And to do that, you need to: Understand, Audit, Disclose, and Control.
Lately, I have been deep diving on AI governance – the ways in which organizations, institutions, and governments can control the outputs and impacts of AI-enabled tool usage. It’s gotten me to work with three notions:
1 / what new tech mental models are required for these times?
AI isn’t a monolith, it’s a kind of analytical and productive tool, and there’s a material difference – like a 100x factor – between aggressive frontier model development, and a private instance of an LLM. There’s not a lot of nuance in the conversation. Some people are hard NO WAYs [22% of my ad hoc poll on instagram, itself a data black hole] and — trust me — LOTS of people are using AI without clear guidelines to transparently disclose. It’s a complicated technology with a giant ??? “we’re not sure how this works” ??? in the middle and lot of…
2 / Avoidance, shaming, and anger at both “the system” and “personally having to deal with this” looks a lot like money / capitalism feelings
I’ve been talking with people about topics they wish they didn’t have to mess with for YEARS. Trust me, I too wish I could safely skip understanding these systems: Money, Data Privacy, the next new technology … that’s not a safe choice for someone without independent economic means.
I’ve said it a thousand times about money: you don’t have to like it or approve of it to understand it. I think the same goes for AI and tech systems at large. There’s a ruleset you can learn, so you’re less vulnerable to the exploitations awaiting people who don’t understand and through which you can reduce unintentional harm and engage with discernment. Why not learn the rules?
3 / Self-lead governance is annoying, and necessary
Much like there’s a part of me that wants someone else to just make it all work out for me, and another part that deeply wants a world in which none of this is material, if I stop there and only revel in those desires, I know I will be operating in fantasy and not facing reality.
The fantasy: I can skip engaging in hard things, I can get a “gold star” by not touching the “bad” things
The error: focus on what’s too big to change as a release valve leading to skipping over what IS changeable.
The reality: focusing on finding and understanding specific things that can be controlled and decisions made about leads to taking actions to reduce one’s impact
For example: Focusing on how shitty capitalism is leads to avoidance, which leads to things like:
- not auditing your spending and ending up more in debt and paying more money to credit card companies or,
- not changing what your investments are invested in so they’re better [even if not perfect]
- spending unconsciously / consumerism doomerism
- not donating in line with your values
We really need guardrails that meet reality.
Without overwhelming ourselves, it’s time.
The video above goes into a framework you can – and I think should – use if you engage in any technology these days [and if you’re reading this on a computer, you engage].
Understand – AI is not a monolith, and developing LLMs, using a system, and using the outputs of an AI-enabled tool are different, specifically in both their impacts and what you can control in regards to your engagement with them.
Audit – take a real look at the AI-enabled search, bot “helpers”, chats, agents, buttons, and so on. Where is AI in your life, today?
Disclose – if an audit is self-disclosure, see if you can start to mention more when you’ve used AI [or managed to avoid it], to do what, as a way to get it out of the shadows. For those of us in business this is extra important from a data privacy and consent standpoint. I recently updated my data privacy and AI use policy on my website to this end.
Control – what you use, what you input, privacy settings, admin settings, what kind of trust you give to outputs, what bias you look for and add thinking to fix.
We also need discernment instead of shame.
There’s not an easy answer to the introduction of a technology that could be – is being – wildly misused and is misappropriating resources.
But.
That’s not a reason to avoid understanding and controlling it.