Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GPT-4

openai.com

Post reply on HN
Print view View original

611–620 of 1001 posts

Re: GPT-4

#611

Post by hanoz »

amelius wrote: post #595
hooande wrote: post #584

After watching the demos I'm convinced that the new context length will have the biggest impact. The ability to dump 32k tokens into a prompt (25,000 words) seems like it will drastically expand the reasoning capability and number of use cases. A doctor can put an entire patient's medical history in the prompt, a lawyer an entire case history, etc. As a professional...why not do this? There's a non-zero chance that i…

> As a professional...why not do this? Because your clients do not allow you to share their data with third parties?

I'd be furious if I found out some professional I'd commissioned had taken a document based on my own personal data, and poured over it themselves looking for errors at the tune of hundreds of dollars per hour, instead of sumbitting it to ChatGPT.

Re: GPT-4

#612

Post by billythemaniam »

hooande wrote: post #584

After watching the demos I'm convinced that the new context length will have the biggest impact. The ability to dump 32k tokens into a prompt (25,000 words) seems like it will drastically expand the reasoning capability and number of use cases. A doctor can put an entire patient's medical history in the prompt, a lawyer an entire case history, etc. As a professional...why not do this? There's a non-zero chance that i…

If you had full source code that fit into the context, do you think it could reliably answer questions about the code, build unit tests, generate documentation? I ask because that is the software equivalent of what you just described.

Re: GPT-4

#613

Post by red75prime »

albertgoeswoof wrote: post #294
Imnimo wrote: post #218

A class of problem that GPT-4 appears to still really struggle with is variants of common puzzles. For example: >Suppose I have a cabbage, a goat and a lion, and I need to get them across a river. I have a boat that can only carry myself and a single other item. I am not allowed to leave the cabbage and lion alone together, and I am not allowed to leave the lion and goat alone together. How can I safely get all three…

LLMs aren’t reasoning about the puzzle. They’re predicting the most likely text to print out, based on the input and the model/training data. If the solution is logical but unlikely (i.e. unseen in the training set and not mapped to an existing puzzle), then the probability of the puzzle answer appearing is very low.

> If the solution is logical but unlikely

The likeliness of the solution depends on context. If context is, say, a textbook on logical puzzles, then the probability of the logical solution is high.

If an LLM fails to reflect it, then it isn't good enough at predicting the text.

Yes, it could be possible that the required size of the model and training data to make it solve such puzzles consistently is impractical (or outright unachievable in principle). But the model being "just a text predictor" has nothing to do with that impossibility.

Re: GPT-4

#614

Post by leodriesch »

While AI gets better and better at creating what I would call "creative output", e.g. poems, texts of any form really, imagery and videos, I think the human skill it takes to produce these becomes less valuable.

In the future I imagine you'd no longer have to be good at writing poems, you'd just have to be good at distinguishing a "bad" poem from a good one. "Bad" is obviously highly subjective in this context.

So it becomes more and more important to have what I would call "good" taste, not the skills to do creative work yourself.

Re: GPT-4

#616

Post by bobsoap »

johnohara wrote: post #189

> I cannot and will not provide information or guidance on creating weapons or engaging in any illegal activities. Please let me know if there is another topic I can help you with. I understand "will not," but "cannot" seems to imply a highly curated "will not." The early GPT-4 response indicates the information was part of its dataset. Has the latest version made that information permanently inaccessible or has it b…

It's a LLM, not sentient. It doesn't know what "cannot" and "will not" means or implies. You're trying to interpret its output as you would a thinking person's.

I'd put it this way: when GPT refuses to answer, it just observes a topical no-go zone and uses the phrase it deems most likely to strongly convey refusal, as that's the phrase that was used most often/most successfully in its training data.

Re: GPT-4

#617

Post by in3d »

josho wrote: post #361
serjester wrote: post #221

Seems like OpenAI is forecasting massive changes to the job market. I highly recommend reading page 18 of the research paper. "GPT-4 or subsequent models may lead to the automation of certain jobs.[81] This could result in workforce displacement.[82] Over time, we expect GPT-4 to impact even jobs that have historically required years of experience and education, such as legal services.[83]"

I work at company that uses AI to automate about ⅓ of the job of trained licensed professionals. Looking at GPT4 those licensed professionals are now completely irrelevant. It's going to take years to build the supporting software around gpt4 to completely eliminate those jobs, but today I am convinced that we are on the verge of massive unemployment. Today thousands of job types have just been made redundant. What s…

Can you share what field you’re talking about?

Re: GPT-4

#618

Post by the_af »

pkaye wrote: post #461
Imnimo wrote: post #218

A class of problem that GPT-4 appears to still really struggle with is variants of common puzzles. For example: >Suppose I have a cabbage, a goat and a lion, and I need to get them across a river. I have a boat that can only carry myself and a single other item. I am not allowed to leave the cabbage and lion alone together, and I am not allowed to leave the lion and goat alone together. How can I safely get all three…

> I am not allowed to leave the cabbage and lion alone together, and I am not allowed to leave the lion and goat alone together. How can I safely get all three across? What if you phrase it as a cabbage, vegan lion and a meat eating goat...

The result is hilariously wrong in all sorts of way: it doesn't understand what "vegan" means, it gives the wrong descriptions, and it carries stuff from the shore where it wasn't.

Here's what ChatGPT told me (I used "wolf" instead of "lion"):

"This is a similar problem to the classic river crossing puzzle, but with a few different elements to consider. Here's one possible procedure to solve this riddle:

1. The man takes the cabbage across the river and leaves it on the other side.

2. The man goes back alone and takes the wolf across the river, leaving it with the cabbage on the other side. It's important to note that the wolf is a vegan and will not eat the cabbage.

3. The man goes back alone and takes the goat across the river, leaving it on the other side.

4. The man goes back to the original side and picks up the cabbage, bringing it across the river to join the others.

5. Finally, the man goes back to the original side and picks up the wolf, bringing it across the river to join the others.

In this way, all of the man's belongings have been safely transported to the other side of the river, without the wolf eating the goat or the cabbage being destroyed."

Re: GPT-4

#619

Post by 0898 »

hooande wrote: post #584

After watching the demos I'm convinced that the new context length will have the biggest impact. The ability to dump 32k tokens into a prompt (25,000 words) seems like it will drastically expand the reasoning capability and number of use cases. A doctor can put an entire patient's medical history in the prompt, a lawyer an entire case history, etc. As a professional...why not do this? There's a non-zero chance that i…

When will the longer context length be available through ChatGPT Plus? Have they said yet?

Re: GPT-4

#620

Post by amelius »

chis wrote: post #46
teruakohatu wrote: post #26

Access is invite only for the API, and rate limited for paid GPT+. > gpt-4 has a context length of 8,192 tokens. We are also providing limited access to our 32,768–context (about 50 pages of text) version, gpt-4-32k, which will also be updated automatically over time (current version gpt-4-32k-0314, also supported until June 14). Pricing is $0.06 per 1K prompt tokens and $0.12 per 1k completion tokens. The context le…

I'm really curious to see if expanding the context length this much will allow GPT to do typical software development tasks on a big codebase. If it can take in a github issue and produce decent code solving a complex issue across many files... will certainly be an interesting time.

It can't even do simple sysadmin tasks like fixing a broken installation, or fixing simple configure/make/make install issues.
Post reply on HN