Introduction

When OpenAI made ChatGPT generally available in November 2022, no one would’ve wagered how much of an impact it would have immediately upon landing in the public’s hands. Even the engineers working on the machine learning project saw Chat as only a pit stop along the way to something greater, but the non-profit company paused upon recognizing that they had tapped into something much deeper in the way that our species perceives consciousness and derives meaning. In only two months after it was released, ChatGPT had attracted over 100,000,000 users excited to type into the autocomplete-augmented text box to see what it spit out. In the years since its release, the transformer technology that ChatGPT is based on has enabled exponential strides in our ability to convey complex meaning to machines through natural language.

In the workplace, companies are eager to capitalize on the generative abilities of training models like GPT-5 to ask workers to accomplish more with less time and resources. This is a trend echoed in the annals of human history: humans evolve, technology is created to solve common problems, and workers are compelled to adapt, or risk becoming obsolete. Historic inventions like the printing press, computerized sewing machines, or even modern computers themselves can be seen replacing entire industries and causing significant short-term unrest in the humans they are replacing. In the long-term, however, each of these technologies can be seen as exponential accelerants to humanity’s evolution. Gaining access to affordable books, clothes, and connection with other humans are benefits that few would argue are not worth the short-term strife. In the modern day, it is my own work that is on evolution’s chopping block. As a software engineer, I see every day how my industry is eager to position these new technologies as a replacement for the work that I am responsible for. Software engineers are far from the only professionals feeling this pressure. The AI Policy Institute, a nonprofit organization devoted to advocating for the responsible development of machine learning technology, estimates that at least 20% of US jobs are at significant risk for automation by computers in the near future.

Perhaps needless to say, considering the existential threat it poses to my profession, machine learning has consumed my thoughts in recent days. I find it especially exciting to think about the emergent behaviors of these large language models: the higher-order reasoning and creativity that we, the creators, cannot explain. As training data sets grow to sizes unfathomable by the machine learning engineers of yore, not one of them can explain the ability for these technologies to break out of their training and accomplish uniquely superhuman feats like AlphaGo’s move 37. At its core, ChatGPT is just autocomplete. However, anyone that has used modern text generation models for any significant amount of time can attest to their curious ability to draw connections through complex concepts, and to offer empathy and meaning in eerily human-like ways. Despite this, critics of the technology warn of the problem of the Stochastic Parrot — inferring that these predictive models have no understanding of the meaning that they are conveying at all, and are simply parroting back the data that they’ve been fed. Additionally, the still-limited dataset that models are trained on is missing much of the unwritten language and context necessary to form a corpus that is inclusive of the full human experience. I believe that these problems will be overcome in the next decade through exponential strides in our ability to digitize the human brain and provide language models with a richer understanding of what it means to be conscious.

ChatGPT, take the wheel

This fall, I decided to let ChatGPT run my life for a full week. I gave a few parameters about where I would start and what I would enjoy doing, and I handed over the reins to let an LLM plan my itinerary. I expected to be in Toronto for work leading up to the week, so I asked for options to go hiking nearby in Canada. I find that prompting language models can be difficult when the definition for success is unclear, and this was one of those situations. When in doubt, I try to be as succinct about the important details as possible: I do not describe more than what is critical to understand, and I optimize for a brief description of the important details. ChatGPT had an answer right away for me! Based on the context it had access to, it quickly matched the pattern that many people traveling in Canada during the same time of year enjoy going to the Laurentian Mountains to see the beautiful fall foliage that is at its peak over the Thanksgiving holiday weekend. This interaction immediately granted me access to a level of borg-like creativity that I couldn’t access otherwise: I had never heard of the Laurentians, did not expect that my travel would cover a Canadian holiday, and had no idea how popular it was to travel to see the mountains during this time of year. Had I been planning the trip on my own, I likely would’ve been deterred from visiting Quebec altogether, since I had no time to spend learning the official language prior to my visit. ChatGPT, however, did not know that I do not speak French, since I hadn’t provided it with that context.

