- Catch of the Week
- Posts
- 500 Days of AI: Part I
500 Days of AI: Part I
Is This It?
500 Days of AI
I’ve been wanting to take the time to reflect on the impact of AI on the industry at large, on how I view my work, and on my day-to-day development and admin workflow. Recently we hit the 500 Day mark of the release of GPT-4, which heralded in the current era of tools that are shaping the world around us. Creatives, and anyone working in a consulting / agency model is particularly exposed to the rapid pace of change in the space – if that’s you, you’ve already probably noticed your role or day-to-day tasks changing because of it.
Part I. Is This It?
Possibly more important than the release of GPT-4 itself was the launch of ChatGPT a few months earlier in November 2022. It ran on GPT-3, a model itself launched in 2020, but the simple productization of AI as a chatbot blew open the doors to wide consumer adoption of AI.
While I played around with GPT-3 and imagined the possibilities, it wasn’t super useful at the end of the day beyond a few specific tasks. In March 2023 I watched the GPT-4 launch keynote and signed up for a $20/mo subscription on the spot – upon launch GPT-4 was an order of magnitude more useful for a much broader category of tasks for me. Despite the actual gap in model releases being a few years, the closeness of those launches inspired a mix of fear and awe in those paying enough (but not too much) attention.
Within weeks there were people loudly expecting double-digit unemployment rates by the end of the year, the end of the web, the end of websites, and basically any technology hyperbole you can imagine.
This timing lined up perfectly with the crypto bear market, which left a lot of online charlatans looking for something to talk about and a lot of unused GPU power. Adjacent to the Twitter charlatans were the entrepreneurial hustle bros, boosted by El*n’s new Twitter algorithm with their claims of 6-digit MRR for their new AI startups they built over the weekend.
All of this was happened simultaneously with interest rates shooting up, VC investment dropping like a rock, and the creative industry going into a relative freeze compared to the previous 2 years.
A year and a half later, where are we?
Generally speaking, I don’t think we’re much closer to the doomsday scenarios being predicted in March-April 2023. We’re a few minor versions – mostly cheaper but not necessarily any stronger – of GPT-4 ahead, and economically things have bounced back (in the US at least) in a way that nobody would have expected, even if it’s all built on top of the Nvidia stock price.
A handful of AI tools have made it into the mainstream beyond ChatGPT – Claude Artifacts, Cursor, Perplexity to name a few – but high churn and high GPU costs have destroyed a majority of businesses that were lazily built on top of frontier models. Meanwhile Meta’s rapidly improving open-source llama models have destroyed any alpha in developing frontier models themselves – once believed to be OpenAI’s unique strength.
Besides the Facebook boomers cheering on generated images of Baby Jesus holding an American Flag, the societal immune system has developed a negative reaction against anything visibly produced with AI. In certain crowds, admitting any association with AI at all is akin to walking around in a MAGA hat in Park Slope. Advertising your new features as being “powered by AI” is as offputting as bragging about your new Cybertruck. At least temporarily, it feels like we’re at the top of an S-curve and we can catch our breath.
Meanwhile unemployment rates are floating up slowly, hiring has slowed dramatically, and overall the relationship we have with work, employment, and creative output has shifted in a way we’re still trying to understand. A recent study from Upwork that highlighted the disparity in experiences: the headline reads “96% of C-suite leaders expect AI to boost worker productivity, but 77% of employees report AI has increased their workload” – I’d expect this dissonance to shake out in one way or another quickly.
From a bird’s eye view things haven’t changed – people still want to build brands that people connect with, want to sell things to them. The economics around that and approaches and tools being used to achieve that are shifting as always, and those subtle shifts will accumulate into bigger ones in time.
What was acceptable in terms of speed and quality a few years ago will be far from acceptable a year from now. There’s a reason for hope there too – I heard someone recently describe it this way: before washing machines were invented, it would take several hours to wash clothes and they would be washed about once a month. When washing machines were invented, people didn’t spend less time doing laundry on a monthly basis, they just started to do it more often.
On a personal level, if anything I’m busier than ever despite these new tools in my belt and I don’t see that changing anytime soon. The major benefit, though, is a general decrease in toil, which in turns frees a lot of time and energy to turn the dial way up on quality. This spills outside of development, too – a lot more time consulting with designers on what’s possible/optimal, more business strategy/UX/Accessibility conversations with clients. For me this is great – I view myself as more of a product builder than “coder” so getting to step back and do this on a holistic level is much more satisfying.
While we might be on the top of the current S-curve, I’m not naive enough to ignore the fact that there are an infinite number of more daunting S-curves stacked on top forever – but it’s nice to sit and take stock here of the things that remain constant in work and in life as those will be the things we keep returning to in the future.
Interested in discussing a project or just want to say hello? Visit us at gonefishing.studio or write us at [email protected] 🎣