Michael Gingras

Interested in lots of things, but mainly -- crypto, browsertech, and interfaces.

2023 Predictions

January 2, 2023 web crypto ai vr


  • ML: machine learning
  • DALL-E: Tool that uses ML to create images from text
  • LLM: large language model, what chatGPT is
  • ChatGPT: a bot created by openAI that is able to answer most questions and can write code, poetry, stories, contracts, etc. Something you really need to experience to understand.
  • VR: virtual reality


ChatGPT / DALL-E excitement will begin to burn out. The AI “improving every week” cycle dies and people realize it’s harder to actually incorporate AI into existing products and workflows. You can’t change entire industries overnight. LLMs are no longer just a toy — there is real value to using them, but they are not going to replace entire industries anytime soon. The last 10% will be just as hard as the first 90%. ML is the new web3 and will thus attract the same amount of gold-rush opportunists flooding the market with low effort products. ML is still in it’s twitter hype bubble… even friends who are in the tech industry but not as serially online as I am don’t know what ChatGPT is. It’s easy to overestimate the influence of certain things when you only hear about it from one source (twitter.)

When ML does eventually start to take jobs, “white” collar jobs” are actually the first to be automated in mass. The easier it is to accomplish a given job remotely, the more likely it is to be automated. Knowledge work is easy for LLMs, and digital artifacts with zero marginal cost of production are ripe for ML to replace. “Hybrid” professions like law are somewhat at risk, although the fact that it is still crucial for lawyers to represent their clients in person means it is unlikely a bot will totally replace the job. “Blue collar jobs” or anything requiring physical labor that would be difficult to create a bot for are the last to be automated. While the nature of the work is repetitive and perhaps easy to teach to a computer, there is a high up front cost to creating the machines themselves that are capable of performing labor in the “real world”.

Specialists are more in demand than generalists. AGI does not mean superintelligence. It means artificial general intelligence. It might as well mean average general intelligence, because even an average level of intelligence would be enough to disrupt entire industries. Consider the hypothetical — “how much more intelligent or productive could a human become if time moved 1000x slower for them, and they were able to collaborate with 1000s of copies of themselves?” It’s easy to assume the answer would be “much more intelligent”. Well, this is pretty much how LLMs operate. They are trained on all of human text, so they have the same knowledge as humans do, but they can “think” 1000s of times faster than we can, and you can cluster 1000s of instances to communicate with each other. So even an “average” level of intelligence for a LLM would likely be huge. Theory aside, it is the specialists who are more capable of taking advantage of LLMs than the generalists. LLMs will be able to automate the entire breadth of what a generalists would be able to do. Specialists however understand a specific domain extremely deeply, and can guide and work with a LLM to solve problems a generalists would not know how to begin with. It will be much more adventagous to specialize in a certain domain than to try to compete with an AI at teaching yourself as much as possible.

Possible counter: generalists are better are thinking “big picture” and could perform better as the “ideas people” capable of working with LLMs to create products.


In person learning communities will thrive. Yes it’s true that you can learn anything you want to learn online, and this is becoming increasingly easy with tools like ChatGPT. But such tools cannot and will not ever replace social/emotional learning. Especially for young kids, there is value in face to face experience. Learning pods will thrive as a third place for kids to socialize outside of remote learning environments.


Nouns will continue to be the biggest project in the web3 space with nouns builder proliferating the pattern. As the daily “revenue” continues to depreciate, there will be growing concern over spend. Gone as the days of effortless requesting 40K a month for a dev salary. At least one proposal will be made to experiment with ways for nouns to generate it’s own revenue outside of the daily auction. There will be discussion and growing concern around how owning a noun can be considered an investable asset aside from speculating that the asset itself will appreciate. The asset itself may pay dividends off of proposals that it funds.

Lil nouns goes through growing pains and faces the challenge of 10k + members all trying to control a treasury. The group continues to brainless buy nouns, despite immense potential to be considered the “idea launchpad” of the nouns ecosystem (trialing small projects in the lil nouns ecosystem before requesting larger proposals from big nouns ala prop lot.) As a result, there is less and less engagement and a small pool ends up in control of both majority vote and in return a stake in 15+ nouns. This causes lots of controversy.


Twitter doesn’t go under despite everyone bitching about it (on twitter!)

No meaningful movement in VR. Meta is wrongly trying to invent the infrastructure of the metaverse rather than creating the engaging experience first (which would be a better approach). The concept of a metaverse is inherently social, and needs to be a “fun” place to hang out. Until meta (the company) understands that a better investment would be to create (or acquire) something with the virality of minecraft or roblox, VR will continue to be a dud. VR targeted at the professional sector (working in VR) is misguided and will not succeed.