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

361–370 of 1001 posts

Re: GPT-4o

#361

Post by caseyy »

yieldcrv wrote: post #331
caseyy wrote: post #319

Few people are talking about it but... what do you think about the very over-the-top enthusiasm? To me, it sounds like TikTok TTS, it's a bit uncomfortable to listen to. I've been working with TTS models and they can produce much more natural sounding language, so it is clearly a stylistic choice. So what do you think?

All these language models are very malleable. They demonstrated changing the temperament in the story telling time.

Looks like their TTS component is separate from the model. I just tried 4o, and there is a list of voices to select from. If they really only allowed that one voice or burned it into the model, then that would probably have made the model faster, but I think it would have been a blunder.

Re: GPT-4o

#362

Post by Jensson »

kleiba wrote: post #312

I cannot believe that that overly excited giggle tone of voice you see in the demo videos made it through quality control?! I've only watched two videos so far and it's already annoying me to the point that I couldn't imagine using it regularly.

Just tell it to stop giggling if you don't like it. They obviously choose that for the presentation since it shows off the hardest things it can do, it is much easier to act formal, and since it understands when you ask it to speak in a different way there is no problem making it speak more formal.

Re: GPT-4o

#363

Post by noman-land »

cs702 wrote: post #115

The usual critics will quickly point out that LLMs like GPT-4o still have a lot of failure modes and suffer from issues that remain unresolved. They will point out that we're reaping diminishing returns from Transformers. They will question the absence of a "GPT-5" model. And so on -- blah, blah, blah, stochastic parrots, blah, blah, blah. Ignore the critics. Watch the demos. Play with it. This stuff feels magical .…

Magic is maybe not the best analogy to use because magic itself isn't magical. It is trickery.

Re: GPT-4o

#364

Post by EternalFury »

cchance wrote: post #310

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…

At some point, scalability is the best form of exploitation. The exploration piece requires a lot more that engineering.

Re: GPT-4o

#365

Post by karmasimida »

cs702 wrote: post #115

The usual critics will quickly point out that LLMs like GPT-4o still have a lot of failure modes and suffer from issues that remain unresolved. They will point out that we're reaping diminishing returns from Transformers. They will question the absence of a "GPT-5" model. And so on -- blah, blah, blah, stochastic parrots, blah, blah, blah. Ignore the critics. Watch the demos. Play with it. This stuff feels magical .…

Very convincing demo

However, using ChatGPT with transcribing is already offering me similar experience, so what is new exactly

Re: GPT-4o

#366

Post by scarface_74 »

cs702 wrote: post #115

The usual critics will quickly point out that LLMs like GPT-4o still have a lot of failure modes and suffer from issues that remain unresolved. They will point out that we're reaping diminishing returns from Transformers. They will question the absence of a "GPT-5" model. And so on -- blah, blah, blah, stochastic parrots, blah, blah, blah. Ignore the critics. Watch the demos. Play with it. This stuff feels magical .…

Some of the failure modes in LLMs have been fixed by augmenting LLMs with external services

The simplest example is “list all of the presidents in reverse chronological order of their ages when inaugurated”.

Both ChatGpt 3.5 and 4 get the order wrong. The difference is that I can instruct ChatGPT 4 to “use Python”

https://chat.openai.com/share/87e4d37c-ec5d-4cda-921c-b6a9c7...

You can do similar things to have it verify information by using internet sources and give you citations.

Just like with the Python example, at least I can look at the script/web citation myself

Re: GPT-4o

#367

Post by croes »

cs702 wrote: post #115

The usual critics will quickly point out that LLMs like GPT-4o still have a lot of failure modes and suffer from issues that remain unresolved. They will point out that we're reaping diminishing returns from Transformers. They will question the absence of a "GPT-5" model. And so on -- blah, blah, blah, stochastic parrots, blah, blah, blah. Ignore the critics. Watch the demos. Play with it. This stuff feels magical .…

>This stuff feels magical. Magical.

Sound like the people who defend Astrology because it feels magical how their horoscope fits their personality.

"Don't bother me with facts that destroy my rose-tinted view"

At moment AI is a massive hype and shoved into everything. To point at the faults and weaknesses is a reasonable and responsible thing to do.

Re: GPT-4o

#369

Post by hamilyon2 »

Image editing capabilities are... nice. Not there yet.

Whatever I was doing with Chatgpt 4 became faster. Instant win.

My test benchmark questions: still all negative, so reasoning on out-of distribution puzzles is still failing

Post reply on HN