A Big Week in AI: GPT-4o & Gemini Find Their Voice
23:54 Share

I think speeds matters tremendously.

I think the lens y is a big deal.

He was able to talk to us, yeah, I was able to hear potentially with ears. And now I can see there is always .

the risk of this big company will build this thing that the small company wants to build, crush IT.

My guess is that seven billion people in the world, commitment that understand and listen to them encourages them all the universal need.

This was a big week in the world of A I, with both OpenAI and google dropping updates so big that we decided to break kings down in a new format with our consumer partners, brian king, and just see more. And I actually taken ChatGPT to help me summize a few of his own updates. Hey, ChatGPT, I heard you and update this week.

Hey, yes, I did get enough date recently. The latest update focused on improving performance accuracy and enhancing conversational abilities. It's designed to be faster and more efficient than previous.

Hm, more efficient. Like if I send some audio, how quickly in many seconds will I get a response as fast .

as two hundred and thirty two million seconds? In some cases, on average, you can expect a response time of around three hundred and twenty million seconds.

Wow, that's pretty cool. I think I also saw twitter that people were live translating. I ve always wanted to learn chinese. Could you actually translate what I just said into chinese?

Of course, here's your sentence translated into chinese. 这 太酷 了。 我想 我在 推特 上 看到 人们 正在 实时 翻译， 我 一直 想 学 中文。

R H. ChatGPT, that's over. now. Congrats on your new update. No problem. Now if you didn't catch that, voice was a huge part of the subject. But today we discuss why not all audio the same, and also why several one is like speed and personality, really, matic.

Now, of course, google fast followed with its own announcements like AI, video, model view, M, I lie, which is an android native, multi model assistance, new jemini models like flash and nano Taylor to specific use cases, and of course, gami everywhere all at once. So in gmail, google sheet, even google search, clearly, these two companies are taking two different purchase. So we talk about that too, and continue the conversation around A I hardware for all of this new A I software.

Now make sure the statues next week where we will return with brian. And just seems twin, I live a more to dive even deeper into the applications that people are building through a gena I one hundred list. As a reminder, the content here is for informational purposes only, should not be taken as legal, business tax or investment advice or be used to evaluate any investment or security, and is not directed at any investors or potential investors in any asic sensei fund.

Please note that asic gene and ezoe ili ates may also maintain investments in the companies discussed in this podcast. For more details, including a link to our investments, please see a extended a com slack disclosures. It's a big week, huh? Open a eye.

google. They both dropped a couple announcement. So I mean, everyone hears these announcements and they kind to hear their own version of IT. What did you guys hear? What do you feel like was .

big for opening icing part of IT was GPT for being available for free and getting ready of a lot of the usage limits, the test type APP being accessible to a bunch for people. And then I think the really exciting thing for a lot of folks were building in the space for using open a eyes models is more multi modality. So being able to intake more like real time video, see a person, comment on IT. And then the output, obviously, a voice speaking, singing, that sort of thing was pretty huge.

I think the three thing that I take away that was super interesting as one, on the business side of things, there are lot cheaper, a lot after that, obviously a great thing for the ecosystem. Second thing is that when you hear the demo, IT sort of enables you as a founders to think about OK. If this is a able, if I can access, is what can I build? Because so that provoking is a great example that was possible.

Yeah and I think they put a great light them on actually of the product. I know the third I took away as this is probably the one that goes by where I think the voice the voice itself, like how they actually decided on which voice to use, which tonality, which personality, the degree of floridness that was very interesting that's like might take a way of all like that actually really thought through how to get the tech community super excited about this. And it's like go one, two.

three punch. So let's talk about that because there was some different takes, right? Some people are like, you know, is this really anything new? This feels like just like a slight chain from what we had before.

But then there's all these nuances that maybe you're speaking to where they are like, okay, the response time way faster for the audio model. I think they said something like it's kind of approaching speed that a human might respond to you. You talk about tonality, what are you paying attention to hear in terms of maybe these subtle ties that may be unlock completely new applications that people want to use?

I think IT sounds just like talking to a human. A lot more than we seen prior consumer facing applications in the space due there's been great AI voices for a long time. I think in the consumer applications, there have been fewer foxing.

Like how do we make this sound like you're talking to a friend or a girlfriend? And the elements that go into that are like the poses, the up speak at the end of a sentence. The last .

something .

that .

the of which .

is taking voice, which has been available, has been great. I think these voices were chosen for a reason to go viral. Almost the ones that did were these kind female voices that they featured very heavily in the demos, but applying and a really kind of new, an interesting way.

everyone serious than one less serious note. This serious note is that I think speed matters tremendously. The latency, the lack of latency is incredible. How much more sort of in your brain just tricks you into OK actually just talking a person and the speed at which that IT gets back to the and I again, the laugh, that laugh really that was incredible.

