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

GPT-4o

openai.com

Post reply on HN
Print view View original

301–310 of 1001 posts

Re: GPT-4o

#301

Post by afro88 »

glenstein wrote: post #265

Text access rolling out today, apparently: >GPT-4o’s text and image capabilities are starting to roll out today in ChatGPT. We are making GPT-4o available in the free tier, and to Plus users with up to 5x higher message limits. Anyone have access yet? Not there for me so far.

I have access to GPT-4o text and audio, but no video. This is on the iOS app with ChatGPT Plus subscription.

Initial connection for audio fails most of the time, but once it's connected it's stable. Sometimes a bit more latency than expected, but mostly just like the demos.

Re: GPT-4o

#303

Post by aixpert »

msoad wrote: post #58

They are admitting[1] that the new model is the gpt2-chatbot that we have seen before[2]. As many highlighted there, the model is not an improvement like GPT3->GPT4. I tested a bunch of programming stuff and it was not that much better. It's interesting that OpenAI is highlighting the Elo score instead of showing results for many many benchmarks that all models are stuck at 50-70% success. [1] https://twitter.com/Lia…

useless anecdata but I find the new model very frustrating, often completely ignoring what I say in follow up queries. it's giving me serious Siri vibes

(text input in web version)

maybe it's programmed to completely ignore swearing but how could I not swear after it gave me repeatedly info about you.com when I try to address it in second person

Re: GPT-4o

#304

Post by fdb »

OutOfHere wrote: post #279

I am observing an extremely high rate of text hallucinations with gpt-4o (gpt-4o-2024-05-13) as tested via the API. I advise extreme caution with it. In contrast, I see no such concern with gpt-4-turbo-preview (gpt-4-0125-preview).

Same here. I observed it making up functions in d3 (`d3.geoProjectionRaw` and `d3.geoVisible`), in addition to ignoring functions it could have used.

Re: GPT-4o

#305

Post by tgtweak »

I feel like gpt4 has gotten progressively less useful since release even, despite all the "updates" and training. It seems to give correct but vague answers (political even) more and more instead of actual results. It also tends to run short and give brief replies vs full length replies.

I hope this isn't an artifact from optimization for scores and not actual function. Likewise it would be disheartening but not unheard of for them to reduce the performance of the previous model when releasing a new one in order to make the upgrade feel like that much more of an upgrade. I know this is certainly the case with cellphones (even though the claim is that it is unintentional) but I can't help but think the same could be true here.

All of this is coming as news that gpt5 based on a new underlying model is not far off and that gpt4(&o) may become the new gpt3.5-turbo use case for most apps that are currently trying to optimize costs with their use of the service.

Re: GPT-4o

#306

Post by blixt »

I don't see any details on how API access to these features will work.

This is the first true multimodal network from OpenAI, where you can send an image in and retain the visual properties of the image in the output from the network (previously the input image would be turned into text by the model, and sent to the Dall-E 3 model which would provide a URL). Will we get API updates to be able to do this?

Also, will we be able to tap into a realtime streaming instance through the API to replicate the audio/video streams shown in the demos? I imagine from the Be My Eyes partnership that they have some kind of API like this, but will it be opened up to more developers?

Even disregarding streaming, will the Chat API receive support for audio input/output as well? Previously one might've used a TTS model to voice the output from the model, but with a truly multimodal model the audio output will contain a lot more nuance that can't really be expressed in text.

Re: GPT-4o

#307

Post by bigyikes »

OutOfHere wrote: post #279

I am observing an extremely high rate of text hallucinations with gpt-4o (gpt-4o-2024-05-13) as tested via the API. I advise extreme caution with it. In contrast, I see no such concern with gpt-4-turbo-preview (gpt-4-0125-preview).

If true, makes me wonder what kind of regression testing OpenAI does for these models. It can’t be easy to write a unit test for hallucinations.

Re: GPT-4o

#308

Post by umbinder »

cube2222 wrote: post #154
msoad wrote: post #58

They are admitting[1] that the new model is the gpt2-chatbot that we have seen before[2]. As many highlighted there, the model is not an improvement like GPT3->GPT4. I tested a bunch of programming stuff and it was not that much better. It's interesting that OpenAI is highlighting the Elo score instead of showing results for many many benchmarks that all models are stuck at 50-70% success. [1] https://twitter.com/Lia…

I think the live demo that happened on the livestream is best to get a feel for this model[0]. I don't really care whether it's stronger than gpt-4-turbo or not. The direct real-time video and audio capabilities are absolutely magical and stunning . The responses in voice mode are now instantaneous, you can interrupt the model, you can talk to it while showing it a video, and it understands (and uses) intonation and…

[flagged]

Re: GPT-4o

#309

Post by ComputerGuru »

I have some questions/curiosities from a technical implementation perspective that I wonder if someone more in the know about ML, LLMs, and AI than I would be able to answer.

Obviously there's a reason in dropping the price of gpt-4o but not gpt-4t. Yes, the new tokenizer has improvements for non-English tokens, but that can't be the bulk of the reason why 4t is more expensive than 4o. Given the multi-model training set, how is 4o cheaper to train/run than 4t?

Or is this just a business decision, anyone with an app they're not immediately updating from 4t to 4o continues to pay a premium while they can offer a cheaper alternative for those asking for it (kind of like a coupon policy)?

Re: GPT-4o

#310

Post by cchance »

HOW ARE PEOPLE NOT MORE EXCITED, hes cutting off the AI mid sentence in these and its pausing to readjust in damn near realtime latency! WTF Thats a MAJOR step forward, what the hell is gpt5 going to look like.

That realtime translation would be amazing as an option in say Skype or Teams, set each individuals native language and handle automated translation, shit tie it into ElevenLabs to replicate your voice as well! Native translation in realtime with your own voice

Post reply on HN