Discussion no.07, Everything is a Spec

Link to the Viewing

(the other videos below are also strongly recommended)

Introduction

We have been roughly alternating between "normie" or normative, canonical, or conventional UI/UX and computing history and readings and work that is critical of that history and understanding. In some respects, this is relatively easy to do as some of this history approaches being a century old. We have a lot information with which to be critical or confirm it. This is not the same with the current context of Artificial Intelligence. As such many people are fighting to be on the cutting edge of this practice.

This talk is by Sean Grove, who works at OpenAI (who makes ChatGPT). He talks about something called "specs" with the presumption that prompt engineering is "dead." Please watch this and note your thoughts and reactions, but I'd like to talk more about how it relates to what we've previously talked about and start to think about what the future might be, and how people with power can or cannot dictate this future. Additionally, language is a tool that can be used to this end.

The above video talks about many of the players in the AI space (Elon Musk, Sam Altman, Peter Theil, Mark Zuckerberg, etc.) misinterpreting the sci-fi and fantasy from which they borrow for their programs and services.

  • Palantir (the company) is named after a tool that Sauron and Sarumon (villains) use to remote view.
  • Grok (X's AI) is from Robert Heinlein (a sci-fi author) meaning "to understand."
  • Metaverse (the new name of the company that owns facebook and Instagram) is a term from Neal Stephenson's Snow Crash

If the future of coding is "spec-ing" or "prompting" how does that relate to the history of coding (which was originally more secretarial)

Nerd Terms

The talk is aimed at developers, so there are a couple nerd terms:

  • binary → programmers generally take source code and build or compile it into a program. This is historically referred to as a "binary." This is because code used to be turned or "assembled" into machine code and then actual 1's and 0's (binary) that a computer understands. That being said the term binary might be used interchangably with an "app" or "program" more generally.
  • lossiness → here, Sean uses the term "lossiness" to talk about the relationship between code (source) and the binary (program). This is a term generally used with compression of files and information, as you typically, lose some kind of information or resolution with this process (think of a low-quality jpeg)

Things to think about

  • Do you experience sycophancy when you use ChatGPT or simiilar? Do you attempt to get the AI to alter how it communicates with you? Does it work?
  • Think more about what a "spec" or formalizing structured communicated means vs "code" as ways to achieve goals. This could be homework, a job, creativity, etc.
  • Think about your experience with ChatGPT, and code itself. How does it relate to or mirror the practice of (spoken, human) language and writing proliferation to you? Think about technologies like the printing press, the ballpoint pen, or even caffeine and how they effected literacy and literacy rates. What is chatting and prompting like in comparison (for you)? Do you feel it helps you? Simply does the thing it is supposed to?
  • What does it feel like that these technologies ("AI" as we call it today via LLMs, Machine Learning) are evolving so fast and so widely, with such an unfathomable amount of funding?
  • Do you have any scepticism about this person's findings and thoughts if they work for OpenAI?

What to submit

Please provide a .PDF, .TXT, .MD, or .RTF file that evidences your reading and comprehension of the material