AGI This Decade: Reflections after attending an AI Hackathon in Silicon Valley
Artificial general intelligence is around the corner: here's what to do about it
After spending a weekend at an AI hackathon in Silicon Valley, I am convinced we will achieve artificial general intelligence (AGI) within 5-10 years. By AGI, I mean a computer system that can reason through life on par with the average American knowledge worker. Then what?
This post will cover my reflections after a weekend among the nerdiest of AI nerds:
Whose jobs are at risk
What society looks like in the post-AGI future
What can we all do now ahead of this inevitable outcome
Last weekend, I was invited to the AGI House’s hackathon, nestled in a quiet neighborhood near the SFO airport. The vibe (and the estate) was magical: dozens of Stanford Ph.D. students compared their achievements, while industry sponsors threw free GPU compute cycles (and tacos) around willy-nilly to attract developer attention.
All eyes were on OpenAI and GPT/ChatGPT. Creative prompt engineering continued to inspire and amaze. I haven’t seen this much energy since the days of crypto, but this time around, the applications are obvious and plentiful. Everyone at the event was working on something in their day jobs that would change how our world operates.
Companies that previously had nothing to do with AI are now tasking product managers to “figure it out asap!” Silicon Valley is buzzing.
Disclaimer: below predictions will likely all be inaccurate. That’s not the point. The goal is to open your eyes to what a potential future could hold so that you best position yourself today (see the end for my recommendations).
What will change
We are bringing the future forward, but what would that future look like?
The greatest “at risk” near-term professions are those whose roles can be framed as inputs and outputs — prompts and responses. In the simplest terms, this is how AI is most easily trained. For example, “our electric vehicle advertising client [input!] needs some new copy for their marketing campaign concerning climate change impact [output!].” With enough examples of client asks and agency responses, an AI can learn how to create the output independently.
This instantly affects the marketing analyst, the copywriter, and the talent manager. Any low mid-level knowledge work that’s routine enough to be learned within a month is ripe for replacement. Next in line are those working supporting roles in technical orgs: the unit tester, the project manager, the technical writer. Such input + output tasks can be learned by ai, too.
Those are just the first years. More dominoes will fall with time.
Artists and creatives aren’t safe, either. AI already generates images, videos, and songs.
Who will be left standing?
One obvious, high-paid portion of society will be building and maintaining the ai; this group will have its moment in the spotlight over the next several years, and later will reduce in size with time because the AI will learn to build and maintain itself more efficiently.
Less obvious: a second part of society will do physical tasks like construction; we are much further along with ai intelligence than ai muscle movements. Physical products have longer development cycles and robot actuators are far behind computing power (ai will be able to explain how to fix a toilet, but it cannot actually fix it). Skilled trade work will be celebrated in the next decade more than the last five (dm me if you are involved with trade schools in some way; I’m interested in seeding a Trade School 2.0).
Later, AI will learn to design toilets that are easier to fix using robots, but I believe that’s more than a decade away. There are higher priorities.
Purpose, Faith, and Politics
Much of society will need to be entertained and offered a purpose. Some of the AI Illuminati envision a world where all of us become creative artists. But that doesn’t make sense, either. Because AI can draw pictures, make music, and create videos.
Religions and new faiths will gain power. But, as with all faiths, they too reach limits because some humans are wired to question, rebel, and create factions… which leads to conflict. New AI-based faiths will appear - those may be more enduring because they aren’t based on the writings of one man walking into the woods and interpreting a magical holy stone. The AI will figure out how to attract and retain followers into its superfaiths.
This will start as humans standing at the pulpit reciting sermons aided by AI, then written by AI, and finally, the pastors themselves will slowly become replaced by digitally rendered versions.
Governments will slowly become usurped by AI. Initially, this will start with the highest stakes, highest dollar elements: elections. Deepfakes are the head fake. The real AI creep will be in developing campaign schedules, optimizing speeches, and, eventually, defining the candidate’s positions. We’ll slowly become accustomed to AI knowing what’s best. Eventually, we will laugh at how we used to let humans set our policies.
