News

ByEagle

May 4, 2023

[ad_1]

Last month’s ideas to explore was much more popular than I anticipated so here I go with the second round! Some of the themes are in a completely new tangent while others trend on similarities in high level ideas (crypto/ai). Anyways, let’s dig in!

Representation and Uncertainty

(This one is longer than the usual Desystemize post; I recommend reading directly on Substack instead of in your inbox by clicking the title above. I’ve grouped it into sections for those who prefer reading in multiple sittings…

Read more

a year ago · 14 likes · 3 comments · collin

https://carcinisation.com/2020/01/27/ignorance-a-skilled-practice/

The word indexicality is one that makes your head hurt just looking at it, but I promise once you internalise what it means you start to view the world just a little bit differently.

A technical definition for it is given as:

Indexicality is essentially the degree to which local context matters for a given concept

Or put another way, the indexicality of a concept refers to how much context is needed to understand that word relative to where it is used. It’s kind of like a pointer. For example, if I said “look at that person”, it’d be an indexical statement because I need to be able to see or understand the situation to know which person is being referred to. “I prefer this”, “don’t do that” and “eat this” would all be more indexical statements.

Indexical statements point at specifics; non-indexical statements describe generalities. “The number 3 is larger than the number 2” is non-indexical, and true in a general sense. “My two cats put together are heavier than your three cats put together” is highly indexical.

Both articles go into why indexical statements are used, how certain groups of people reject to use them and most importantly, how they can go wrong.

You see, the problem is that when you have a non-indexical word which is actually an indexical word, that’s where the issues start to form. Everyone is pointing to a word/concept, but no one is actually speaking about the same concept or word. In emerging fields, indexicality errors run astray. The article I wrote last week about “users” in crypto is a perfect example. Everyone thinks that they’re talking about the same concept, but they can mean very different things. Once you start thinking with the lens of indexicality you realise the errors and opportunities that exist all around you if you can pay close enough attention.

The full read is quite a long one but if you can set aside 15 minutes of your day to read it you’ll come out of it feeling stunned. Language is powerful, but also mislead. Understanding the origins and surrounding context around the language you receive may lead to discoveries you may have never thought you’d come across.

To give one last concrete example, in crypto people refer to the term “DAO” to describe anything that doesn’t look like a traditional company. However, there’s a few flavours of DAOs that people are all simultaneously referring to:

  • An entity that attempts to avoid regulatory rules

  • An entity that aims to be a technocracy of some sorts

  • An entity that has no path to profitability but can sell dreams of the future

  • An entity that would like to be a digital-native company

All of these are wildly different in their intent but they’re all given the same word “DAO”.

I could go on about indexicality issues but I hope that you learn something new through this as a start!

Talking of indexicality, the word “consciousness” is one that I don’t know if we’ll ever be able to fully define but in this instance I’ll use it in a way to describe how one experiences & perceives the world around them. I can’t promise if this will make complete sense but I’ll do my best here!

The concept of demographical consciousness is basically thinking about how a group of people is currently thinking and where their collective consciousness is trending towards next. It isn’t meant to dehumanise anyone but rather come from a place of we’re all one, than we are different.

One way of thinking about this is really trying to empathise and understand the point of view of a group of people that may be completely different to your life and then understanding how the same information you perceive will be perceived by them given their life experience and background. There’s no perfect way to do this and you’ll always have your own biases creep, but even attempting the exercise is helpful to see where different demographics perception of a phenomena is now and where it goes next.

My favourite example of thinking about this has been the whole AI race. You have one demographic which is AI Creators that are cautiously releasing models and in an arms race to outcompete each other. Then you have a whole new class of adversarial users who you can see aggressively pushing the bounds of what LLMs can do. For example, the first wave of usage was primarily around how it could tell decent jokes, automate some basic tasks and show an impressive baseline of intelligence. Then you started seeing the collective consciousness of this group pushing to understand if the AI has political beliefs to then trending towards working around these beliefs and now to a whole class of “jailbreaking” where you avoid the limitations through advanced prompt engineering. Now the consciousness of this demographic group is centred around pushing an AI to places that the creators didn’t anticipate. With Sydney coming out, it took less than a week for this group to extend their existing thought process to new extremes. The AI creators responded by shutting down Sydney given the eerie responses it was starting to spew. While you have this arms race between these two demographics, you have capitalists attempting to max extract efficiencies in existing processes with AI. All while the average un-initiated AI people are thinking about how this changes their work situation. There’s a few more demographics in this race and many more springing up but being cognisant of the group’s state now and what it might be in the future is helpful to think about.

This is more of a freestyle idea I’ve been discussing with friends but do let me know what you think about this since I’d like to refine getting better at communicating more abstract ideas clearly 🙂

As if AI isn’t the talk of everyone’s conversations but I’m convinced that the highest leverage thing anyone can do right now is to become a prompt engineer. Learning how to “think in data” is critical. Prompt engineering is the last tool in the toolkit you need to get comfortable with. As large context sources can be imported and joined, knowing how to ask the right set of questions is going to be almost everything to turbo charge the amount of leverage you have.

The dream of having small, high leverage organisations is becoming increasingly real. We’re easily less than 1 year away from this happening in my prediction. You already have Bing, research papers and countless YC startups running at this problem full speed.

As more relevant contexts are being given to you, learning how to prompt engineer is going to give you even more leverage. Practice on the simple stuff while you can because slowly this is going to eat everything and we’re going to accelerate as a species together. There’s so many intricacies to prompt engineering that require time to achieve mastery with. One fun encounter that I had recently:

  1. Asking ChatGPT to write a script for me in a very sub-par way

  2. Re-prompting it multiple times to give me what I want

  3. Finally landing on the thing I was after

However, rather than leaving things there. I took it a step further. I asked “how would you have restructured my original prompt to be more concise and give the end result I have now?”. To my amazement, it actually told me how I should have written the prompt. Machines and humans working together yay!

I’m still at the starting stages of my prompt engineering journey but if you don’t start now with yours it’s going to feel harder and harder to catchup with every passing month to the point where a year from now is far too long.

Anyways, that’s a wrap for now with this edition of Ideas to Explore! As you can directionally see, the AI stuff is starting to consume my head space more given the wide design space. As meme-y as it sounds, combined with crypto we’ll have one of the most powerful branches of automated computing rise in history in parallel. Right now, the crypto people think that the AI is a new fad and the AI people think crypto is just a fad. As soon as those perceptions break the collaborations are going to become even more fascinating.

[ad_2]

Source link

By Eagle

Leave a Reply

Your email address will not be published. Required fields are marked *

PHP Code Snippets Powered By : XYZScripts.com