Fascinating tool of Generative AI
This past weekend I did a self-directed weekend intensive bootcamp focused on Generative AI (GAI). The output from my weekend is my (very humble) first prototype of PIE Bot.
Here are my takeaways from that weekend:
Consider a powerful toolkit
A hammer is a simple object that can be used in a variety of ways. A skilled craftsperson can build unimaginably beautiful art installations. A skilled carpenter can build pragmatic structures for a wide-array of people and places. An unskilled hobbyist can miss a swing and break their own hand. A murderer can bash someone's brains in. A frat brah can open a beer. Same exact tool, many different outcomes.
Magnified to the nth degree, GAI feels like a self-propagating toolshed that keeps expanding and refining.
Wield power tools safely and wisely
I've worked in a handful of artist workshops, some more safe than others. In the case of GAI, I'm especially keen to understand the use cases, limitations, and considerations as early as possible in its development. In middle school, shortly after a class on how to look up references in a library (ha!) we had a class on how to Google effectively and how to triangulate trustable sources. Nowadays, the equivalent should absolutely be how to prompt smith for GAI and how to validate information through trustable sources.
And my voice-cracking peers who honed their skills at using the internet to retrieve and synthesize information are likely to be much further ahead in the modern era. I expect those who learn to use GAI now at any age will substantially outpace their cohorts in the short and medium term.
Dedicate a learning sprint to explore and unlock new levels
For all the reading and thinking I've done, I wanted to roll up my sleeves and try my hand at a passion project. Like millions of others, I've opened the GPT chat page and fiddled but I wanted to get a richer, more nuanced understanding. I learn best experientially and wanted to delineate hype from reality. So I drafted a very small project scope that is immediately applicable to my Thriving work, recruited a willing buddy, and carved out a weekend creators retreat (thanks for the support honey!) to focus on learning GAI.
Impressed and inspired
Despite all the obnoxious hype, I simply could not fathom how powerful GAI is already. Used thoughtfully and aligned with my ethical standards, GAI enables me to do a lot more, in a lot more domains, at a much higher level than I was capable of just a week ago. I've already used both text and visual applications at work, and had some fun playful interactions with the visual tooling at home. Even approaching base camp on a rapidly generating mountain range has inspired me to think and create in novel ways.
I have spent more time considering the text and language side, so that felt more natural and understandable to me. Given that I have not developed my visual creative capacities, and have devoted zero time to studying that side, I was most deeply touched by my short playing with Midjourney . And now I am infinitely more capable to make relevant, attractive visual assets as well as communicate and interface with professional artists.
When photoshop came online, it was a common refrain that the art of photography is dead. When smartphones started to get good cameras, photographers again moaned and groaned. In fact, these tools created new capacities, new forms of art, and made that art more accessible to vastly more humans. There are technological parallels in nearly every medium - GarageBand and more professional variants for music; iMovie and more advanced Premiere for movies; and so on. GAI is already unleashing a wave of new forms of creativity and we haven't seen nothing yet.
Excited and enthusiastic
Although already very impressive, GAI is really a toddler bumping into tables as it shakily learns to walk. If this much is possible already, I'm very curious and excited to see what novel applications arise in the short and medium term. Extra exciting is how quickly things are progressing. OpenAI first released ChatGPT in Dec 22, hit 1M users within 5 days (!!), and within 3 months we've seen substantial step functions of improvement plus a breakneck pace of industrial and consumer applications. A lot is going to change in a lot of directions within the next 12 months.
The positive applications are substantial and notable. We'll be able to organize human knowledge in ever more useful and accessible ways. Creative professionals will be able to make ever more interesting and complex art in a fraction of the time for a fraction of the cost. Knowledge workers are already cutting busy work and saving significant resources - time first and foremost - with equal or even better output. I know doctors who are applying these tools for precision medicine and say GAI recommendations reading patient files are as good or better than human doctors. Appropriately bounded and used, GAI can time, money, energy, and human lives.
And, the negative applications are just as or more substantial.
Uneasy veering on terrified
Starting now, I assume everything I read, see, watch, or hear to be generated partially or entirely by AI. Most of the time, that's just fine. But it irrevocably drops signal to noise in a subtle and intense way.
When barriers to entry lower, as was the case with Photoshop, smartphone cameras, and music simulators, it means a lot more people can make a lot more trash content a lot more quickly. The speed of GAI is so extreme that we will certainly see a mass proliferation of mediocre content that will be oozing out of every conceivable nook and cranny of society.
As I'm partial to written word, I'm especially concerned about the echo chamber of AI generating content by training on content generated by AI. As Dan Shipper has written , GAI can be a great brainstorming partner to improve writing skills . The vast majority of folks, though, will side-step the thinking and learning phases and just have GAI write m'eh content that sounds good enough. Just like TV and social media have eroded the average attention span and creative capacity, so too I expect GAI to further dilute average writing skill and imaginative capacity.
At worst, the proliferation of GAI citing GAI content creates a warped hall of mirrors that best serves malicious actors. After the "fake news" election cycle in the US, institutional trust is already at an all-time low. And the scariest thing is that GAI's hallucinations - the technical term for generated text that's factually incorrect - are compelling, well-reasoned, and often *sound* right. That means only subject matter experts can discern fact from fiction and, in a society becoming intellectually shallower with shorter attention spans, I doubt many will go to the necessary lengths to identify trustworthy sources then verify GAI content against fact. Amplified to the logical dystopian conclusion, society becomes ever more confused and entrenched in attitudes and behaviors shaped by nefarious intent. I've already seen my spam tick up notably, and convincing misinformation is right around the corner. If we thought Cambridge Analytica was bad ...
Overall, I came away from the weekend convinced that GAI is a powerful tool to address the first gap of Thriving. For that reason alone, and many more still, I must learn to wield this tool with skill and grace.
Also, there are significant drawbacks to GAI's impact on society. I have no idea how those can be mitigated beyond personally managing it's impact in my sphere of influence.
I will continue to learn and grow in this domain.