
-Mel
Il 07/08/2021 18:23, Mel Smith ha scritto:intro to ML. !
On Friday, August 6, 2021 at 5:55:12 PM UTC-6, Mel Smith wrote:
On Friday, August 6, 2021 at 5:20:29 PM UTC-6, Dan wrote:Hi Dan:
Il 04/08/2021 23:18, Ella Stern ha scritto:Hi Dan:
...
One more thing.
I answered you, but the post is intended for the OP. I credit you of a >>> vaste knowledge of IT, so I didn't want to teach you anyhting... maybe >>> I'd have something to learn from you, instead :-)
Dan
I was never able to 'Save' the Rules.txt. It was always empty. i.e., the size of rules.txt == 0
I also note that 'saving' requires an 'S' rather than a Y for Yes.
Anyway thank you for the intro to machine learning in the Harbour language.
I also quickly compiled your proggie under the Harbour fork too (using hbmk2 and MinGW11.2.0 and Harbour Version 3.2)
In xHarbour and Harbour, the program reactions were identical. But, in both case I got a zero-sized rules.txt
Thank you !
-Mel
It appears that your use of 'Handles' has a small flaw in the closing statements of saving the 'rules.txt' file. i.e., lines 98 and 99 where you use fcreate() and fopen(). I fixed this for myself, and it works great. Anyway, I again appreciate this
-Mel
Oops Mel, I quickly translated my little program, and since it was in Italian and "Yes" sounds "Sì" here, I missed that line.
You are right about the file rules.txt. The lines 98-99 are wrong. The correct syntax is:
nHandle:=fcreate("rules.txt")
There is no need to fopen after fcreate, fcreate returns the file handle
of the already opened file. The prog was written in Clipper many years
ago, and maybe there is a difference in the way fopen and fcreate do
work. For sure, that syntax worked in Clipper.
Since you are interested, here is a little explanation of the behavior
of the prog:
the matrix holds the weights of the properties for every animal. When
the prog has collected the 5 answers, it calculates the row that
totalizes the higher score, simply adding the weights for every "yes" (property present).
The interesting part is how the program corrects the weights when he
gives the wrong answer:
- it increases the weight of the properties of the right animal where it
got a "yes"
- it decreases the weight of the properties of all the others animals
where it got a "yes".
That's all. This very simple mechanism is incredibly efficient.
In another program, I was able to simulate a creature (let's say an ant) that starts from a point in the upper row of a "field" (a grid of cells)
and tries to reach its "home" in the bottom row.
The ant moves from one cell to the adjacent at every loop.
To be honest, the ant is more like a pigeon, because it is able to know
the direction it should take to reach home. It knows its coordinates
(x,y) on the grid, and the coords of home (x1,y1), so it can simply
subtract the x,y values to know if it's approaching or stepping away.
But the creature resembles to an ant for the other thing it can do: it leaves a chemical track on every visited cell.
The ant has two simple rules to follow: approach home and not enter a
cell already visited. Both rules are not mandatory: only discouraged.
But the chemical track is reinforced, should the ant enter a visited
cell, and the smell can become so strong that the ant can overcome the primary directive to not step away from home!
The ant can calculate which cell is the nearest to home at every loop,
and move accordingly.
The problem that the pigeonesque ant must solve is what to do when it encounters an obstacle.
If the obstacle is a simple horizontal wall, it is not difficult to
write a program that makes it to follow the wall until it ends and
restart approaching. In fact, the 3 cells below the ant are forbidden,
the 3 cells above discouraged because more far away from home, so the
only choice is to go horizontally.
But when the wall is like a "U" and the ant ends up inside the "U" it's
hard to make it to follow the wall, because it must step away from home
in order to overcome the obstacle. Without the chemical track the ant
would enter an infinite loop going left to right and right to left.
With the ES, the ant adjusts the score of the next move and "realizes"
that it must climb the wall, even if that is against the main directive,
if it wants not to be intoxicated by its own chemical substance.
I'll post the source here if someone is interested.
I was very proud to solve the problem with exactly the same algorythm
seen in neurnetw.prg, used in a ingenious way!
Dan
is 'siri' AI ? (in the literal sense - if AI in literal sense truly exists...)
is 'alexa' AI ?
am i right in believing?
any system/machine/robot/driverless car/appliance....
will execute EXACTLY 'ONLY' the 'code/syntax/function/command...' it processes.
(however wrong/off-logic the code/function...may be)
Sysop: | Keyop |
---|---|
Location: | Huddersfield, West Yorkshire, UK |
Users: | 339 |
Nodes: | 16 (0 / 16) |
Uptime: | 10:17:22 |
Calls: | 7,467 |
Calls today: | 3 |
Files: | 12,692 |
Messages: | 5,626,494 |