Using characters instead of tokens because token size can be changed, and characters are what humans actually care about. If they advertise a context window in tokens, I'll convert it to characters at the average rate of that tokenizer on representative human text.
Something "cheaty" doesn't count, it has to be, say, at least as smart as GPT-3 on similar inputs.
Related questions
🏅 Top traders
| # | Name | Total profit |
|---|---|---|
| 1 | Ṁ149 | |
| 2 | Ṁ59 | |
| 3 | Ṁ24 | |
| 4 | Ṁ24 | |
| 5 | Ṁ8 |
Nice time capsule of a market... A year ago GPT-4 was 8k or some ridiculous-expensive 32k that wasn't even that good. 76% in April 2023 is not a strong vote of confidence given "end of 2028" timeline. That probability means "speculative, far in the future, but maybe somebody will get it working".
Today I put my codebase into a 1 million context LLM on my Mac to ask it questions, write code, and use tools. You can just download them and run them on any Mac. It's not some heroic feat like "76% by end of 2028" makes it sound.
@IsaacKing Gemini 1.5.
“We’ve been able to significantly increase the amount of information our models can process — running up to 1 million tokens consistently, achieving the longest context window of any large-scale foundation model yet.”
1 million tokens > 1 million characters.
https://blog.google/technology/ai/google-gemini-next-generation-model-february-2024/#sundar-note
Let's keep in mind that LLM's are not purely generative in nature and don't have to be based upon a GPT or any pre-determined architecture, they are merely a probability distribution over sequences of words. So as written, this question has a wide interpretation. I would almost advocate for narrowing down the definition further to make it more interesting.
@PatrickDelaney It should probably be "equivalent to GPT-3 on some benchmarks", otherwise a random tree search or markov chain would qualify. (Well, a "large" markov chain)
Maybe the "Evals" repo that was introduced with GPT-4 would be a good one? openai/evals: Evals is a framework for evaluating OpenAI models and an open-source registry of benchmarks. (github.com)
@IsaacKing would be much easier to use some kind of external knowledge/state store rather than a massive context window
@jonsimon No reason? How do you know that a-priori, do you know every industry? Do you know every possible architecture that might come out in 6 years? 1 million characters is 600 to 800 pages. I could imagine existing some odd esoteric application for that. If it was 100,000 or 1 million pages, increasingly less likely. But what about...government/intelligence summarization?
@IsaacKing I don't want to argue too harshly because this is all metaphors we're talking about but...humans do not have an unlimited context window analogous to an autoregressive GPT's context window...right? You would have to be able to, "remember," e.g. "tokenize," every conversation you ever had in detail, e.g. including every word to fulfill the condition, "*unlimited* context window." I think you might mean...humans can have a context window that stretches back selectively for years if not decades...something like that? To me, "unlimited," means just a massive billions or trillions of words long corpus including everything one ever heard, read, wrote or spoke. I hardly can remember what I did 10 minutes ago.