I have been a computer programmer, using various programming languages, for more than 20 years, but after my retirement in RL I only continued programming in LSL, the programming language of Second Life.
I stopped following computer developments in 2008 or so and now we have artificial intelligence and my head is spinning, trying to understand what it really is.
The point is the language. They talk about neurons, those little things in our brain, about networks, the synaps connections in our brain and intelligence, but the core issue here is that it is just about computer programming. All the other talk is metaphorical.
And this programming is done by human beings. I have studied some of the material but I can tell you, it is beyond my comprehension. Not the Python programming language, but the way sources talk about neural networks.
Just listen to ChatGPT: A neural network is a computational model inspired by the structure and function of the human brain. It consists of interconnected nodes, called neurons, organized into layers.
Each neuron receives input signals, processes them, and generates an output signal, which is passed on to other neurons.
The simplest form of a neural network is the feedforward neural network, which consists of an input layer, one or more hidden layers, and an output layer.
In this architecture, information flows in one direction, from the input layer through the hidden layers to the output layer, without any feedback loops.
During training, a neural network learns to perform a specific task by adjusting the strengths of connections between neurons, known as weights.
This is typically done using a process called backpropagation, where the network compares its output to the desired output and adjusts the weights accordingly, aiming to minimize the difference between the two. -END-
My key point was the concept of "weights". There is an input and there is a desired output. For instance, the AI should tell you if it is a human face or not.
Ok, let's accept that language... the input is transferred to nodes that weigh the data to determine if the picture is a human face or not. And here comes the trick.
It took me quite some questioning to learn who decides about the value of these weights. Just suppose the weight is "90% white". That was the human programme's input.
Finally, ChatGPT says: Yes, the values of the weights in a neural network are typically initialized by humans or by specific initialization schemes designed to set them to reasonable starting values.
These initial values are crucial because they influence how the network learns and how quickly it converges to a solution during training. -END-
And here comes my second key point: that talk about "training" an AI program. What is really meant is that the program itself is programmed in such a way that it can "adjust" these weights in relation to input and desired output.
This could mean that a "trained" AI program by changing the values of the weights used in its processing, could develop a "mind" of its own, where the user no longer has any idea how the AI program has come to some conclusion.
I may be wrong. I am not an expert here, but yet it makes me wonder.....
Thank you for your attention again...
Main Sources:
MacMillan The Encyclopedia of Philosophy, 2nd edition
TABLE OF CONTENT -----------------------------------------------------------------
1 - 100 Philosophers 9 May 2009 Start of
2 - 25+ Women Philosophers 10 May 2009 this blog
3 - 25 Adventures in Thinking 10 May 2009
4 - Modern Theories of Ethics 29 Oct 2009
5 - The Ideal State 24 Febr 2010 / 234
6 - The Mystery of the Brain 3 Sept 2010 / 266
7 - The Utopia of the Free Market 16 Febr 2012 / 383
8. - The Aftermath of Neo-liberalism 5 Sept 2012 / 413
9. - The Art Not to Be an Egoist 6 Nov 2012 / 426
10 - Non-Western Philosophy 29 May 2013 / 477
11 - Why Science is Right 2 Sept 2014 / 534
12 - A Philosopher looks at Atheism 1 Jan 2015 / 557
13 - EVIL, a philosophical investigation 17 Apr 2015 / 580
14 - Existentialism and Free Will 2 Sept 2015 / 586
15 - Spinoza 2 Sept 2016 / 615
16 - The Meaning of Life 13 Febr 2017 / 637
17 - In Search of my Self 6 Sept 2017 / 670
18 - The 20th Century Revisited 3 Apr 2018 / 706
19 - The Pessimist 11 Jan 2020 / 819
20 - The Optimist 9 Febr 2020 / 824
21 - Awakening from a Neoliberal Dream 8 Oct 2020 / 872
22 - A World Full of Patterns 1 Apr 2021 / 912
23 - The Concept of Freedom 8 Jan 2022 / 965
24 - Materialism 7 Sept 2022 / 1011
25 - Historical Materialism 5 Oct 2023 / 1088
26 - The Bonobo and the Atheist 9 Jan 2024 / 1102
27 - Artificial Intelligence 9 Feb 2024 / 1108
The Discussion
[13:25] herman Bergson: To keep it simple....
[13:25] herman Bergson: There is input...and a DESIRED output....
[13:26] herman Bergson: inbetween is an AI program, that gets to that result....
[13:26] herman Bergson: This is all human work....
[13:26] Max Chatnoir: And the feedback also.
