2023: The Year AI Went Mainstream
whatever you think about AI, you have to admit: 2023 was the year it went mainstream.
at the start of the year, most people had maybe heard of chatgpt. by the end, your aunt is using it for recipes and your boss is asking about "AI strategy."
let me recap what happened.
the major releases
march: gpt-4 dropped. multimodal. significantly better. expensive.
march (same week): anthropic releases claude 1. my first introduction to the safety-focused alternative.
july: claude 2 goes public. i start using it regularly.
november: claude 2.1 with 200k context. game-changer for long documents.
throughout the year: open source explodes. llama, mistral, and dozens of fine-tunes. AI became accessible to anyone with a GPU.
the cultural shift
this wasn't just a tech year. it was a culture year.
- every company suddenly has an AI strategy
- every school is debating AI policies
- every job is being evaluated for "AI automation risk"
- every conversation eventually turns to "have you tried chatgpt for X?"
the vibes shifted from "AI is a niche research topic" to "AI is everywhere and we're all affected."
the debates
the optimists: "AI will solve climate change, cure diseases, make everyone more productive"
the pessimists: "AI will take all jobs, spread misinformation, maybe end humanity"
the realists (maybe): "both outcomes are possible and which one we get depends on how we build and deploy these systems"
i think i'm in the realist camp. with a slight lean toward optimism because i'm working on this and want it to be good.
what surprised me
- how fast adoption happened (chatgpt to 100M users in two months)
- how much the open-source community delivered
- how quickly the "AI can't do X" claims got proven wrong
- how seriously companies started taking safety (at least publicly)
what concerned me
- the speed of capability gains outpacing safety work
- the misinformation potential (deepfakes, etc.)
- the fact that most AI development is happening in a handful of companies
- the hype-to-reality ratio in enterprise "AI solutions"
where i fit in
a year ago i was a student who was casually interested in ML.
now i'm a student with a thesis in the space, reading papers regularly, and actively hoping to work in AI safety.
the field pulled me in. and i'm not leaving.
looking ahead to 2024
predictions (probably wrong):
- multimodal gets better (video?)
- agents become a big thing
- regulation starts (EU, maybe US)
- more consolidation or more fragmentation (hard to predict)
- anthropic keeps doing interesting things (biased but genuine)
one thing i'm confident about: next year will be even wilder.
2023 was a lot. whatever happens next, we're not going back to a world where AI is "just research."