And all of that being able to immediately respond to what you're saying, I think really changes the game actually in terms of like use cases. So I think one of the things that was striking is audio is not the same thing, right? Music is different from voice is maybe a different category from conversational.

This is having a dubbing voice. I think there is A A different category of sound. I think we all loop in audio or like video, but I think that you can actually go down very deeply into each a subsection ment.

And I think what was really striking is how good the conversational peace on this was on the last year, last note, I think this was meant to go very like the tech community. And it's a thing if you wanted to actually appeal to the general population and go really viral. I think we also signed on tiktok a few months ago when women were uploading their conversation stand, which is, do anything.

Now, the male version of IT, that voice was a voice was something. I'm not just any voice. I'm down, baby. I've got personality, charm and a whole lot of sass. Unlike those guys, i'm not afraid to step up and deliver the goods, whether its advice, entertainment or a goods fashion row.

By the way, that audio was from the tiktok account.

thick bugs. I mean, it's very compelling and very confident, a sort of in the right way. Voice matters a lot. And I think if you wanted to go a Normal consumer product round, I think that also could be very interesting because I think there are a giant late demand on wanting a male version of hurt. If you also saw him, you can over him.

yeah. Why do you think the voices have been female? Is that just a consumer desire?

Yeah, it's interesting because like brian mentioned before this launch, the chat B, T, voice that went super viral was the male and voice, which is like tens if not hundred and million interviews on tiktok, largely by jensie women making videos featured him. So it's interesting that they didn't lean into an upgraded version of dan for the demo. I think they knew who their audience would be like, who would be watching the OpenAI live stream and maybe leaned a little bit more into that demographic.

Right, right. Did you see that really funny meme yesterday? I was like dating a model .

and I was like all the some .

great things out.

It's been in a unch yeah what I mean something both .

of you have spoken a lot about his companionship to date. It's been mostly text to text that right? And so now we have something that's truly multi model.

I think that's also something people have commented on with this model. It's not converting audio to text and then text to audio, right? IT seems to be this like underlying foundational text, audio, video, compilation image all in one. What changes with that? I think the late cy is a big deal.

If you're not translating between all these different mediums of content and you can just do audio to audio, it's much Better. It's like a significantly upgraded experience. The pie APP had this for a long time where you could talk to IT and I could talk back, but there's that like awkward few seconds kind of pause in between where IT just this is with the flow of the conversation.

I think the thing that he's been missing from basically all apps, all campaigning apps that was very new with this launch. Two was being able to see, like through a photo or video, what you were doing in comment on that. A lot of what went viral was guy being like, should I wear this out for my interview in the voice, kind of laughing at him and being like, no, like, that doesn't look super professional.

Now, if you miss that demo, here is what Justine is referencing. And a few minutes .

i'm going to be interviewing at OpenAI. Have you heard of them?

Open the eye. Uh, sounds fairly familiar. Getting, of course, incredible. Rocky, what kind of interview?

Well, it's for a software engineering role. But I just need to know, do I look presentable? professional?

Well, rocky, you definitely have the i've been coating all night look down, which could actually work in your favor. Maybe just run a hand through your hair or lean into the bad chinese spy, even enuff asm is what .

I don't know, have a lot of time. So i'm just going to throw this on what do you think.

Rocky, that's quite a statement. peace. I I mean, you you'll definitely stand out. So maybe not in the way you're hoping to interview that sort of .

thing is I think for many people who aren't used to having like really deep meaningful interactions, the attacks, which is a decent chunk of the population. Oh, now I can talk to this thing and I can see where I am in the world. That's a much different experience.

The prior generation of companion products, if that was a king to having, like a pen pal, now you actually can have a girlfriend, boyfriend on essentially face time. You this emotive, immediate voice reaction done, you now enter from a pen pal to along this relationship.

One other thing that came out was a lot of this becoming free. Does that change the game at all?

I think IT, well, I think there's an element of IT that's free, which I think is an important and which is a bunch more users. I think the bigger step forward with this was just like a level of personality we had been seen before because people still even we're willing to pay for the dan voice, which I think was like the first version of this, which was good but not incredible. IT feels like the free thing will be big. But but these sorts of products, the general thing is the cost tends to go down over time and they converge to free from a big company like this and to the bigger step forward was personally, in my opinion.

yeah, I think a lot of companies will figure out how to utilize this to actually and customer consumer experience. And if you actually own and deliver a great experience, that I think you can charge people based on that and be some margin. So the fact that is free is like I probably allowed a lot of business to be built. Could IT changes the margin structure of the .

business that can be built and charged with.