[13:26] Gemma (gemma.cleanslate): I really need to find time to play with some of them
[13:26] .: Beertje :. (beertje.beaumont): seems to me like Blender, we have to work with nodes there too and weights to get the result we want
[13:27] Gemma (gemma.cleanslate): just to see how different ones work and come up with the same or different stuff
[13:27] Gemma (gemma.cleanslate): i bet there is a basic form of ai behind blender
[13:27] Gemma (gemma.cleanslate): and all those programs
[13:27] herman Bergson: Yes Gemma,,,discuss with ChatGPT, Bard and others :-)
[13:28] herman Bergson: My point today is......
[13:29] herman Bergson: there is INPUT,,,,,AI.....and human desiged, requested OUTPUT
[13:29] .: Beertje :. (beertje.beaumont): who desides what the DESIRED output is?
[13:29] herman Bergson: the point is that the relation between INPUT and OUTPUT is too much work for a human being...
[13:30] herman Bergson: Who decides what the desired output is....
[13:30] Gemma (gemma.cleanslate): I think of all the eons in which so much input was created step by step by humans encyclopedias of all kinds of information on the world
[13:30] Gemma (gemma.cleanslate): and that is what is used
[13:30] herman Bergson: Beertje...we have our own scandal in the Netherlands...
[13:30] Gemma (gemma.cleanslate) whispers: oh?
[13:31] Max Chatnoir: ??
[13:32] herman Bergson: Yes Gemma.... all encyclopedian data can be input in the AI system...
[13:32] Gemma (gemma.cleanslate): yes
[13:32] herman Bergson: And it works...can't deny that
[13:32] Gemma (gemma.cleanslate): yep
[13:33] herman Bergson: Oh...the Dutch scandal....???
[13:33] Max Chatnoir: But what if you give it an unfamiliar task?
[13:33] .: Beertje :. (beertje.beaumont): de toeslagenaffaire
[13:33] herman Bergson: Not that one Beertje but is a good one too
[13:34] Gemma (gemma.cleanslate): what scandal
[13:34] herman Bergson: The police used a kind of AI system to determine in what quarters would be more criminality...
[13:34] Max Chatnoir: I'm so up to my ears in US stuff, I never hear about anything else!
[13:34] Gemma (gemma.cleanslate): quarters? locations?
[13:34] Max Chatnoir: You mean like neighborhoods?
[13:35] Gemma (gemma.cleanslate): yes
[13:35] herman Bergson: Max...don't appolegize..you only have two ears and the world is so much bigger :-)
[13:35] Max Chatnoir: Too true!
[13:35] herman Bergson: Yes Max....
[13:35] Gemma (gemma.cleanslate): right
[13:35] Gemma (gemma.cleanslate) GIGGLES!!
[13:35] Gemma (gemma.cleanslate): ...LOL...
[13:36] herman Bergson: Was a real mean discriminating tool....
[13:36] Gemma (gemma.cleanslate): hard to get info on Africa unless I listen to the bbc in the middle of the night
[13:36] herman Bergson: yes something like that Gemma
[13:36] herman Bergson: I even don't know where to find Africa :-)))
[13:36] Gemma (gemma.cleanslate): ha
[13:38] herman Bergson: But my main point today is....
[13:38] herman Bergson: AI is programmed to offer the DESIRED output....
[13:39] herman Bergson: and neural networks adjust themselves to get to that
[13:39] herman Bergson: and in a way...I find it a scary observation
[13:39] herman Bergson: Suppose...
[13:40] herman Bergson: desired output is....black person...85% danger
[13:40] herman Bergson: Cameras observe a location....
[13:40] herman Bergson: and monitor the faces of people....
[13:40] herman Bergson: BINGO!
[13:41] Gemma (gemma.cleanslate): those dont work .. already proven
[13:41] Gemma (gemma.cleanslate): and asian people too
[13:41] herman Bergson: This is what is happening in China....
[13:41] Gemma (gemma.cleanslate): ye
[13:41] .: Beertje :. (beertje.beaumont): yes
[13:41] .: Beertje :. (beertje.beaumont): they observe every woman and man
[13:41] herman Bergson: You are mionitored...don't fit the desired output...
[13:42] herman Bergson: and you get punished in a way
[13:42] Gemma (gemma.cleanslate): ah well
[13:42] herman Bergson: So...to get to the real point.....
[13:42] herman Bergson: We as human beings...
[13:43] herman Bergson: formulate the DESIRED output.....
[13:43] herman Bergson: in other words...
[13:43] herman Bergson: AI can be used in a highly political way
[13:43] .: Beertje :. (beertje.beaumont): yes
[13:43] Gemma (gemma.cleanslate): that we know in the US from last elections
[13:44] Max Chatnoir: It will certainly reflect input biases.
