OpenAI's GPT-4o Voice Alpha Release, Midjourney & Runway GEN-3 Updates & More AI News

AIForHumans69AudioRaw
===

[00:00:00] 

Welcome, welcome to AI for Humans, your weekly guide into the wonderful world of generative AI and all of the AI news, tools, and other stuff that's happening. We have an unusual show this week.

I am by myself in a car in Fresno, California because we are actually supposed to take this week off, but I had to come down from the mountains. And I figured why not record a very quick audio podcast to get you guys updated because there's a lot of big breaking AI news this week. 

We'll get back to our regular show next week. Of course, as always, please leave us a five star review on Apple podcast or leave us any review. It always helps us. We are continually growing the show and each week more and more of you listen, which we're really excited about, but that's more than enough BS for now.

We are going to get straight to the news today because there are some quick stories and I promise we'll get you out of here in about 15 minutes. It's time for the news.

[00:01:00] 

A surprisingly big week in AI this week, OpenAI has started to drop, it's voice mode, it's advanced GPT 4. 0 voice mode is out to a very few select , chat GPT Plus users. Kevin and I are not on that list yet, but please OpenAI, we need you. Bring us to the table. We're excited to play around with it.

there is a lot of excitement here. We've seen some early clips. There is a Reddit user named Begara who has been sharing access to the clips that he's gotten. He did a really interesting back and forth where he asked chat GPT to help him pronounce the word croissant the French croissant, take a listen to that.

Hey, how's it going? I'm fine. I'm trying to practice my French. Can you help me with that? Bien sûr. How would you like to practice your French today? Am I saying this word correctly? Croissant. Pretty close. Try emphasizing the nasal sound at the end a bit more, like croissant. [00:02:00] How does that feel?

Croissant. That's it. And He also showed off where he asked it to sing happy birthday for him but do it in a bluesy way. And in this clip you get a real sense of the ability for the voice mode to modulate and try different things. 

Hey, if I wanted to sing happy birthday to someone, but like a soulful blues singer, how should I sound like? You probably need some emotion and some growl to your voice. Drawing out the notes and maybe throwing in some bluesy riffs. It might sound something like, you

know, really feeling each word. 

So obviously this is in a very small group of people so far.

, we have seen a couple people describe the fact that they aren't able to do anything that involves copyright material, which is no big surprise. But [00:03:00] assuming that this is going to go well, OpenAI has said that everybody's going to get access to this, all plus users. That means if you're a paid user, you're You will get access by September.

So that's a pretty big deal. Other open AI news, search GPT was announced earlier this week. This is something we've been talking about for a while. The fact that open AI is actually spending time now working on a search product.

People talk about, , AI companies not making all that much money. There's been a lot of narrative around how generative AI as a whole has flopped. I say wait until this fall, as we've said in this podcast, but also search GPT is going to change the overall search experience.

I suggest really highly you go and listen to the Verge cast from a couple of weeks ago with Nilay Patel and that crew, and they talked about what a different world search is going to be in going forward. Yeah. And I think in a lot of ways, I'm looking forward to this search experience. For me, Google has been broken for quite a while.

If open AI can solve search in an interesting way, or at least a different way, it will be interesting. This also is partly why [00:04:00] they made those deals with content companies. Cause they want to be able to surface news. The company Perplexity ai, which I know a lot of people our audience are familiar with and use on a daily basis has complained that OpenAI is coming to their business model.

But I would say that OpenAI was probably planning for this for quite a while. So I don't think it's a shock they're doing this, but I do think it's a big deal. Alright, another big update that happened this week was MidJourney dropped 6.

1. Oftentimes with MidJourney, the 1 updates they seem small, but they're actually much bigger than they seem. Much better looking Much more photorealistic, much better looking hands, better use of texts. Now it's not going to allow you to do in painting yet. It's only for image generation, but it's a pretty big update. I've played around with a little bit, not enough yet to really have a judgment call, hopefully next week on the podcast, we'll do some deeper dives and in connection with that, not the same company, but something I think you're going to find really interesting.

If you care about creating AI video, Runway Gen 3 has now opened up image to [00:05:00] video for paid users. So before this point, Runway Gen 3, one of the biggest AI video generation software was only allowing text to video, which is much harder to get consistency with character.

It's So image to video, Allows you to upload an image which can often be created in something like mid journey and then create the video from that So this allows content creators AI content creators to create consistent characters because within mid journey It is much easier to keep a consistent look to create essentially style guides all sorts of other stuff Now runway is definitely having a little bit of a moment.

