Leaked Info Reveals BIG Open AI Event on Monday

Thank you Incogni for sponsoring this video.

▼ Link(s) From Today’s Video:

Tibor:

-------------------------------------------------

▼ Extra Links of Interest:

-------------------------------------------------

Thanks for watching Matt Video Productions! I make all sorts of videos here on Youtube! Technology, Tutorials, and Reviews! Enjoy Your stay here, and subscribe!

All Suggestions, Thoughts And Comments Are Greatly Appreciated… Because I Actually Read Them.

-------------------------------------------------

[MattVidPro]

Thank you Incogni for sponsoring this video.

[TeddyLeppard]

It looks like developers still have no idea what to do with all of this incomprehensible power and capability.

[FilmFactry]

Google is becoming Yahoo. Remember Kodak invented the digital camera, but didn't want to cannibalized their analog business. Where's Kodak now.

[jonmichaelgalindo]

Where's the open source humanoid bots though? 😢

[uranus]

There is actually another model in the Lmsys arena called I am-also-a-good-Gpt2-chatbot

[rebeccamiller]

When I asked the models if it knew they were in the chatbot arena, the im-a-good-gpt2 bots claimed they knew. Wonder if it was hallucinated, otherwise.. something strange going on there.

[mijailmead]

There seems to be another new model called

[FrotLopOfficial]

Ah so maybe GPT-2 will be announced and used for a free ChatGPT search engine. I doubt they could afford to run a global search engine running on 3.5 or 4

[starblaiz]

Well it just happened, and it did not disapoint! 😮 GPT-4o, integration of voice so that it's instantly responsive and interruptable so that you can have truly natural conversations with it, a heck of a lot of personality in the voice to the point where it's now basically passing the Turing test, live vision capabilities, and it's all free! 🤯 And the GPT-4o API will also be half the cost and 5x the rate limit of GPT-4turbo 🤯 I am SHOCKED at how this has STUNNED the industry! 🤯

[Chuck_Hooks]

Needs an integrated Second Brain to save and search the contents hundreds of saved chats

[VastCNC]

Maybe open ai is delivering an open source model that wipes the floor with the current field of play?

[MarkSokolov]

You: give your data to incogni, so it removes it
Incogni: steals your data

[mattelder]

FYI, people DO eat rocks. Ever heard of salt? That's a rock.

[NakedSageAstrology]

Exciting times! I can't wait for July!

[Dude_Wassup]

Imagine if their big announcement is releasing Chat GPT 4.6

[louis-dieudonne]

If that’s true, my final report would be saved.

[mickelodiansurname]

OpenAI must be internally streets ahead if they can continually wait until Google have a special annual thing like Google IO or a new model release to gizump them every time like clockwork. Cos what we end up doing here is saying, "okay so what else have they up their sleeve if this is just to troll google...". We even know the day GPT-5 will likely be released, it'll be the day just before Google announce they have something else to say!

[Edbrad]

I bet Sora news. And maybe even a music model. Does everyone remember OpenAI released a relatively good open source music model in 2020, called Jukebox? There’s no way they haven’t improved it. I wouldn’t be surprised if it’s Sora-level good, and better than Udio.

However when I did some research I found out it took a lot of processing power but most importantly it was EXTREMELY slow.

I have to wonder if Sora and other lesser music models are built off of any of the OpenSource code. 🤔 Anyway… it’s possible with Udio (and Google’s Lyria they’re annoying still keeping very limited) they might come out with a music model.

I also think this is likely because they have teased some impressive vocal models as well. And what is very impressive… Udio does vocals better than ElevenLabs… it’s crazy… and if you experiment enough it shows there’s something odd going on, it must have some sense of the meaning of the sentence/lyrics, in order to give you so much inflection that is way too accurate too often to be chance.

This might be partly why Udio is so good at music, like somehow it must have some actual understanding of what the idea is. As if there’s two dimensions going on, where you’d expect some aspect of randomness to the music it generates, but then on a higher level actually ends up giving you something that ended up that way because it understood it. Like a sad song that it makes sound sad because it understands it’s supposed to be sad in some limited level. … seriously experiment with making Udio do various spoken word stuff. Both lectures, standup, radio shows and musical theatre with spoken word. It’s astonishing. Takes a while to craft a great outcome, as it requires more detailed attention. It’s even more amazing if you take it’s generations and edit it outside of Udio. If you spend enough time on it you could make it sound practically perfect. Anyway… I had my Adderall. So I’ll shut up.

[Copa]

Open ai is already sitted on the largest model out there

[matetheking]

I don't think open ai is going to release GPT5 like other models. Based on comments from Sam, I see a iterative release, and just enough to stay ahead of competition