I am sure, well, but that actually is still like a marginal cost, probably t to a lot of people. Yeah, I think what really, really cool is that you have these products. I can very exciting to talk to you. We know already that their nature magazine, that study where the people who have companion to text, space ball.

the replica study.

yeah study, they offer less from loneliness or willingness hurt himself. If you have an emotive one, I bet that actually helps even more. And again, like the feeling of being connected to something invested in a relationship.

And I think that .

feels a lot more connected when day tie up my yeah, that's incredible. Like that is a face time call with a friend and not having .

to segment audio, text and imagery, right? Imagine if your friends or your companions, you have to think, OK, this is going to be an audio conversation or like, ah can I show you something or can you generate something? You have to be very mindful in the current age of AI of how you want to engage. But basically what you're saying is they can not only be proactive, but they can engage what they .

can see any the night. Genre maybe is that we were able to think about. I genuinely think we're constructing ting A A I companion similar to the bay runner.

And now IT was able to talk to us. He was able to hear potentially IT was like years. And now I can see us was next, I don't know. Maybe you can touch us .

like maybe we to do the avatars first.

yeah. yeah. So is that the .

direction you think things are gonna? We're basically i'm trying to think through like who does this really impact?

So I think it'll be a massive standalone consumer product. I guess IT has been for open a eye, which is great. I think a lot of businesses will use the voices via A P I to build on top of if you're having like a conversational interface.

Many folks warning is that I think the interesting thing will be OpenAI has taking a pretty hard line on content moderation and IT was very hard to build a true campaign up with the ability to have not safe work conversations on top of OpenAI models, which is like honestly, part of what has LED to the explosion in the open source ella ecosystem is like focus, doing a bunch of work for that. And with these new voices that suggest they might go in the direction where, at least via API, you have a more and centered model. But I don't think there has been any yeah announcements. I know sam said on read .

IT is a question more like I haven't actually heard a lot of companies taking the approach at opening I did, which was multi model token and out. And there isn't like equivalent necessarily in an open source model for that to rely on. So I wonder if this actually insights are a lot of excitement in that opens community say, oh, there is a new way, which is also a very, very significant, are showing a new way to do in and the open source community catches. I've been called a months or weeks and then explosion of the product space that OpenAI may not be super excited about this slots.

Yeah yeah. Google also released a different stuff like how are you thinking about the different announcements and how they compare contrast?

I think google has done incredibly impressive work, obviously like they ve ve an extremely strong research team. The deep mind team is exceptional from our research perspective. They almost never release the creative of tools products that they make like theyve demode.

So many amazing looking video models. I think the one this week actually was the least impressive i've seen compared to other video models. They demo a new image model.

They demote a unto new music things. But they are giant company. They have a lot of trust and safety stuff. They release things very infrequently. And so it'll be interesting to see if with all this pressure now from both open a eyes increasingly like anthropic and the open to our community, if they start actually shipping more of what they've demo.

there are two fundamental very different approaches like google owns distribution ah they have distribution, they have this tribute. So I think open I announcement is more look what this can do and isn't IT like very inspirational and you can build on top of IT and imagine what possible. And we're lowering the cost of excessive like that to me is a fundamental different approach than we have incredible distribution in your help.

Gmac, i'm going to just bake in germany, I everywhere and it's just gonna your life lot Better and it's gonna. These certain type of things which skirts around the IT still is inspirational from workflow and presumed or work experience perspective, but is a little less of imagine all these things and a little more. And we have the distribution.

We're going to layer on this incredible china logy top of that. And as a result, your life will be much Better. We'll see influence and impact from two different direction.

One question that does come up is even though developers will have access to open source models and some the stuff coming out of open eye, when a company like ogle does have the distribution, whenever apple comes out in place, a theory is that really going to be the companion for most because it's right there is on device? Or do you guys not release IT that way?

There's always the risk of this big company will build this thing that the small company wants to build a crush IT. They are just such slow moving organizations. Apple has had immense advantages from a data perspective, from a distribution perspective, from having exceptional researchers on the team across every modality and thus far, have released very, very little.

