Like clockwork, every time they need to drum up excitement:
https://www.theverge.com/2019/11/7/20953040/openai-text-gene...
To be clear, it is true that transformers did change things but the merchants are still over selling it and everyone else laps it up while not meta-thinking about it for even one second.
There's also a huge range of practical demonstrations of non-aligned, monomaniacal and not particularly smart intelligences, that literally eat humans: bacteria.
(Also lions and tigers and bears, if you want to insist that evolution doesn't itself count as intelligence).
Personal opinion: I think this means we've probably exhausted all the low hanging fruit in LLM land. This was the last thing I was reserving judgement for. When the most hyped up big idea openai has rn is basically "we're just gonna have the model dump out a wall of semi-optimized chain of thought every time and not send it over the wire" we're officially out of big ideas. Like I mean it obviously works... but that's more or less what we've _been_ doing for years now! Barring a total rethinking of LLM architecture, I think all improvements going forward will be baby steps for a while, basically moving at the same pace we've been going since gpt-4 launched. I don't think this is the path to AGI in the near term, but there's still plenty of headroom for minor incremental change.
By analogy, i feel like gpt-4 was basically the same quantum leap we got with the iphone 4: all the basic functionality and peripherals were there by the time we got iphone 4 (multitasking, facetime, the app store, various sensors, etc.), and everything since then has just been minor improvements. The current iPhone 16 is obviously faster, bigger, thinner, and "better" than the 4, but for the most part it doesn't really do anything extra that the 4 wasn't already capable of at some level with the right app. Similarly, I think gpt-4 was pretty much "good enough". LLMs are about as they're gonna get for the next little while, though they might get a little cheaper, faster, and more "aligned" (however we wanna define that). They might get slightly less stupid, but i don't think they're gonna get a whole lot smarter any time soon. Whatever we see in the next few years is probably not going to be much better than using gpt-4 with the right prompt, tool use, RAG, etc. on top of it. We'll only see improvements at the margins.
its both too linear, and hard to revise. it's hard to undo parts of the conversation that poison it. it's too hard to save important bits that shouldn't be forgotten or drowned out. its not word processor like enough.
i envision the next generation of these products being multipane by default.
on the left I have a chat, in the center I have a whiteboard, and on the right I have a rendered document. throwing a clip of something onto the whiteboard makes it modifiable by chat. "store this, categorize it, summarize it, place it in the document."
whatever comes next needs to function more like onenote or obsidian. just to use an example from current events today, lets say I want to make a Parody Dossier on Walz, similar to todays Vance leak. I should be able to describe the project to chat. It builds a document structure. I tell it we are going to scrape all of the internets jokes on Walz at a bbq or other non-scandals. I should be able to quickly click through a table of contents, and "chat" with each paragraph. "This one needs fleshing out, this one needs summarization." As we scrape a reddit post, we want to incorporate not only the original post, but all the best comments. I should be a be able to "chat with the document editor" and put together a 200 page document in the amount of time it took me to write this post, just by describing what is and isnt working, and dragging and dropping.
chat, the simplicity of it, understanding of complex sentences, and multi sentence conversations was a UI paradigm leap. it went well past keyword search and the new command line of the internet. its a great first step, and a nice reset after a decade of interface stagnation, but now its ubiquity and simplicity, like the search box, is clouding peoples imagination and ability to dream up the next new interactive interface, which I expect to involve more mouse and visual relationships.
tldr: the llm is a component of the next generation interface, not the entire interface itself.
Re "all the basic functionality" if you are talking intelligence then you are kind of talking the things humans can do and LLMs kind of do basic language but are not much good and common sense reasoning, spatial awareness and shape rotation, motor skills and running around, emotional intelligence, figuring what people are thinking, planning and probably some other stuff. It'll be iPhone 4 when those are covered.
I think it's as they say: reinforcement learning applied to cause it to generate a relatively long 'reasoning trace' of some kind from which the answer is obtained through summarisation.
I think it's likely a cleverly simplified version of QuietSTaR, with no thought tokens, just one big generation to which the RL is applied.
The way I believe it's trained in practice is as follows: they have a bunch of examples, some at the edge of GPT-4s ability to answer, some beyond it, some that GPT-4 can answer if you're lucky with the randomness. Then they give it one of these prompts, generate a fairly long text, maybe 3x the length of the answer, and summarize that to produce the final answer. Then they use REINFORCE to reward the generated texts that increase the probability of the summary being correct.
Is it? I feel like if you don't care about the cost it's pretty easily replicable on any other LLM, just with a lang-chain sort of approach
this may have consequences for how useful it is.
There were months worth of articles on how strawberry is considered almost dangerous internally its so smart. I know we only got -mini and -preview but...this doesn't feel like AGI