CertAIntly Urgent
I'm sure many of us has seen a version of this chart, which shows how technologies get adopted over time:
Source: Craig Chelius
I've historically been a "late majority" kind of guy in terms of tech adoption. I like my trusted approaches to doing things, and have always been skeptical about the shiny new technologies. Let others figure it out before I dive in - if the technology is really worth its salt, I can always hop on the wagon later.
This has been a useful approach for me as it's enabled focus in my life on the things that matter, in a world that constantly cycles through new "in" technologies that are supposedly going to revolutionize the world, but that eventually, don't. Crypto was supposed to change the world of banking and money as we know it, but has until now remained a playground for speculators and gamblers. VR and AR was supposed to change how we work and play, but we are still dealing with overly heavy and expensive headsets. The jury is still out on these technologies - but I'm glad that I spent little time trying to understand them as it would've entailed taken away from other things that matter more to me.
The reality is that revolutionary technological change typically happens slowly and steadily. Innovators and early adopters go through the pain of trial and error as they try to find the right markets and products for the innovative technologies they are evangelizing. And sometimes, these technologies never end up finding the right markets or products, and die off in the common consciousness - resulting in "The Chasm" in the figure above. And by withholding involvement until products that are directly relevant to us come to market, we retain our time and attention for the other parts of our lives and careers. Not a bad trade.
Gen AI feels different though. Unlike other game-changing technologies like the smartphone and television, there feels to me like there's more of an opportunity cost to not being early Gen AI for a few reasons.
First, the substantial productivity benefits that likely accrue from Gen AI tech makes it an extremely economically useful tool that feels like an eventual "must-have" in any individual's life and career - in a way that the television did not given it's focus on providing entertainment, which is inherently unproductive from an economical perspective. We are already seeing substantial progress made in a variety of fields - enhancing software engineers' ability to code and potentially even replacing them, identifying novel drug candidates for previously uncurable diseases, among others.
Second, the seeming generality of Generative AI - i.e. applicability across many if not all industries - means that it is likely to touch and impact all of our lives and jobs, even if some are impacted more than others - in a way that other more "specific" technologies like genetic modification and aviation did not despite their substantial impact on the world.
And finally and potentially most importantly, the importance of the human / technology interface, i.e. the importance of knowing how to interact with the technology to maximize the benefits of the technology, which seems very high for Gen AI given the importance of knowing how to ask the models for what you want - an element that other technologies where humans are more "passive" users and products are more "cookie-cutter" and "user friendly", like the smartphone or the car, do not require. In the same way that our ability to make full use of Microsoft Excel is highly dependent on understanding the features on offer and the capabilities and limitations of the application, I wonder if a keen understanding and familiarity with Gen AI will be key to maximizing its usefulness. And this keen understanding and familiar will take far longer to develop should an individual delays their exposure to Gen AI.
As such, Gen AI feels to me like an essential technology to get in early on given failure to do so means missing out on what is likely to be a substantial productivity-enhancing tool that will almost certainly touch all jobs and careers given its relevancy across industries. Furthermore, despite the costs of being early, first movers likely could have an eventual first-mover advantage in terms of a deeper and more fundamental understanding of how to utilize the technology, which will be essential in how much value we derive from the technology.
It reminds me of the personal computer revolution. Massively productivity-enhancing, relevancy across many if not all industries, and where knowing how to use it and whatever derivative industries it inspires (e.g. how to code software with computers / how to effectively use spreadsheets) was key. For a while, the personal computer split the world into two binary halves - those who knew how to use it, and those who did not. And for those who leaned early into learning to code, or learning to use spreadsheets, they developed a competitive advantage for themselves that enabled outperformance, at least while others fell behind.
Now, this competitive advantage likely faded over time, as the share of the workforce that were computer-literate grew, as more of the existing workforce took computer classes and the workforce with each passing year included more younger entrants educated on the computer in school. However, the important distinction here is that the later adopters of the computer were not trying to gain an advantage, rather, they were trying to reduce a disadvantage, and close the gap. Moving early allowed a leg up on others in a crucial and valuable technology, while the late movers are stuck playing catch up. And even when they did catch up, late movers at no point gained an advantage - they were just trying to catch up. Yes, they got more time early on as early adopters figure out and trial and error the technology, but it was a poor ROI (return on investment) on that saved time given the opportunity cost of missing out on the massive benefits that being computer literate eventually enabled. And even when they did catch up, at best they put themselves back at parity with their peers and competitors, which is really just back to where they started, while their competitors in the meantime enjoyed the advantage of being more efficient and productive at work, or the ability to build and/or use new products and services enabled by the technology before anyone else did - since they had a differentiated skillset for a period of time
I also suspect that in reality, early movers were able to sustain their early advantage over time, even when later movers attempted to close the gap. This likely came either via 1) having early momentum, which enabled rising up the ranks faster for a number of years and opening a gap on late adopter peers that in reality is hard to close given the relative scarcity of senior roles, or 2) assuming continued investment in deepening understanding of the technology, a persistently better, more nuanced and foundational understanding of the technology than peers, which like a chef who is one with his pots and pans or a marksman who has a bond with his rifle, enabled a better ability to wield the computing technology that in turn enabled better results.
