The AI Iceberg: Understanding ChatGPT
Analogies are useful for understanding complex ideas, and there are plenty of complexities for educators trying to wrap their heads around ChatGPT. In this post, I’ll try to explain some of the features of the chatbot and the model it’s built on top of. I'm deliberately avoiding any kind of analogy that represents the AI as magical, mythical, human, or godlike - we've seen enough of them. I’m not claiming that this analogy is watertight or that there is no better way to conceptualise […]https://leonfurze.com/2023/05/18/the-ai-iceberg-understanding-chatgpt/
I am a #ComputerScientist who is not a trained theoretician. As such, I find #DependentTypes not only intellectually stimulating (that is, challenging) but also practically attractive (albeit potentially, at present).
Using a dependently typed language, many vexing run-time checks that pollute the code could be off-loaded to the compile-time type checking, thereby leaving the logic flow clean, concise, comprehensible, and well, you know. Off-the-cuff examples abound: matrix multiplications without run-time checks for dimensional compatibility, sorted lists that need no run-time assurances of order, etc.
But these thoughts niggle me:
• BOUNDARY—Many modern, strongly, statically typed languages, like TypeScript, Zig, etc., have type systems that are Turing complete (unintentionally and with the attendant ⊥). It would be nonsensical to transfer all computing from run-time to compile-time. So, how do we, #programmers, discern this allusive boundary?
• ERGONOMICS—At present, most dependently typed languages and simply typed languages with Turing-complete type systems have rather ill syntax for type expressions. I am tempted to save Agda from this rather sweeping swipe, but even my beloved Agda's type-level syntax could use a bit of tidying up. If language designers insist upon foisting dependent types upon us, they ought to pay attention to the syntactic ergonomics of the type-level expressions as much as they do that of the value-level expressions.
Happy birthday to #mathematician & #computerscientist US Navy rear admiral Grace Hopper (1906-1992) who popularized the revolutionary idea of developing machine-independent programming languages based on English.
She began teaching at Vassar in 31 & got PhD (Yale) in 34. She found marriage & teaching less fulfilling & tried to enlist in the Navy, but was rejected then got a special exemption to 🧵
https://minouette.etsy.com/listing/748918489
#printmaking #womenInSTEM #histsci #artAdventCalendar #mastoArt
The man who created #CAPTCHA and #Duolingo
By the way I am very tired of Duolingo streak system. I want to form habits which are more important, like exercising my body, not regularly spent concentration on a language learning game.
In short, I think Duolingo is dubious.
Liam O'Mara IV, PhD
Dr Leon Furze
David Grayless
amen zwa, esq.
m bonbon
Ele Willoughby, PhD
Benjamin Carr, Ph.D. 👨🏻💻🧬
恐龍化石 Dino Fossil
Marco Hackney