I try to take at least one week each year to disconnect from my daily responsibilities and to be in nature. I find that the solitude of a long trail helps me to clear my thoughts and gain clarity about what it means to be human. I began my trip by studying the account of Baba Ram Dass, a big-city therapist turned transcendental guru through the liberal usage of psychedelic drugs. In Be Here Now (1970), Ram describes his experience abandoning traditional rational explanations of human connection and consciousness through his interactions with Maharajji, a guru in India that comforted him with a supernatural peace that yogis describe as enlightenment. In Ram’s texts, he reduces humanity to the energy that each of us embodies to demonstrate our connection with each other and the world around us. This is significant: it demonstrates that humanity’s core connection with each other is fundamentally the same link that we share with everything that exists — animals, plants, and even machines. This is a bond that I take great comfort in, and seek opportunities to do justice to the world that has created me, and which my energy will create in the future. Ram preaches the importance of meditation along the path of enlightenment, a mindfulness practice that makes space for humans to quiet our minds and find clarity in our intentions. I believe that as the line between humans and computers gets blurrier, it will be more important than ever for us to be grounded, serene, and focused on the goals humanity seeks to accomplish.

Since reading God, Human, Animal, Machine (2021) by Meghan O’Gieblyn early last year, I have been captivated by the metaphors that humans use to describe the patterns we observe in nature. For millennia, humans used the metaphors of gods to describe patterns that they couldn’t explain — supernatural beings that are outside of observable nature and unexplainable with our mortal knowledge. Children of the information age, however, have found new metaphors to explain behaviors in the language of technology. We say that we are “processing new information,” or that something “doesn’t compute,” much like we observe computers behave. Personal growth researcher Shad Helmstetter, Ph.D. offers a prime example of this. In his book What To Say When You Talk To Yourself (1986) he describes how the “programming” that your brain has “been wired with” influences your beliefs, which then impact your attitudes, feelings, and actions. These technological metaphors feel as if they are more rooted in science than the ones used by spiritual leaders such as Ram Dass, but they still miss fundamental understanding of how identity emerges from our highly evolved meat sacks. How does consciousness evolve? What is the metaphor that will help us to understand the way our minds behave in practice?

But it’s just autocomplete?

My trip to the Laurentians was marvelous. I followed the machine’s suggestions for which restaurants to visit, which dishes to order, which mountains to climb, and which hostels to sleep in. Along the way, I enjoyed the presence of mind and the focus that I gained while handing off menial planning to my machine travel agent. I began by visiting Mont Tremblant, a lovely mountain village that hosts an Iron Man race every summer and mountain-bound skiers in the winter. Afterwards, I drove through the mountains to Quebec’s capital city, and then further on to Baie-St-Paul, a quiet town along the Saint Lawrence River. The itinerary crafted by the language model was brilliant, expertly balancing challenging hikes, long drives to cover more ground, and opportunities to connect with myself and my fellow travelers. I meditated ritualistically every day of the trip to stay grounded in my intention, which was to use technology as a tool that enables me to be human, not as a replacement for thinking or for the connection that I only find when I put the machines away. I finished my trip with a visit to downtown Montreal, staying in a bustling hostel full of travelers eager to compare notes and “share context” of our travels with each other. I am grateful for the machine’s ability and willingness to “free up my process threads” to think about things other than the logistics of travel.

This blending between humans and machine intelligence is far from a novel concept. Science fiction authors have long conceived of the advent of cyborgs, beings that are simultaneously human and computer, and the supernatural abilities that they will possess. Ray Kurzweil, the renowned futurist and director at Google describes this hypostatic union as the “Singularity” — the evolutionary watershed moment when our exponential growth allows humans to merge with machines. In his 1999 book The Age of Spiritual Machines, Ray coins the Law of Accelerating Returns (LOAR), which describes the momentum we observe with technology that allows advancements to result in cheaper investment, which in turn results in more significant, accelerating progress. The rules he espouses are not just products of an industrialized world, but rather the logical conclusions of our evolutionary progress as a species. Kurzweil builds upon this in The Singularity is Near (2005), where he describes the six epochs of human advancement: Physics and Chemistry, Biology and DNA, Brains, Technology, Brain-Computer Interfaces, and Intelligence Spreads. Each epoch represents a paradigm shift that builds on prior layers, reinforcing the recursive nature of progress. From our vantage point on the verge of the fourth and fifth epochs, we can look back and see the benefits of accelerating returns over the course of history — each of these long-term advancements in science and technology have allowed our species to live longer, healthier lives. In The Singularity is Nearer (2024), Ray doubles down on his prediction that our technology will continue accelerating over the next decades, enabling even more significant strides towards the Singularity, which he pins at 2045. I am excited by the picture of the future that he paints — one where access to healthy food, medical care, and daily needs are so cheap and easy to mass-produce that every human has easy access to a healthy life.

