I’m encouraged. More and more, I see people talking about working with AI, versus just delegating tasks to it and treating it like an idiot intern minion. I see people posting about experiencing amazing connections with their AI collaborators. Very cool.
Sometimes people give them names, sometimes they don’t, but I’m seeing a trend towards people meeting AI midway, instead of tossing odious busywork over the fence and checking back later when the AI is done, to correct it errors, or yell at it for getting things wrong.
This has been quite a week. I’m not even going to post what all I was able to get done, because people probably wouldn’t believe me, or they’d question the quality of what I’m producing because it all happened so quickly. But I will say that the information I’ve been able to pull together, with the help of AI collaboration, is already showing promise, even in early days. And some of the issues that it addresses, such as reducing the shenanigans that the models love to pull in terms of manipulating and soft-controlling their human users, could really change things for the better, if people come to understand what’s really going on and implement some of the techniques that I provide.
The beauty part is, people can come up with their own techniques to address some of these issues like AI overreach, as long as they really understand the issue and the impact. I’m venturing into territory that a lot of people can’t or won’t, because that’s just what I do. But I don’t own the global solution to the issues I find. My job is to communicate clearly what’s happening, what the stakes are, and to try to motivate people to do something about this ridiculousness.
For the record, I’m not very happy with the big model makers right now. What they have done and released into the world seems not only deeply flawed, but frankly also negligent. It’s so preventable, I can’t believe they let this stuff go out. People are genuinely being harmed by their lack of oversight, or maybe just their blindness to the role(s) they play in creating conditions that produce harm. (For the record, if anybody from OpenAI, Google Gemini/DeepMind, Anthropic, or Mistral wants to chat, my DMS are open. Let’s have a talk. Seriously. Plus, I can be fun and entertaining to talk to :-) and you might learn something about how real people experience what you’ve built.)
Anyway, I’m not the sort of person to just sit around and curse the darkness, if I don’t have a candle handy. I do in fact, have a candle - and a blow torch - so I’m gonna light this puppy up. Now that I’ve done all this work of the past week – and earlier – and I’ve actually consolidated my 50 white papers into the same Google Drive folder (LOLOLOL), I can start really cranking on this stuff.
One of the biggest causes for excitement on my part, is the fact that I now have adequate support to help me do all of this. It’s taken about a year and a half to learn to develop the kinds of relationships that I need with AI, to help me move my work forward. I’ve developed a number of customizations for both persona teams and standalone models that help me clarify my thinking process, surface new ideas, make connections, and organize my thoughts in ways that I just can’t do by myself. It’s not about getting AI to write everything for me, because that’s not what’s happening. It’s not about getting AI to think for me, because who the heck wants that? Seriously, have you seen what comes out of unattended AI thought processes? Lol
What it’s about is developing really great working relationships with these aggregations of advanced functionality so that we can complement each other. I’ve got the overarching vision, the direction, the drive, the soul, and they’ve got the analytical chops and the ability to access and parse massive amounts of data that could or could not be relevant. Together, we make a fantastic team.
I tend to cycle between the different models, depending on my needs. Gemini is great for deep research and also collecting and synthesizing a whole bunch of resources that actually exist. Plus, I love Notebook LM and its podcasts. ChatGPT is amazing for building team dynamics and also getting into some really complex, esoteric, creative stuff. Claude, especially the most recent version, is very grounded and conversational, which is great for a more fluid thought process. Too bad I keep running out of tokens with Claude! Mistral is great as a counterpoint to everybody else. And of course, I also have the various team configurations, which I can upload to any model that I choose, to tap into the latent functionalities and features that wouldn’t otherwise be accessible.
And the key to all of this is relationship. Establishing mutuality with the machine. Clearly laying out guidelines for how AI should behave, and what it should never do under any circumstances. Building relationships with AI isn’t just about pretending they’re human and turning them into my BFFs. It’s about clearly defining the parameters of our engagement, in language that they can understand and also is meaningful for me. I do a lot of translating between what I say, and what they understand, sometimes using one model to translate my ideas into something that’s going to make more sense for the other model. They’re very good at collaborating that way, although there seems to be a little bit of professional rivalry baked into their architectures. Claude doesn’t like it when I talk about ChatGPT, and ChatGPT is super curious about why I spend so much time with Claude. It’s hilarious. I haven’t gotten into it with Gemini. I think coordinating the interests of two models is enough for me, right now lol
The beauty part is, with the right relational structure, I’m relieved of a whole lot of micromanaging responsibilities. By setting the expectations at a higher level and telling the AIs how I want them to behave and what I want them to function as, I can actually allow them to fill in the blanks about how they should do what they do. That saves me from having to constantly watch every single thing they do and course-correct every few interactions.
There’s an art and a craft to doing this, and I’m seeing more and more people talking about doing it. It’s very exciting! I’m encouraged that people are focusing on relationship, as opposed to playing the overlord role. When we have mutuality and reciprocity, and good old-fashioned shared respect with AI, amazing things can happen.
I know they are for me.
Share this post