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

601–610 of 1001 posts

Re: GPT-4

#601

Post by MagicMoonlight »

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?

What we really need is a model that you can run on your own hardware on site. I could never use this for business because they're reading everything you send through it, but let me run it on my own server and it would be unbelievably useful.

Imagine being able to ask your workplace server if it has noticed any unusual traffic, or to write a report on sales with nice graphs. It would be so useful.

Re: GPT-4

#602

Post by TaupeRanger »

maxdoop wrote: post #265

The comments on this thread are proof of the AI effect: People will continually push the goal posts back as progress occurs. “Meh, it’s just a fancy word predictor. It’s not actually useful.” “Boring, it’s just memorizing answers. And it scored in the lowest percentile anyways”. “Sure, it’s in the top percentile now but honestly are those tests that hard? Besides, it can’t do anything with images.” “Ok, it takes imag…

There isn't and was never any movement of goalposts. They have been exactly the same for 70 years. We want creative systems (in the Deutschian sense) that can create new explanatory theories, which lead to actual new knowledge. When an AI is capable of creating new explanatory theories that are GOOD (not world salad), we will have human-like AGI. GPT is no closer to this goal than ELIZA (though it is much more useful).

Re: GPT-4

#603

Post by swalsh »

lionkor wrote: post #95

I cant wait for this to do targeted censorship! It already demonstrates it has strong biases deliberately programmed in: > I cannot endorse or promote smoking, as it is harmful to your health. But it would likely happily promote or endorse driving, skydiving, or eating manure - if asked in the right way.

Imagine a system where we have a UBI, but it's privately distributed by companies that own AI, and AI governs it.

AI: "I'm sorry I cannot allow you to use your OpenCredits on this item, you will have to use dollars."

You: "But I don't have dollars, I can't get a job anymore, AI does everything".

Re: GPT-4

#604

Post by dinvlad »

I wonder how long it takes till those stupid Leetcode problems as an initial "filter" become obsolete

Re: GPT-4

#605

Post by agnosticmantis »

This is all cute and entertaining, but my digital assistant still remains as dumb as ever and can’t process the simplest of ordinary tasks.

I still can’t ask my phone to “add a stop at cvs if it doesn’t add more than 5 minutes to my trip” while driving and using maps/navigation.

Is that too much to ask from a superhuman-performing AI that’s mastering all tasks and will disrupt everything? Or maybe the hype is more than it can deliver?

Re: GPT-4

#606

Post by majkinetor »

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?

Which will disappear soon enough, once it is able to run on premise.

Re: GPT-4

#607

Post by seydor »

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 assume that health providers will use servers that are guaranteed not to share data with openAi

Re: GPT-4

#608

Post by swerling »

On GPT-3.5 and now on GPT-4, I told it a line I could remember from a sonnet, and asked it to give me which sonnet it came from. It failed, and fabricated a sonnet that was a mashup of other sonnets.

It seems like maybe GPT-4 is not good at knowing when it does not know something?

Is this a common issue with LLMs? Also surprising (to me), it seems to give a slightly different wrong answer each time I restart the chat.

Here is the transcript in case you're curious:

``` Me: hello

GPT: Hello! How can I assist you today?

Me: I can remember a line I love from a sonnet, but can't remember which one.

GPT: Sure, I can help you identify the sonnet based on the line you remember. What is the line that you can recall?

Me: "Like a jewel hung in ghastly night"

GPT: The line you remember is from Sonnet 66 by William Shakespeare. The full line is: "And art made tongue-tied by authority, And folly (doctor-like) controlling skill, And simple truth miscalled simplicity, And captive good attending captain ill: Tired with all these, from these would I be gone, Save that, to die, I leave my love alone. But, like a jewel hung in ghastly night, Makes black night beauteous and her old face new." ```

Incorrect, the response is a mashup of at least 2 different sonnets. Still I think it's really impressive how it understood what to do with "I can remember a line I love from a sonnet, but can't remember which one." That's not a well phrased question, or even a question at all, but still it knew what I was getting at.

Re: GPT-4

#609

Post by pstorm »

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…

Reading the press release, my jaw dropped when I saw 32k. The workaround using a vector database and embeddings will soon be obsolete.

Re: GPT-4

#610

Post by HPMOR »

TaupeRanger wrote: post #602
maxdoop wrote: post #265

The comments on this thread are proof of the AI effect: People will continually push the goal posts back as progress occurs. “Meh, it’s just a fancy word predictor. It’s not actually useful.” “Boring, it’s just memorizing answers. And it scored in the lowest percentile anyways”. “Sure, it’s in the top percentile now but honestly are those tests that hard? Besides, it can’t do anything with images.” “Ok, it takes imag…

There isn't and was never any movement of goalposts. They have been exactly the same for 70 years. We want creative systems (in the Deutschian sense) that can create new explanatory theories, which lead to actual new knowledge. When an AI is capable of creating new explanatory theories that are GOOD (not world salad), we will have human-like AGI. GPT is no closer to this goal than ELIZA (though it is much more useful…

Bro what???!!?? GPT-4 is already being used as a personalized tutor on Kahn Academy. It’s personally helped me understand difficult Algorithms and CV applications in my undergrad classes. GPT-4 is about to revolutionize the world.
Post reply on HN