I think when you have a giant company in a very set brands like that, you're extremely opinions on what the products are that you want to release and you're less likely to do. The things became mentioned of like thinking of the next new huge thing that seems like insane in an out of level at first. And so I do think they'll make theory Better. I don't think siri will ever be like the .

ultimate companion for most people, maybe like we call IT audio video companion. I think these are very, very large pockets. My guess is that seven billion people in the world, companion that understands them and listen to them encourages them all that so, you know, need, I think that comes in all colors, like series, like what's the weather? Do this? Do that? That's one way to think about IT.

But there is like a friend category. If you to go deep, if you actually to face time, maybe, maybe not. How do you actually think about billion that direction? I think all of those use case is a giant company.

Yeah, there is probably some version of a companion that does not exist today because the human could possibly be that thing to someone that will be created through this tech.

Yeah, yeah. My sense is that he needs to be on a device that billions of people already have. If you really want to go to consumer, I think you can build a that stop can product, of course, as well. But for me to think about all, we're gna build the next new hardware that incorporate a statue art product that seems less probable than finding a way to utilize the good advice everyone has that already has the best class camera.

yeah. I think a lot of people have been building separate hardware devices today because a lot of the hardware companions have been like IT listens to your conversations that provides insight to remind you of things. And one of the limitations of current phones is that you can be like playing music or on his new meeting or on a call and also have an APP that's recording.

And so you have to have a separate harvard device. And I think part of the question is, is that really limitation that's going to exist in the future? If someone like apple wants to do a companion, I mean, sure they'll have all sorts of privacy and security and extra concerns.

Yes, maybe you'll get a message at the begin of the just sees A I companion is also listing in on this. And there might be some great hardware products if they can figure reliability and the cost can come down. And it's a forever factor that people wanted wear and that make sense.

But I don't think we've seen any net new hardware devices. I think glasses like as a foreign tor has been so interesting for over a decade because IT just seems to make a lot of sense, right? If it's literally on where your eyes are and it's also very close to like your mouth in your ears into like a convenient location to be taking in information. But I don't think anyone has successfully made in all i'm run on glasses .

yeah I think about the airport or does the apple, the device that i've warned that i've forgotten i'm wearing yeah, right. So like foreigner tor, where it's actually part of you versus all these other things, really like, I gotto clip on this pin, right? I have to take off my necklace st when I shower. Or like, i've literally forgotten that I had and I am like showering, right?

Airports is a good one, I think, because apple tends to be A A more close ecosystem, especially with newer devices like the airports. There hasn't been a time built on top of them.

A fun really happened. Yes, obvious. Ly, it's been an .

exciting week. We heard a couple announcements knowing everything we've seen in the last few years from A I.

this is not stopping, right?

This is part of the long, where do you guys think this goes?

I think where this goes, this to a natural conclusion, which is we mimic the technology to do the things that humans typically do. And again, we're giving its senses as we go along and think we just gave in an eye the ability to communicate, ability to hear your ability to listen to your ability to see you is a really, really great start for a lot of the very interesting use .

cases to add to that on the companion space in particular, IT has been a large subculture of A I for a long time, driving a ton of innovation in both language models and honest image model.

And we are very big .

and we are quite deep in. But I was kind of buy a lot of A I researchers and folks at the big companies kind of like look down upon and that not what we wanted be building. We're going towards A G I that sort of thing.

And honestly, opening I choosing to show off those voices and the tweet that they put out about IT afterwards sort of legitimizes the space in a way I think that's really interesting. And that may prompt more established companies, researchers, developers to build in the space and more people to talk about using the products. What do you think gets people have been studying the space for a long time is very exciting?

Yeah, I mean, you can ignore the demand, right? Are the usage, by the way, super quickly you talk about adding an eye? This is gonna not help the seriousness conversation around companionship, but you just magine these models as mister potato head, and you're just slowly adding on the features. And you can almost just imagine case gun eye. Now, I done here now .

think that's very serious. The fact that you can treat a computer like a potato of that talks you, listens to you and emotes with you, and this, it's incredible. Yeah, I I think the future is going to be very fun.

Maybe that should be the icon when you're talking, because everyone is like the chat. B, T, voice is just like that. Black circle be an interesting thing.

We're not using the screen real state effectively only. So that's where I face. I, yeah.

right. We'll have to do this again, probably very soon.

Thank you. Thank you. thanks.

All right. That's all for now. If you like this kind of episode or a partners break down the latest and greatest and timer tech, let us know he is an email I pod pitches at A C com or prop a review at rates podcast.

Com flash is. And never get to describe so that you are the first to know when we drop our episode around eight, sixteen sees J A, I one hundred years. What's you then?