There was a story that broke early last week that runway gen three trained a lot of their video on YouTube. Like the other story that we had talked about the pile 404 media, again, breaking the story.

Shout out to those guys that are doing incredible journalistic work. They found a spreadsheet that actually showed all of the different types of YouTube creators. And even went so far as trying to recreate. Specific YouTubers and their shots within gen three and got pretty [00:06:00] close. So AI video and AI imaging, as we've said from the beginning is a little bit tricky because there is an original sin based on who and where they have trained their data.

That said, we are looking at the upgrades happening faster and faster. journey 6. 1 and runway image to video, we are going to see some really incredible creative material be created. But also it's important to realize what's going on in the background. Another big story, mostly big in some ways for the price that was paid for the URL, but there was a new company launched called friend and friend is an AI pendant.

It has a kind of funny video that Marques Brownlee, one of our favorite tech Bloggers said he thought was a dystopian fake ad turned out to be real.

I know the effects are crazy. It's dank. I could eat one of these every day.

The idea with friend is that is AI as a companion is there to be with you as you have good days and bad [00:07:00] days. I'm not entirely convinced by this. But the bigger story here again from our friends at 404 media, is that the company that is friend now raised a VC round of 2.

5 million. And they paid for 1. 8 million for the URL friend. com. They've got to feel very strongly about that. The fact that their tech may be relatively cheap enough to make that they can spend, what is that? Like 85 percent of their funding around on a URL, a granted friend. com probably has some incredible Google juice.

And now who knows how long that'll last with search GPT, but they definitely have some incredible Google juice there. a story we're going to be following. I don't personally love the idea of wearing a giant thing around my neck.

I still think if we're going to have AI companions that are with us at all times, it's likely going to be in the mode of either our phone or Meta's glasses that have come out. Or conceivably, I still think there's a really interesting play to have something just in my earbuds. I wear my AirPods around more than almost anything else, and [00:08:00] to me, that feels like the future of where the voice assistant will live.

And again, if ChatGPT's voice assistant is as good as it seems so far, That feels like where we are going to be seeing the AI companion come to the forefront. And finally, a dumb story, but also a fun one. Taco Bell is folding themselves deeply into the world of AI. You may remember a couple months ago, we talked about how McDonald's was canceling their AI trial for their ordering and all sorts of other stuff.

And we said at the time that it might just be that they didn't love the implementation. It was a test. Taco Bell Loves the implementation of ai. 

Yum. Brands has specifically said that they think that AI ordering has allowed less problems, has stopped mistakes on their orders, and they are going full force into AI customer service. They are doubling down on their efforts rather than pulling back.

And I think this is an interesting thing as a wrap up here in that there's a lot of conversation that's happening right now around the idea that AI is going to go through a real bubble pop moment. And I [00:09:00] just want to say to everybody who's listening to this podcast and has been listening for a while, or if you're new, Keep in mind, we are in a lull in part because of the cycles that happen in AI.

There was a story briefly that said a rumor. This is just a rumor. That was floating around the last couple of days that GPT five finished training in April. And I really do believe this is Gavin saying this and Kevin's not here. So we can't dispute it.

I really do believe we are going to see a big new model come from open AI by the end of this year. Why do I think that? GPT 4 0, the text model has been out for a bit, and yet we haven't seen the voice model or the image capabilities on GPT 4 But I think they've been waiting on something rather large.

We will see. There's a really interesting person to read out there. If you're interested in the anti AI take, especially anti generative AI, Gary Marcus, if you're familiar, is a very loud voice saying that the AI bubble is about to pop. But even he had some interesting positive things to say about what deep mind has been doing and how they've been working in the AI space.

As much as people are out there are talking [00:10:00] about how AI might be a bubble. I think it's a really important thing to couch everything that you're listening to. It's never as good and it's never as bad as people say. I think there is clear.

Progress that is happening. And I think from Kevin in my perspective, it's obvious that this is just going to get better and better. many people in this space say, this is the worst it will ever be. That's it for this week. A very short truncated edition of AI for Humans.

But we will be back next week with a full episode and thank you everybody for listening. Feel free again, take some time this week to leave us a five star review or again, any review on Apple podcasts or any of your podcast platforms. It really helps the show.

Please talk about us, share us on threads or on X. We're definitely spending some more time on threads and just tell somebody about the show this week. All right, I'm heading back up to the mountains, everyone. Thank you for listening. We will see you all next week.