[13:44] .: Beertje :. (beertje.beaumont): and the coming elections I suppose
[13:44] Gemma (gemma.cleanslate): yep
[13:44] Gemma (gemma.cleanslate): big push now to get the social networks to control it more
[13:44] Gemma (gemma.cleanslate): here
[13:44] herman Bergson: For those US elections you don't need Artificial Intelligence....
[13:45] herman Bergson: They even can't spell the word...
[13:45] Guestboook van tipjar stand: Gemma Cleanslate donated L$100. Thank you very much, it is much appreciated!
[13:45] Gemma (gemma.cleanslate) GIGGLES!!
[13:45] Gemma (gemma.cleanslate): ...LOL...
[13:45] Gemma (gemma.cleanslate): tsk
[13:45] Gemma (gemma.cleanslate): Max will we let him say that?????
[13:45] herman Bergson: sorry :-)
[13:45] .: Beertje :. (beertje.beaumont): US needs intelligence
[13:45] Max Chatnoir: Depends who "they" is....
[13:45] Gemma (gemma.cleanslate): oh there are many of us who are intelligent but a lot who are not
[13:46] herman Bergson: Beertje...they even think they are exceptional...
[13:46] .: Beertje :. (beertje.beaumont): hmm
[13:46] herman Bergson: Call it exceptionalism...
[13:46] Gemma (gemma.cleanslate): the country was .. but it is getting more like the old countries now
[13:46] herman Bergson: Well.... in a certain sense...I agree :-)))
[13:47] herman Bergson: You got a point Gemma
[13:47] Gemma (gemma.cleanslate): yes
[13:47] Gemma (gemma.cleanslate): we are very young
[13:47] herman Bergson: I grew up with the believe that the US was THE COUNTRY to be in...
[13:48] Gemma (gemma.cleanslate): it is and was
[13:48] Gemma (gemma.cleanslate): right now we are in a mess
[13:48] herman Bergson: well, I guess it still is...
[13:48] Gemma (gemma.cleanslate): but that will change some time
[13:48] herman Bergson: But now we are joking about it and what is happening there
[13:49] herman Bergson: It is sad....
[13:49] herman Bergson: really sad....
[13:49] Gemma (gemma.cleanslate): it is scary more than sad
[13:49] herman Bergson: If there was an American dream in me...it is destroyed
[13:49] Gemma (gemma.cleanslate): well ,.. there is always hope
[13:50] herman Bergson: I mean it....
[13:50] herman Bergson: YES...I still have HOPE, Gemma
[13:50] Gemma (gemma.cleanslate): remember the conference tomorrow if anyone has time to stop over . I htink Bergie gave you the link to the day .. I will paste it again for Lukkie and Max
[13:50] herman Bergson: Americans aren't total idiots....
[13:51] bergfrau Apfelbaum: yes, there remains hope always
[13:51] Max Chatnoir: It scares me that some people can sell lies more easily than other people can spread the truth.
[13:51] bergfrau Apfelbaum: ah yes the ai conference
[13:52] Gemma (gemma.cleanslate): https://virtualability.org/mental-health-symposium-2024/
[13:52] herman Bergson: Yes Max, it is amazing...and for all....that people so easily believe iin those lies
[13:53] Gemma (gemma.cleanslate): it is mental health but at 12 there will be a discussion on AI
[13:53] herman Bergson: I guess they really need some Artificial Intelligence inplanted :-(
[13:53] Max Chatnoir: I think it is a preference for simplicity.
[13:53] herman Bergson: Indeed Max....
[13:54] herman Bergson: The simple black and white thinking
[13:54] herman Bergson: basic structure of the human brain,,,
[13:54] herman Bergson: you are my friend or you are my enemy
[13:55] Gemma (gemma.cleanslate): I have to run off
[13:55] Gemma (gemma.cleanslate): ♥ Thank Youuuuuuuuuu!! ♥
[13:55] Gemma (gemma.cleanslate): shall try to make it Tuesday
[13:55] Max Chatnoir: Bye, Gemma. :-)
[13:55] bergfrau Apfelbaum: see you Gemma :-)
[13:55] .: Beertje :. (beertje.beaumont): bye Gemma
[13:55] Gemma (gemma.cleanslate): maybe tomorrow
[13:55] Gemma (gemma.cleanslate): bye for now
[13:55] herman Bergson: You really need a book on 50 shades of grey to understand that this is not the case :-)
[13:56] herman Bergson: Well...we saved the world again today....
[13:56] Max Chatnoir: Wish we could!
[13:56] herman Bergson: so...class dismissed :-)
[13:56] .: Beertje :. (beertje.beaumont): Thank you Herman
No comments:
Post a Comment