Artificial Intelligence: Difference between revisions

Revision as of 10:26, 31 January 2025

ChatGPT is all the rage; stock market valuations of major companies like Alphabet (Google) fluctuate billions of dollars overnight due to perceived strength or weakness of the product's AI-powered features.

This page will capture some of the interesting points about AI and its use or relevance in Knowledge Management, MediaWiki, and probably some other tangents like deep fakes or politics.

One interesting essay I read was "Creative Commons and the Face Recognition Problem" by Adam Harvey. He describes how 100 million images from Flickr were used to train facial recognition systems using peoples wedding and vacation photos.

Understanding AI edit

An excellent introduction to Artificial Intelligence and Large Language Models (LLMs) is an article Large language models, explained with a minimum of math and jargon by Timothy Lee and Sean Trott - July 27, 2023

Tim Lee is a journalist with a master’s degree in computer science. The article is the result of two months of in-depth research. Co-author Sean Trott is a cognitive scientist at the University of California, San Diego.

Vectors are not just for graphics edit

SVG is cool for graphics. But vectors aren't just for graphics.

The Nordic Language Processing Laboratory, Language Technology Group at the University of Oslo, Norway publishes their research tools which help visualize how word vectors work in LLMs For instance, here's the vector for cat. (Click the link that says "Show the raw vector").

Biases in the Hive Mind edit

In Semantics derived automatically from language corpora contain human-like biases the writers show that AI "learns" the same biases we live. The biases are embedded in our language. TLDR; if a million documents with the word 'nurse' contain predominantly female people and the word 'doctor' in that same corpora contain mostly male figures, then AI learns that nurse is female and presumes a doctor to be male.

Google AI Studio edit

Google AI Studio is a tool that lets you use Google's Gemini API 'easily'. In other words, if you want to develop something using AI, this is a starting point and you don't need to even know how to program. But the current site is laughably ridiculous - offering sample prompts like

"Test if AI knows which number is bigger." (something a 5 year old knows)
"Get recipe ideas based on an image of ingredients." (like, Don't know what to cook for dinner? Take a picture of your pantry! If you don't know how to do meal planning, I don't think artificial intelligence will help much.)

So, if you truly want to build something with AI, you might need a programmer ;-)

At Google, they have two series of LLMs: Gemini (cookbook) and Gemma

@@ Line 1: / Line 1: @@
 ChatGPT is all the rage; stock market valuations of major companies like Alphabet (Google) fluctuate billions of dollars overnight due to perceived strength or weakness of the product's AI-powered features.
 This page will capture some of the interesting points about AI and its use or relevance in Knowledge Management, MediaWiki, and probably some other tangents like deep fakes or politics.
 One interesting essay I read was "''[https://adam.harvey.studio/creative-commons Creative Commons and the Face Recognition Problem]''" by Adam Harvey. He describes how 100 million images from Flickr were used to train facial recognition systems using peoples wedding and vacation photos.
 == Understanding AI ==
 An excellent introduction to Artificial Intelligence and Large Language Models (LLMs) is  an article [https://www.understandingai.org/p/large-language-models-explained-with Large language models, explained with a minimum of math and jargon]  by Timothy Lee and Sean Trott - July 27, 2023<blockquote>''Tim Lee is a journalist with a master’s degree in computer science. The article is the result of two months of in-depth research. Co-author Sean Trott is a cognitive scientist at the University of California, San Diego.''</blockquote>
@@ Line 18: / Line 14: @@
 == Biases in the Hive Mind ==
 In [https://www.science.org/doi/full/10.1126/science.aal4230 Semantics derived automatically from language corpora contain human-like biases] the writers show that AI "learns" the same biases we live. The biases are embedded in our language. TLDR; if a million documents with the word 'nurse' contain predominantly female people and the word 'doctor' in that same corpora contain mostly male figures, then AI learns that nurse is female and presumes a doctor to be male.
-[[Category:Artificial intelligence]]
+== Google AI Studio ==
+[https://aistudio.google.com/prompts/new_chat Google AI Studio] is a tool that lets you use Google's Gemini API 'easily'. In other words, if you want to develop something using AI, this is a starting point and you don't need to even know how to program. But the current site is laughably ridiculous - offering sample prompts like
+* "Test if AI knows which number is bigger." (something a 5 year old knows)
+* "Get recipe ideas based on an image of ingredients." (like, Don't know what to cook for dinner? Take a picture of your pantry! If you don't know how to do meal planning, I don't think artificial intelligence will help much.)
+So, if you truly want to build something with AI, you might need a programmer ;-)
+At Google, they have two series of LLMs: [https://deepmind.google/technologies/gemini/ Gemini] ([https://github.com/google-gemini/cookbook cookbook]) and [https://ai.google.dev/gemma Gemma]
+[[Category:Artificial Intelligence]]