The following below is with AGI (potentially possible by 2030 based on field experts) and ASI (TBD) taken into consideration, but not the main focal point as current AI models with finetuning can still be capable of accomplishing what is described below. Kudo's to Sam Altman for touching upon these topics in recent times... Artificial Intelligence epitomizes a pivotal transformation within the realm of digital labor. Its integration has swiftly shifted from speculative to imperative, underpinning a present-day reality that redefines our relationship with work and our perception of labor market value. The purpose of this article is to offer a lucid and methodical exploration of the extent to which AI alters the digital workforce, with a nuanced focus on the consequences and opportunities it presents across various sectors. In the forthcoming sections, we will dissect these industry-specific impacts with precision, offering a thorough understanding of the evolving digital labor landscape in the age of artificial intelligence.
When I first heard about modern AI chatbots and GPT in approx. 2018 with GPT-2, I had a guttural, almost primal, sensation that things were about to fundamentally change in the world. This was another advent of technological change that was akin to a fresh Pandora's Box unboxing (not the kind of unboxing video we were prepared for, I jest). I remember I discovered it when I was looking into new trends on the subreddit r/futurology. It was used in the context of a choose your own adventure story telling game. This is where you would give it prompts or a generic story line before the contextual prompt settings (a feature now called "Custom Instructions" in ChatGPT) were native to the applications interpretation of user input. I imagine they were just prefacing your input into their API access to the old GPT2 model, or the model was small enough where they could download the models locally and could influence the weights and biases for localized training.
Nevertheless, it wasn't until October 2022 that my genuine appreciation and esteem for the progress within the machine learning community truly began to soar. Since then, I've reflected on, considered, and anticipated the economic consequences that this emerging technology introduces. This prompted a compelling realization that I must strategize for an unpredictable future, one in which my current role may become obsolete. I figured that AI imposed labor cost disparity will likely expedite the adoption of AI in software engineering and other abstract concept based jobs, leading to significant layoffs and reshaping the labor market as well as the common expectations related to what is in demand in terms of human labor. If we look at what most bigger organizations love to use as a indicator of job performance (yes, the dreaded acronym, KPI), there isn't any feasible way that a human can compare to an AI when it comes to speed, accuracy, and overall productivity. Unfortunately I can't just duplicate myself, like AI can with distributed systems, load balancing, and additional compute FLOPS. Even if I could clone myself, and trust me nobody wants that, I still wouldn't stack up to the speed that these systems can operate at, nobody can.
The answer of what the future looks like is clear to me based on existing corporate work culture and cost cutting nature of organizations, It raises the question of how society will manage such a concentrated displacement of jobs without any clear answer for workers as to what's next in their future and how they keep food on the table and maintain a decent life.
Above is a picture that displays what GPT-3's tokens were calculated out at, and what it would cost to replace the coding capabilities of a Software Engineer, now granted GPT-3 does have coding issues, hallucinations, and bugs; but extrapolate the same concept out 5-10+ years and replace GPT-3 cost and coding capabilities with better models and cost structures. In the picture above Dr. Matt Welsh speaking to a Harvard CS50 class, indicates that based on the math he did using GPT-3's cost structure for tokens as a reference and with the daily cost of a software engineer's salary and benefits to be around $1,200, that the cost to generate a similar output through GPT-3 would only be around $0.12 per day - a difference of 10,000x. This dramatic cost difference was used to argue that large language models have the potential to greatly increase efficiency and lower costs compared to human programmers. Here is a transcript quote "The other thing is, yes, the robot makes mistakes. But those mistakes can happen incredibly quickly, to the level of speed where iterate, iterate, iterate, iterate, iterate, iterate, iterate is perfectly fine. You can say to the robot, you know what? This whole thing, 5,000 source files, 20,000 lines of code, whatever it is, blow it away. Start over, boom. Five seconds later, you have a brand new version of it. Try that with a live human engineer team."
He also stated:
"Robots are not going to take breaks. The robot is not, today, expecting free lunches and on-site massage. That could change. The robot takes the same length of time to generate its code, whether it's the rough proof of concept or the final production-ready code."
And provided the cost comparison:
"So if you do the math, then the total cost for the output of one human software developer on GPT-3 is $0.12."
As mentioned earlier, one mitigation strategy might be the adoption of Universal Basic Income (UBI), which would provide a financial cushion for displaced workers but that obviously comes with a myriad of ethical issues that boil down to what typically comprises of political view points, and no political view points are welcome here in this blog or in the comments. Moreover, the dramatic shift in labor economics will necessitate a surge in entrepreneurial ventures as displaced workers and other professionals seek new pathways to apply their skills and replace income that has been trampled by the adoption of an operational and financial executive. That driven narrative being that "Our AI provided product/model/service will provide you the essentials to cut costs and increase productivity and workloads for your business" and pushed by Big Tech influencers and trending vendor provided solutions at the largest of company's. That will of course, being the new and shinny object in town, skyrocket market prices and shareholder value for those companies just out of pure speculation if the product holds promising functionality, and will force most other board of directors to try and adopt the same methodology in their products to stay competitive... This is the real "trickle down" economics in a very literal sense, I jest again...sorta.
What ultimately will determine peoples ability to earn a living, the way I foresee things panning out, will come down to creative solutions to pain points. Wait..... that's what business is about now! Well yes, but the context will be so much different as people with creative solutions to problems will no longer be limited in knowledge access (presuming all these systems are kept at a reasonable price point or within public accessibility), and the viability of learning the required skills to bring an entrepreneurial startup off the ground becomes substantially easier to obtain and accomplish by ones self, especially in the computer interaction aspect of startups (which, lets be real, encompasses a big portion of modern businesses success criteria via marketing, sales, underlying functionality, etc etc).
The imminent transformation wrought by AI in the software industry is a microcosm of the broader changes to come. Policymakers, educators, and business leaders must work together to ensure that the workforce is prepared for this future, equipping individuals with the skills to thrive in an AI-augmented job market and trying to determine what skills will still be around. Preparing for this shift means not only rethinking economic structures but also embracing a culture that fosters continuous learning and adaptability in the face of AI's inexorable advancement.
That sums up my feelings for now on this topic...
P.S As always, if your feelings are different on the matter, as long as you have reasonable evidence to support your claims, I encourage you to respond below in the comments. I love to have an open conversation going as times change.
Comments