Welcome to our regular Tech Tuesday, a weekly challenge where you explain one technical concept intuitively.
Pick any idea, technology, concept, formula, …, from any technical field —math, science, economics, politics, humanities, art— and share an article, note, thread, or any other written communication explaining it as intuitively as possible.
Link back here your contributions, and make sure to check all others.
Just couple of days I wrote about something interesting. Exact computation on the closed form expression of Fibonacci numbers which involves the golden ratio. We know that the golden ratio is an irrational number and any computation using it in computers requires using floating point data types. The problem with these data types is that they have limited precision and cannot represent the values accurately. And, as we perform computations using these approximate representations, we accumulate errors, which ultimately may result in incorrect results.
The trick involves expressing the quantities involving the golden ratio in the form a + b * phi, where a and b are integers. You can read how exactly this works in the article, it's a short read: https://codeconfessions.substack.com/p/fibonacci-numbers-2
I'm going to go back to the archives a bit since the topic of Bias in AI/ML keeps coming up and most of it ignores the fact that AI/ML is intentionaly coded, mathematical bias. This essay breaks down the three layers of bias in AI and helps us understand what to do with it!
Loved this. It is true that bias is often a catchall word thrown around without really understanding the implications of the different types of biases. I always say there's no unbiased learning. You just have to pick your biases.
Just couple of days I wrote about something interesting. Exact computation on the closed form expression of Fibonacci numbers which involves the golden ratio. We know that the golden ratio is an irrational number and any computation using it in computers requires using floating point data types. The problem with these data types is that they have limited precision and cannot represent the values accurately. And, as we perform computations using these approximate representations, we accumulate errors, which ultimately may result in incorrect results.
The trick involves expressing the quantities involving the golden ratio in the form a + b * phi, where a and b are integers. You can read how exactly this works in the article, it's a short read: https://codeconfessions.substack.com/p/fibonacci-numbers-2
This was really facinating
Thanks, Michael.
This was an incredible follow up, I learned a lot ;)
Me too. I think this was the first time I saw a practical application of rings and fields. Although, I think cryptographers do these things regularly.
This was a collaborative piece on the history of the Mechanical Turk I did with Devansh: https://goatfury.substack.com/p/the-turk
I'm so stoked that this is still going on! It's such a great place to share and discover work.
I really liked that one, especially the final twist ;)
"The elegance of the ruse."
I'm going to go back to the archives a bit since the topic of Bias in AI/ML keeps coming up and most of it ignores the fact that AI/ML is intentionaly coded, mathematical bias. This essay breaks down the three layers of bias in AI and helps us understand what to do with it!
https://www.polymathicbeing.com/p/eliminating-bias-in-aiml
Loved this. It is true that bias is often a catchall word thrown around without really understanding the implications of the different types of biases. I always say there's no unbiased learning. You just have to pick your biases.
Or as I like to say: "Everyone is biased, the key is knowing what your own are!"
Yep