I’m no stranger to being a contrarian – so when ChatGPT exploded onto the scene a few years back, I resisted the urge to start using it for everything and instead decided to go as long as possible without. At the one year mark, I wrote about how my non-use of ChatGPT was going. The TL;DR – the ethical concerns related to large-scale, hosted LLMs around copyright, labor in training, climate, and monopolization/market capture led me down the path of self-hosted, personalized, on-device AI. As a result, I was exposed to the industry second-hand and first-hand in a unique way: I have extensive knowledge of AI generally, having worked with it since 2014, and every week I heard new updates about what GPT4 was capable of, but I hadn’t actually worked with the consumer side of things since an initial test in November 2022. That changed this week.

I didn’t actually change my perspective (yet) about ChatGPT and “Open”AI. The problems and issues I have still stand. But I’m enrolled in ‘Generative AI for Business’ as part of my MBA program at Columbia, and many of the assignments require using ChatGPT and OpenAI’s API, which means that as of now, I’m in enemy territory.

T”he enemy’s gate is down” – hacks.mit.edu

For the first assignment, I analyzed the differences between GPT3.5 and GPT4 for a personal branding exercise. The ability to pull real-time data from the web in GPT4 made the results more immediately relevant (of course) but I’ve become used to AI systems with a little more personality. Having built MemoryCache as my baseline, I still find ChatGPT to be a bit too generic (statistically average, if you will) and the GPT3.5 was actually a bit more interesting to me in some regards:

GPT3.5 – Empowering Innovation, Driving Change
GPT4 – Innovating for a More Open and Joyful Internet

ChatGPT: “Generate… a new tagline for the side [sic – should be “site”]

GPT4 basically nailed my work as the Ecosystem Development Lead in Mozilla Innovation, but that tagline is less about what I do, and more about what my organization does. GPT3.5’s tagline is more generalized, but it speaks more to individual motivation than the specific role that I play at this point in time. Which do you prefer?

In the coming weeks, we’ll be continuing to dig into ChatGPT and OpenAI API behaviors, which means that I have at least a month of poking around at the system-that-should-not-be-named. I have a lot of mixed feelings. I’m impressed with the capabilities that I’ve seen, acutely aware of the human labor that was stolen to make this work, and uncomfortable with the tensions at play between capital-holders and the individuals who stand to improve individual productivity with their tools.

I really hope that the transformation that AI brings is a 4 day work week for everyone and universal basic income.