Think 2022 was wild for AI? Just wait

If you thought 2022 was wild for artificial intelligence, the people actually building these systems think you haven't seen anything yet.

On New Year's Eve, OpenAI president and co-founder Greg Brockman dropped a blunt forecast on Twitter: 2023 will make 2022 look like a sleepy year for AI advancement and adoption. Elon Musk’s one-word reply summed up the vibe for many: “Yikes.”

Moving faster than the benchmarks

Over the past year, generative tech went from an obscure research playground straight into the mainstream. Stable Diffusion shook up visual artists, Meta launched Cicero, and ChatGPT ended the year by signing up millions of users in record time.

Historically, AI researchers built benchmarks to measure state-of-the-art systems. Lately, new models shatter or bypass those benchmarks almost before researchers can finish writing the papers.

GPT-2 limitations vanished in GPT-3, and ChatGPT largely bypassed the bottlenecks of its predecessors. While ChatGPT still trips up over its human-feedback fine-tuning, GPT-4 is reportedly scheduled to land sometime this winter or spring, widely expected to jump another level.

Visual art took the first punch, but music composition, video animation, coding, and translation are lining up right behind it.

Concrete calls for the year ahead

Several clear predictions are on the table for this incoming wave.

First, image models are expected to reliably handle multiple characters and complex object interactions, fixing the most glaring flaws of current art systems.

Second, language search is hitting an inflection point. According to The Information, Microsoft is already preparing to integrate ChatGPT into Bing, swapping link dumps for direct conversational answers.

Third, developer workflows are flipping. Coding assistants like Copilot could soon see daily adoption pass 10% of software engineers, with up to half using them if pricing remains affordable.

Finally, personal conversational agents and digital companions are projected to leapfrog legacy voice assistants like Siri and Alexa.

Why it matters

For regular knowledge workers and developers, the automation frontier is no longer confined to digital illustrators. If tools like Copilot deliver on their trajectory, adopting them won't be an experimental hobby; it will be baseline workplace table stakes.

For the tech industry, Microsoft’s reported Bing overhaul marks an existential threat to traditional search economics. Replacing blue links with generated summaries directly challenges Google's core cash engine.

The concerning part is the pure velocity. Musk famously warned in 2014 that developing advanced AI was like summoning a demon, co-founding OpenAI in 2015 before departing in 2019.

We think the world would genuinely benefit from a slow, sleepy year to study model alignment, establish safety guardrails, and let society digest the shift. But looking at current competitive pressures and commercial incentives, nobody has any reason to pump the brakes.