Even performing a map-reduce over large documents to summarize or analyze them for a specific audience is largely beyond it. A 32K context size is a pittance when it comes to a single Title in the USC or CFR, which average into the millions of tokens each.
How do you think people become lawyers and how smart do you actually think the average lawyer is? The problem is that there's hundreds of thousands if not millions of pages, not that it requires superhuman intelligence to understand.
Even if it were capable of intelligence in the bottom quartile of humanity it would be SO MUCH more useful than it is now because I'd be able run and get something useful out of thousands of models in parallel. As it stands now GPT4 fails miserably at scaling up the kind of reasoning and understanding that even relatively stupid humans are capable of.
[1] https://www.amazon.com/Introduction-Legal-Reasoning-Edward-L...
It quickly breaks down even at 8k with legislation that is even remotely nontrivial.
What if someone finds a good practical way to expand the context length to 10M tokens? Do you think such model won't be able to do your task?
It seems like you have an opportunity to compare 8k and 32k GPT-4 variants (I don't) - do you notice the difference?
I can mow my lawn yet I still hire landscapers. That doesn't say anything about the difficulty of cutting grass or the intelligence of a DeWalt lawnmower but about specialization and economic tradeoffs - like the liability insurance accountants carry for their client work.
> What if someone finds a good practical way to expand the context length to 10M tokens? Do you think such model won't be able to do your task?
Not based on the transformer architecture (aka predict next token). It already fails at most of my use cases at 32K by default, unless I go to great lengths to tune the prompt.
> It seems like you have an opportunity to compare 8k and 32k GPT-4 variants (I don't) - do you notice the difference?
32K works better for my use case but requires much more careful prompt "engineering" to keep it from going off the rails. In practice, actually getting full 32K use out of it is a disaster since the connection will drop and I have to resend the entire context with a "continue" message, costing upwards of $10 for what should cost $2-4 per call. I haven't actually tried 32K on as much as a whole USC Title because that would costs thousands.
That would work.
I'm not sure that pointing out that LLMs are as useful for parsing legal code as the average human is something to brag about though.
All we've shown is that LLMs are as useful for parsing legal text as the average human. Which is to say, not.
A dog is also as useful for parsing legal texts as the average human. So is a rock.
First hand experience -I’ve been using it daily for the past two months.
I've used Google search for decades and it's been able to answer questions better than humans ever could. Same for Google Maps, though arguably they're the same system at this point. My calculator is far more intelligent than any human I've met, at least when it comes to adding large numbers. My compiler can detect even the slightest syntax error with impeccable accuracy. Microsoft word has an incredible vocabulary. Wikipedia knows more historical events than any human dead or alive. And so on.
Shit, users thought Eliza was intelligent in the 60s.
If what you really mean is that LLMs are cool and useful, then sure. Just say that instead of couching it in some vague claim of intelligence.
When you say "intelligent", what do you mean exactly?
What might help is describing what specific interactions give you the impression that GPT-4 is intelligent?
My interactions with GPT-4 include quite a wide range of questions: "how to talk to my kid in this specific situation", "what could have happened if Germany had won WW2", "what does this code do", "here's an idea for a research paper, let's brainstorm the details and implementation". I can also discuss with it anything that's been written in this thread and I'm sure it would provide intelligent responses (I haven't, btw).