So yes, I have come to the belief that regardless of our industry or occupation, it's essential that we spend the time to experiment and understand Gen AI as soon as we can, to ensure that we develop a better understanding of how we can wield it effectively as a tool in our professional and personal lives. For not doing so will likely come at the expensive opportunity cost of falling behind our peers and competitors and leave us having to play catch up. But how do we do that? Especially for those of us, me included, who have little to no foundational knowledge in computer science / AI, how can we best approach this? As always, I don't have the answers but I have some hypotheses that I'd like to put out there, try out myself, and continue to evaluate over time.
First, this is obvious but it's important to keep abreast of all Gen AI developments that are directly (or even indirectly) relevant to our lives and work. Is there a new product being released that purports to replace essential parts of our jobs? Is there a competitor business that has built AI into their workflows to increase productivity? Is there a tangential industry with a cool new tool that could easily be adapted for our own jobs? Not keeping up with developments, which is easy to do given all the other things we have to deal with in our day to day lives, would immediately put us on the back foot as it prevents us from being offensive in identifying opportunities to build Gen AI into our jobs. One of my tasks for the week is to find a number of newsletters / substacks to follow to track developments in the field, and I'll share a few when I find them.
Second, I think it's essential that we spend some time understanding the fundamentals of Gen AI. What it is, how it came to be, foundational technologies and advancements that enabled its proliferation. This deep and foundational understanding will help us better sift through the volume of content that will start coming at us from news sources, friends, family - as everyone will soon, if not already, have their own opinions of what Gen AI is and will be. The ability to sort through the noise will be valuable given the uncertainty associated with how Gen AI will develop over time. And it is only through doing our own primary research of foundational pillars of Gen AI - academic papers, podcasts from leaders in the field, as examples - that we can better build out our noise filters and develop our own understanding of Gen AI. Again, I hope to distill and share some of this work as I dig deeper in the weeks ahead.
Finally, I also suspect that experimenting with existing products - even if they aren't fully formed - will be valuable. I've been resistant to paying the $20/month for ChatGPT Plus, given the free ChatGPT option, but the opportunity to play around with the more powerful GPT-4 model, as well as image generation AI tool DALL.E, is a valuable part of experimentation. Similarly, I intend to pay up to try out Microsoft's Copilot, which has various email and Powerpoint productivity enhancing tools (including building an entire slide deck for you from a single word document), along with any other Gen AI products that are available. And of course, as I experiment and test the limits of the products, I will try to actively make thoughtful and rational decisions to subscribe / unsubscribe based on the value each of these products provide to me, while also trying to figure out the best way to optimize my work and life. Will hopefully have a few posts on these endeavors here in the coming weeks and months.
Despite all this thinking, I can't completely shake the fear that by trying to understand Generative AI this early, I will end up wasting a bunch of time and energy on something that will end up being hype - I've seen it happen to others too many times. I'd hate to be that person constantly chasing the "shiny new thing" since that never really works out. That being said though, I do believe that it is still the right thing to do, given my conviction about the strong probability that Gen AI ends up being a paradigm-shifting technology for the world. And ultimately, even if I'm wrong, the work I'll be doing - if I do it right, that is - will help me realize that my conviction is misplaced, which in turn will enable me to de-emphasize sooner rather than later. It would still have been a worthwhile endeavor then.
Watch this space!
The Quickest Reader in the World
Wanted to leave this chart here as a teaser of all the posts I'll write on Gen AI in the future. Quite mind boggling that GPT-4 has already read almost all of the words ever written by humans, which is both an indication of the reading speed of these models but also perhaps an indication of how inefficient these models are at learning (for now)? If it took reading that many words for ChatGPT to be able to hold a decent conversation with humans, certainly the models can't be that effective learners?
--
That's it for this week! Posts might be shorter from here on out as I'm trying to find the right balance between creating content and also ingesting content (i.e. reading), which I haven't been able to do enough of.