Letting AI In
AI IS GOING TO RUIN EVERYTHING
AI IS GOING REVOLUTIONIZE EVERYTHING
These two divergent views are dominating the public conversation. Where do you fall? I used to waffle myself, but I tend to fall in the middle and lean toward the positive now.
My Previous Stance
I have always been a lurker on the internet and focus most of my online activity toward a privacy focused approach. AI gave me existential dread and I felt like I was watching the collapse of society happening in real time.
Most of this was driven from the viewpoint that wraps up the internet (and especially social media) as a cesspool of the worst humanity has to offer. I disengaged from nearly all social media over a decade ago. I keep a few accounts active, but there is very little activity and I assume it will take more time and practice blogging before I am willing to become more active across these platforms.
As these LLMs began to explode, their ability to create slop threw my mental picture of the internet all the way to the negative extreme. The extreme world views are always going to be the loudest, and now they can accelerate their screaming? Boy did that paint a pretty gloomy picture.
In addition to the screaming, the biggest companies racing to get to AGI are run mostly by people so out of touch with general reality that I saw no way for this pendulum to swing toward the positive.
The Flip in the Script
Working with AI on my own during my masters degree opened the topic up to my curiosity against the media portrayal and constant noise of social media. Sure these tools (because that is what they are for now) can be used for both good and evil goals, and I can see myself starting to voice my own ideas of how to adopt them with continued skepticism. Most people look at the biggest projects today (OpenAI, Anthropic, Meta, Google, xAI...) and only see a giant barrier that is gatekeeping these tools, but there are loads of smaller to micro projects looking to bring these tools to the masses without the AGI race in mind at all.
These projects are looking to use "AI for good" and can provide some measure of hope against the apocalyptic views. There are also numerous engineers and product people looking to solve the challenges of safety surrounding the ever-increasing speed of iteration we are seeing with these frontier models. It is here that I have settled into a more positive outlook for our future.
I started out working with these models using tiny open source options in Ollama and keeping as much of locked down as possible (no direct file access, absolutely no account access, and no way for personal information to leak outside of my home lab). Slowly, as I built trust in the way these models work and how to extend their capabilities with custom code I have adopted a more open approach to working with the bigger companies.
I finally opened up my interactions with ChatGPT pro and Claude Pro (I still refuse to directly integrate with my personal accounts like gmail), but there truly are massively powerful systems that can be built with these platforms already. I have become fascinated with the two drastically different approaches OpenAI and Anthropic are taking while developing these systems, and recognize how dangerous an unchecked program can go (looking at you Elon and xAI...) or how miserable society or working could be (thanks Meta!)
Closing Thoughts
I recommend anyone interested (or terrified) by these tools ask all the questions that you want with anyone or any model that you want. Please don't take anything these tools output as gospel, and challenge anything that feels negative of icky. You do not need to have a deep tech background to help guide these projects for good!