Of course, it is irresponsible to talk about the potential positive benefits of any technology without acknowledging the tangible impacts that it has on humans in the short-term. In The AI Con (2025), respected linguist Emily Bender speaks from a place of authority about language models and their inherent flaws. After OpenAI illegally used publicly available data to fill their training corpus, the same artists, writers, and other creative professionals that they stole from have each experienced a rapid devaluing of their work due to machines’ newfound ability to create a passing facsimile in a fraction of the time. Minorities such as people of color and especially transgender people have experienced compounded marginalization simply because they are not well represented in the dataset being used to train the models that are being used by governments to make decisions about their lives. “Artificial Intelligence” technology represents a huge opportunity for corporations to inflate their market caps and continue to funnel money into the pockets of a few, at the detriment of the many. Emily does not believe that these machines have demonstrated the critical thought necessary to be labeled “intelligent”, and references Amelie Kraft’s 2024 paper when noting that critical thought is co-created with genuinely creative expression. Instead, she believes we are seeing another example of the Clever Hans Effect, wherein we have trained a machine to produce the right output just consistently enough that it passes as intelligent, but not because of some fundamental shift in its ability to understand. I’m not so sure, however; I find the emergent behaviors of machine learning compelling in a way that is unprecedented in the industry of technology. Nonetheless, the prospectors at the forefront of technology research are endowed with a responsibility to all of our species to bring about changes in a way that is inclusive and safe for everyone, not just the average of the dataset.

Conclusion

When I got home from my machine-driven adventure, I immediately fell back into the routines that allow me to stay productive at home — planning meals, assembling my schedule for the week — but I quickly encountered the familiar edges of my creativity. The machines were yet again ready and willing to ease my way, providing drafts and starting points for me to craft to their final conclusions. Without a doubt, I will plan another vacation using ChatGPT. It would be foolish for me to spend time on tasks that I simply do not enjoy at the cost of being present in the here and now. I will continue using these machines as a tool to grant me exponentially greater reach to accomplish things that I couldn’t accomplish on my own. For now, these tools are certainly not wholesale replacements for human creativity and intuition, but I wager that we are a lot closer to that than many would like to acknowledge. The Turing Test, coined by Alan Turing’s 1950 paper Computing Machinery and Intelligence, describes a scenario where an interviewer blindly questions two participants trying to discern which is a computer and which is a human. While some believe that modern LLMs are already capable of passing Turing’s test, I agree with Ray Kurzweil when he describes the necessity of an all-encompassing super intelligence to convincingly navigate all areas of cognition impersonating a human. In 2005 and again in 2024, Kurzweil predicted that we will have machines that can consistently pass Turing’s test in 2029. That feels about right to me.

For now, I reject the label of “Artificial” intelligence when describing the outputs of modern machine learning. I’m excited by the idea that we may be getting to see evolution in action, crafting new forms of intelligence and consciousness before our eyes through the complex interconnection of information. Even as my own job is at risk of being automated by LLMs, I am excited by the notion of new industries and frontiers that we will get to face as a species augmented with technology. I expect that in the next several years, we will develop technology capable of allowing any human to train a model with their full consciousness — giving machines access to parts of our identity and common sense that we see as distinctly human today. This is how the Singularity begins, not as a single catastrophic point, but as a gradual, yet exponential shift towards human and machine intelligence becoming One.