GTPeter training data #1: how to get inspired to start a huge project

GTPeter training data #1: how to get inspired to start a huge project

One thing I realised is that I'm bad at asking for or taking advice from myself.

I've had so many scenarios where, if I wasn't me, I would've told myself to look at things from a different perspective,  to not stop worrying, or to stop being a dick.

But I don't, because it's hard to listen to yourself when you're you -  you're too close to the situation to see things clearly.

I was thinking about how I could do better at this and it struck me that, I could ask ChatGPT for advice when I was struggling with random things to see what it could come up.

However, below is the kind of advice typical for ChatGPT:

Basically, just generic, shallow, not situation-aware advice.

I realised, that if I wanted to get good advice - from AI or anyone - it'd need to be deeply tailored to how my mind works - to understand how I think, what I'm going through, and to communicate in a way that I can easily understand.

The best possible advice for me would probably be from me!

So, I decided to try to train a model based on advice I'd give to myself if I weren't me. But this wouldn't be advice from the lazy, half-assed me - this would be me at my best, most thoughtful, and most consientious - like what OpenAI is going for when they build training data for their models for software development:

However, to do this, I require a huge amount of high-quality training data -  from all kinds of situations, different types of problems, and with varying levels of data and context.

To build up this up, I decided, over the next year, I'm going to try something: every time I'm struggling with something, I'm going to have a conversation with myself where I play both roles.

I'll play me, coming to myself for advice, while playing me, giving advice: GTPeter!

I'm going to record these conversations - via both voice and text - in a structured manner.

The goal is that I can use this to fine-tune a large language model in order to create an AI version of myself that can give me advice, challenge me, and share a detached perspective that's distinctly my own.

The goal is to have a version of me I can consult on everything I'm trying to figure out and understand.

I'm also going to record meta-data and reasoning related to the advice - that I'll hopefully be able to feed into GPT-3/GPT-4 to provide rich context it can leverage.

Below is one of the first conversations - which helped me get out of a mini-slump and start to make fast progress on a project I was struggling to get inspired on lats week:

A conversation with GTPeter - how to get inspired to start on an intimidatingly big project:


"I’m struggling to get inspired about an artwork I’m working on, and the tool I’m building to make it."


"Why do you think that is?"


"I have another project that’s been taking up a lot of my headspace and I keep thinking about that and don’t feel as though I have ‘space’ to be excited by something else."


In what way does that keep grabbing your attention?


"Well, every morning I check the notifications it and then get drawn down a rabbit hole of thinking about it. Then, it’s on my mind and I just keep thinking about it over and over."


"Well, the first thing I’d suggest is that you reclaim your morning - make it so that you don’t get caught up in thinking about the new project first thing. You can do that by locking your phone away somewhere, making sure it’s not easy to check things first thing (e.g. by blocking certain websites), and then making sure to not start checking it when the temptation arises as you become more aware.

The key thing is that you don’t default your brain into thinking about it. Then, you need to get yourself to take action to get inspired about the artwork project - so it fills the space that the other one is demanding. I’d suggest that you do this by doing something to get re-excited about it first thing - just as you would by checking the notifications for the other project - read things that reacquaint you with the big vision you’re working towards, or get you in the mood to think about it.

This could be something by listening to music, thinking about it, or reading your long-term plan. You should also consider a meditation where you think through your vision and how the action you need to take will lead to it. If you do that, you can build up enough motivation to at least get started - that’s half the battle!

Do you know what action you can take?"


"I guess not, any suggestions on how to figure it out?"


"Well, in order to build momentum, I wouldn’t think about what the most important possible thing you could do is but rather try to get a win that will help build up more and more inspiration - do you have something that you know you could probably complete in 1-2 hours that would feel inspirational to have done?"


"Yeah, I do actually!"


"Then, once you inspired, do that! To summarise:

1) Take action to ignore the distractions from the artwork project
2) Take action to rebuild inspiration by listening to music
3) Use this inspiration to tackle a part of the project that can get you a hit of inspiration quickly.

Does that make sense to you?"




"Then go do that! Make sure to get a good night’s rest in advance - go to bed a while before you need to sleep and read a book to help make sure this happens. You should also do all your morning habits properly first. You choose to do them because you believe that they’re good for you - now is no different!

Best of luck!"


"Got it, thanks!"

That's it! I'll share more here if I think they're generally interesting!