Can’t wait to get interviewed by this model!
GPT-4o
471–480 of 1001 posts
Re: GPT-4o
#472Ohhhhhhhh, boy... Listening to all that emotional vocal inflection and feedback... There are going to be at least 10 million lonely guys with new AI girlfriends. "She's not real. But, she interested in everything I say and excited about everything I care about" is enough of a sales pitch for a lot of people.
Pretty much, tech is what we make of it no matter how advanced. Just look at what we turned most of the web into.
Re: GPT-4o
#473feature request: please let me change the voice. it is slightly annoying right now. way too bubbly, and half the spoken information is redundant or not useful. too much small talk and pleasantries or repetition. I'm looking for an efficient, clever, servant not a "friend" who speaks to me like I'm a toddler. felt like I was talking to a stereotypical American with a Frappuccino: "HIIIII!!! EVERYTHING'S AMAZING! YOU'RE BEAUTIFUL! NO YOU ARE!"
maybe some knobs for the flavor of the bot:
- small talk: gossip girl stoic Aurelius
- information efficiency or how much do you expect me to already know, an assumption on the user: midwit genius
- tone spectrum: excited Scarlett, or whatever it is now Feynman the butler
Re: GPT-4o
#474The most impressive part is that the voice uses the right feelings and tonal language during the presentation. I'm not sure how much of that was that they had tested this over and over, but it is really hard to get that right so if they didn't fake it in some way I'd say that is revolutionary.
That was very impressive, but it doesn't surprise me much given how good the voice mode is in the ChatGPT iPhone app is already. The new voice mode sounds better, but the current voice mode did also have inflection that made it feel much more natural than most computer voices I've heard before.
Slight off-topic, but I noticed you've updated your llm CLI app to work with the 4o model (plus bunch of other APIs through plugins). Kudos for working extremely fast. I'm really grateful for your tool; I tried many others, but for some reason none clicked as much as your.
Link in case other readers are curious: https://llm.datasette.io
Re: GPT-4o
#475I’m a huge user of GPT4 and Opus in my work but I’m a huge user of GPT4-Turbo voice in my personal life. I use it on my commutes to learn all sorts of stuff. I’ve never understood the details of cameras and the relationship between shutter speed and aperture and iso in a modern dslr which given the aurora was important. We talked through and I got to an understanding in a way having read manuals and textbooks didn’t…
I think this is probably one of the most compelling personal uses for a tool like this, but your use of it begs the same question as every other activity that amounts to more pseudo-intellectual consumption; what is the value of that information, and how much of ones money and time should be allocated to digesting (usually high-level) arbitrary information?
If I was deliberately trying to dive deep on one particular hobby, or trying to understand how a particular algorithm works, there's clear value in spending concentrated time to learn that subject, deliberately focused and engaged with it, and a system like your describe might play a role in that. If I'm in school and forced to quickly learn a bunch of crap I'll be tested on, then the system has defined another source of real value, at least in the short term. But if I'm diving deep on one particular hobby and filling my brain with all sorts of other ostensibly important information, I think that just amounts at best to more entertainment that fakes its way above other aspects of life in the hierarchy of ways one could spend time (the irony of me saying this in a comment on HN is not lost on me).
Earlier in my life I figured it would be worthwhile to read articles on the bus, or listen to non-fiction podcasts, because knowledge is inherently valuable and there's not enough time, and if I just wore earbuds throughout my entire day, I'd learn so much! How about at the gym, so much wasted learning time while pushing weights, keep those earbuds in! A walk around the neighborhood? On the plane? On the train? All time that could be spent learning about some bs that's recently become much easier to access, or so my 21 y.o self would have me believe.
But I think now it's a phony and hollow existence if you're just cramming your brain with all sorts of stuff in the background or in marginally more than a passive way. I could listen to a lot of arbitrary German language material, but realistically the value I'd convince myself I'd get out of any of that is lost if I'm not about to take that home and grind it out for hours, days, move to a German speaking country, have an existing intense interest in untranslatable German art, or have literally any reason to properly learn a language and dedicate real expensive time to it.
Re: GPT-4o
#476Are these multimodals able to discern the input voice tone? Really curious if they're able to detect sarcasm or emotional content (or even something like mispronunciation?)
Yes, they can, and they should get better at this over time.
There is a demo video where the presenter breathes heavily and asks the AI is able to notice it as such when prompted.
It can’t just detect tone, it seems to also be able to use tone itself.
Re: GPT-4o
#477it's crazy that Google has the Youtube dataset and still lost on multimodal AI
Re: GPT-4o
#478Apple and Google, you need to get your personal agent game going because right now you’re losing the market. This is FREE.
Tweakable emotion and voice, watching the scene, cracking jokes. It’s not perfect but the amount and types of data this will collect will be massive. I can see it opening up access to many more users and use cases.
Very close to:
- A constant friend
- A shrink
- A teacher
- A coach who can watch you exercise and offer feedback
…all infinitely patient, positive, helpful. For kids that get bullied, or whose parents can’t afford therapy or a coach, there’s the potential for a base level of support that will only get better over time.
Re: GPT-4o
#479it really feels like the quality of gpt4's responses got progressively worse as the year went on... seems like it is giving political answers now vs actually giving an earnest response. It also feels like the responses are lazier than they used to be at the outset of gpt4's release.
I am not saying this is what they're doing but it DOES feel like they are hindering previous model to make the new one stand out that much more. The multi-modal improvements here and release are certainly impressive but I can't help but feel like the subjective quality of gpt4 has dipped.
Hopefully this signals that gpt5 is not far off and should stand out significantly from the crowd.
Re: GPT-4o
#480For all the hype around this announcement I was expecting more than some demo-level stuff that close to nobody will use in real life. Disappointing.
Why are you so confident that nobody will use this in real life? I know OpenAI showed only a few demos, but I can see huge potential.