hckrnws

This is a really poor selection of events. Even if you restrict “AI” to mean “LLMs and diffusion”, this timeline starts years too late to cover the complete history of GPT models alone; I first became aware of LLMs a bit after Talk To Transformer launched in May 2019. Super frustrating.

Interesting, but... starting in 2022? That's not an "AI Timeline", that's an "LLM Timeline" or maybe a "Generative AI Timeline". But AI proper goes back much, much further. At least to Dartmouth in 1956[1] and if you accept some of Jurgen Schmidhuber's reasoning[2] then back to around 1800 (if not 1676).

That said, there's a fairly good "history of AI" page[3] at Wikipedia that covers a lot of the early material.

[1]: https://en.wikipedia.org/wiki/Dartmouth_workshop

[2]: https://arxiv.org/abs/2212.11279

[3]: https://en.wikipedia.org/wiki/History_of_artificial_intellig...

shouldnt it start from "Computing Machinery and Intelligence" (1950)

AI didn't exist before 2022. One day an LLM appeared and that is where AI began. /s

I think we should add

- microsoft build supercomputer cluster for OpenAI to train gpt3

- gpt-3 was regarded as moonshot and BERT was hot buzz in those days

- gpt-3 bots spotted on 4chan

- google fired a guy claiming AI was "sentient"

I’d consider all major players committing to support the Model Context Protocol a pretty big prominent event!

Timline? See, AI still can't spell.

How many times does the letter "r" appear in "timeline"?

I'd argue the modern AI timeline started with the deep learning revolution and the development of AlexNet around 2011.

AI is moving fast. To help track it all, I built an open-source timeline that maps out every major milestone in generative AI, from GPT-4 to Gemini, from Midjourney to Claude.

It’s updated monthly and covers the key developments across models, tools, and breakthroughs in a clear, no-frills format.

MIT licensed and fully open for contributions. Ideal for staying oriented in the chaos.

Are these prominent events? Some of these models are just trained a little longer or cooked with different data?

I think it would be better to list pivotable events, i.e. significant changes in architecture or approach, or a benchmark being surpassed. Otherwise this is just a list of models released by big tech companies irrespective of their importance.

Also I think I would prefer it be renamed from "Artificial Intelligence Timeline" when it's only since 2022 and AI encapsulates a hell of a lot of important work other than LLMs. I've been around long enough to see a few AI bubbles now.

Important events of AI are definitely missing anyway, for example when in 1956 some of the greatest minds in the field of the time set about trying to solve AI [1], only to realise it was far more complex than they imagined. It's only now we even approach addressing some of those original aims, some 70 years later.

[1] https://www-formal.stanford.edu/jmc/history/dartmouth/dartmo...

> Are these prominent events? Some of these models are just trained a little longer or cooked with different data? >I think it would be better to list pivotable events, i.e. significant changes in architecture or approach, or a benchmark being surpassed. Otherwise this is just a list of models released by big tech companies irrespective of their importance.

I don't think it's that easy to dismiss changes in recipes, data and training regime, tbh. There have been plenty of "huh, that's strange" moments with LLMs even without major breakthroughs in the overall architecture. The case of one model being really strong at chess while others from the same family (larger to boot) aren't is such a "huh" moment for me. Or the new "reasoning" thing, where it's obvious the new models are better at some tasks than the previous gen. Training steps is another interesting one, where two teams - one fine-tuning on 100k+ samples and one fine-tuning on 1k samples w/ 15 epochs and getting similar performance is also a good example. Or the (extremely readable) simple paper coming out of meta fair ~6 mo ago where they found that repeating some (5-10%) of the samples leads to much better generalisation. All of these are "huh" worthy, and I don't think we thoroughly understand why they happen, or why they sometimes happen and not other times (i.e. w/ 20+T training sets on massively scaled LLMs like the purported gpt5).

I’m familiar with the field of machine learning — I’m not an expert, but it’s quite clear that the developments in recent years have been very significant and very rapid.

There’s actually a tendency among experts in the field to underestimate the power and potential of current AI progress.

As the well known saying goes: 'Garbage in, garbage out.' I believe the same logic applies in reverse when extremely high-quality data goes in, an exceptionally good model comes out. At the end of the day, it's all about data, training time, and all those so-called 'details'that’s exactly what really matters

Thanks!

One nitpick from my side: It's not clear to me what the difference between blue and red dots are in the list...

This emphasizes significant events. As a model with a new architect, or an event that significantly influenced the industry

Reverse the timeline, it is more natural to start from now and scroll to the past.

Typo in the title, TIMELINE.

Uh-huh, but why is the starting date for this Feb 2022? Is it still work in progress, or are big discoveries in ML/AI prior to 2022 not considered part of modern transformer based LLM "AI" ?

I'd say these were quite a big deal too, in general. 1943! https://en.wikipedia.org/wiki/Multilayer_perceptron

Yeah that was my thoughts. GPT2 was huge.

Why are some bullet points red?

> 2022 - Present

lol