Universal basic income will make sense until we realize that we don't need so many people. Nations currently celebrate population growth because it creates GDP growth and geopolitical power. But that will no longer be true. Plus, humans create pollution which continually wears down the planet, and any semi-intelligent AI we view as our voice of reason won’t stand for it. Programs to incentivize rearing children will likely fade.
Some of us (likely those with more resources) will lean in and find our meaning in having children since that's the one thing robots cannot create. Until stem cell cloning is perfected and we can cultivate a human embryo-like we currently cultivate bacteria. This is probably at least 30 years away because it involves software + biology + physical actuators acting on fragile items.
Video games as a job
Our minds need purpose. Without it, we become depressed or hostile (e.g., what’s happening with America’s young men today). But, there will be less and less for us to do in the world. Aside from a small portion of us who leave society to live like 18th-century farmers enjoying the simple life, the rest of us will focus on faith and playing video games to give us meaning. These feed our need to compete, feel friendship, feel community, and feel love.
The video games are further optimized by ai to hit all dopamine receptors to the max. If Insta and TikTok are addicting, imagine what can be created with an all-knowing ai over the next decade.
Later, instead of joysticks and controllers, we connect the games directly to our brain stem. Bodies become less important, as we can actuate the same receptors for pleasure without needing skin and bones. How we look in the meta world is far more important than how we look in the meat world. Note: they might not feel like games; we might be convinced that our jobs are to move those joysticks and rack up points because this is how we gain economic currency in society… but… we today know what they really are…
Yes, but when???
None of this will happen dramatically or with some kind of big bang. It will happen slowly but surely, just like software ate the world very slowly but surely. You don’t remember the exact day when your accountant switched from using printouts to desktop software to cloud-based software. Similarly, you won’t notice when they let go of their junior staff and bookkeepers, instead using ai to close your books, input your 1099s, and recommend deductions for your situation.
What can we do?
In short, Zuck (metaverse ), Elon (neuralink), Sam (openai), and Larry (goog acquired deepmind, which started the latest cascade of dominos) are a decade ahead of us. Plan your stock portfolios accordingly.
Oh, and maybe have some more admiration for the Amish… they seemed to have caught on to the realization that a close-knit intergenerational community brings more lasting joy than scrolling ai-curated feeds. It takes courage to resist a flashier, more efficient society.
Cheeky quips aside:
Put in the reps. Using AI is a discipline like anything else. Start weaving ChatGPT into your life on a weekly basis. For any task, start with “how could I prompt ChatGPT”? Become the most proficient person in your role with the help of AI.
Make sure you and your kids learn the core concepts of coding. Data, logic, and front-end layers are the new building blocks of our society. Fundamental concepts are more important than syntax: ChatGPT can write code these days…
Prompt engineering is the newest high-paid consulting profession. Knowing how to query OpenAI to get the right answer in the right format (and knowing the full breadth of capabilities to know when to ask). Make sure your kids/nephews/nieces know this. It will become imperative for the near term while ai augments rather than replaces. The humans who know how to wield AI’s power have an advantage.
Stop mocking it. Don’t be the guy on his horse-drawn carriage mocking the model T. Best case, you will realize you were wrong. Worst case, your ego gets the best of you, leaving you behind
Stop and smell the roses. All things considered, you’re probably in a good place. Feel joyful in the elements that persist. Enjoy the sunlight beaming through the trees. The rain plopping in puddles. The close friend who has borne witness to your journey.
The workforce will evolve. Always does.
Any prediction about technology displacing jobs has to include a nod to the fact that humanity has adapted every time we’ve had a technological revolution. We have far fewer farmers today than we did 200 years ago, but our society still has plenty of non-farming jobs. It’s safe to say that we’ll figure it out this time, too. We will find something to do to create perceived value (see above: video games).
My goal with this essay was to offer a wake-up call and some suggestions so you can become empowered rather than displaced by AI’s future.
DM me with any and all feedback.