🚀 Add to Chrome – It’s Free - YouTube Summarizer
Category: AI News
Tags: AIGamingGoogleImageEditingProductivity
Entities: 11 LabsAlibabaGoogleGoogle PixelGrammarlyMicrosoft CopilotNano BananaQuinn Image EditRunway ML
00:00
There was a ton of really cool AI updates this week. We have AI image editors that could potentially replace Photoshop.
Tons of updates out of Google and their made by Google event. And even teases around GPT6.
Yes, only a couple weeks after GPT5, we're already talking
00:16
about GPT6. All that and more in today's AI news breakdown.
I don't want to waste your time, so let's dive right in. Starting with the new image editor out of Quinn called Quinn Image Edit.
Oh, I mentioned a new image editor that could replace Photoshop. Well, this week there's actually two.
00:33
No. God, please no.
You've got this Quinn image edit which comes out of Chinese company Alibaba and is actually opened via the Apache 2.0 license. And similar to what you can do inside of Chat GPT where you can give it images and then tell it to, you know,
00:50
change the angle of the person or put a hat on them or things like that. Well, Quinn imageedit is an open version of that.
And if we look at their page here, we could see some examples of like this input image and then the various sort of uh derivatives they built off of this
01:06
input image. Feel free to pause here and see the prompts that they used.
But you can see they gave it one input image and then gave it a prompt like the bear is wearing an elaborate dance dress with arms spread open performing an elegant dancing pose. And you got this image
01:21
using this same bear here. More examples.
Input image. Obtain the front view and it turned the person around.
And we have some more examples of the same type of prompt. Here we can see an input image where they change the style to like a Giblly style and a 3D cartoon style.
Here's an input image where they
01:37
added a sign that says, "Welcome to Penguin Beach." So, pretty much the sky's is the limit on what type of edits you can make to these images here. And you can use this model totally free right now by going to the Quinn Chat website, which is just quin.ai.
And then once you're logged in, you'll notice
01:53
that there's a little button down in the bottom that says image edit. We select this image.
It says upload a picture and describe how you'd like it edited. So this is one of their demo images here.
And it automatically plugged in this prompt while I was testing that says replace the polka dot shirt with a light blue shirt. And we can see it did just
02:09
that. I'm going to go ahead and feed it an image of myself here.
And then say give me the side view of this image. And it actually did a really good job.
I mean that actually kind of looks like my profile. Here's an image I uploaded of me and my friends Molly and Joe, and I
02:25
said, "Make it nighttime with a city skyline behind us." And a few seconds later, I got the exact image I was looking for. I mean, see how this could potentially replace Photoshop.
And this is just one of two image editors that everybody's talking about this week. Maybe you've been hearing the term nano
02:41
banana flying around this week as well. Well, that's because that does the same thing, but almost feels even more powerful than this one.
But it's not totally accessible yet. There is a workaround to use it which I made a whole video about a few days ago.
So check that video out if you haven't already. But let me give you a quick
02:57
overview of what it's capable of and then I'll give you the quick TLDDR of how you can actually generate with it right now. So here's some examples from this Xacount D studio project.
And we can see they gave it a base image here of this woman on the left with a white purse. They gave it this image of a new
03:12
purse with the prompt change the bag with the second image. And it put this purse on this woman here.
Another example. Here's the base image where she's reading a book.
They give it this image and say replace the book cover with the cover with the second image. And it replaced just the book cover,
03:28
leaving the rest of the image pretty much exactly the same. Another one where they gave this girl these glasses and this girl this like puffer jacket here.
Here's another example from Mariano Partardo where he was able to give it three different images. He gave it an image of Satia Nadella, Sundar Pinchi,
03:45
and an image of the beach and said, "Put the two persons together on the beach in a natural way." And you can see it made this image with Satia and Sundar hanging out together on this beach. And everything stayed consistent.
It's the same beach. Maybe they're wearing different outfits, but you can
04:00
definitely tell who it's supposed to be. And it did a really good job.
I thought Mad Pencil here also had a really cool example where she took this image, which is like a realworld image. I don't know if it was actually AI generated or this was a real image.
I'm actually guessing it was AI generated, but she gave it the
04:16
prompt, make the ramen bowl and ramen look like 2D whimsical handdrawn anime illustration, keeping everything else in the image exactly the same. And this was the result.
And I just thought this was a really fun use case to sort of blend real image with cartoony image. And
04:31
again, I do believe this background image is probably AI. I can't confirm 100%, but this would work just as well on a real image.
When she gave ChatGpt the same image and same prompt, it just turned the whole thing into a cartoon. Same with when she used the Quinn image edit, the one that we were just looking
04:47
at a minute ago, Nano Banana was able to figure out how to cartoonize just the ramen. All the rest of the models wanted to turn the whole image into a cartoon.
Really, really impressive. It's also really good at colorizing black and white photos.
I actually found this
05:02
photo online and I fed it into Nano Banana. Gave it the prompt to colorize and upscale the photo.
And here was the output. Again, here was the original.
And here's what it made from my prompt. I then wanted to test combining two images.
So, I gave it this image that I
05:18
found online of Steve Jobs and this image of myself. And I gave it the prompt, make these two men sitting next to each other.
Adjust lighting and colors so it looks like they're in the same room chatting. And here was the output.
I mean, it kind of gave me a different haircut, but other than that, everything looks really, really good. I
05:34
then fed that same image back in and said, "Make it so the two men are sitting on top of a skyscraper with the city below." And this was the output image that I got of me and Steve Jobs sitting on a skyscraper. I then took this same image, plugged it back in, and gave it the prompt, make it nighttime.
05:49
And now I've got a picture of me and Steve Jobs sitting on a skyscraper at night. I don't know why I'm still wearing sunglasses in it at nighttime.
I'm sure there was some bright light shining on us or something, but really, really, really impressive model. Now, as of right now, it hasn't been publicly
06:04
announced who made this model. They've been kind of keeping it a mystery.
Not really, but kind of sort of. It's a mystery.
Most people believe it's Google. And the reason most people believe it's Google is because Logan Kilpatrick, who is the lead of product over at Google AI Studio, tweeted a
06:21
picture of just a banana. and Josh Woodward also at Google.
He's the VP of Google, Google Labs, and Gemini app. Well, he posted this is banana, which I think is a pretty good indicator that Nano Banana came from Google.
But again, I mentioned it's not the easiest to use
06:36
right now. The way people have discovered Nano Banana and started using it was through this site LM Arena, which is a site where you can kind of do a blind test of models and then pick your favorite and then it sort of ranks the models based on which ones were blindly
06:51
picked as the favorites. And if you come to LaMarina, you can use Nano Banana for free, but it takes some work.
So head on over to lmarina.ai. Up at the top, you have three options: battle, sidebyside, and direct chat.
You're only going to get Nano Banana in the battle mode. And well, you may not even get it on your
07:08
first try. It's the mode where it gives you two models side by side, but they're blind.
It's not available in sideby-side chat or direct chat, only battle mode. Once you're in battle mode, you can turn on generate images, and then you can drag and drop any image in.
So, I'm
07:23
going to drop this old image in of me being interviewed by Joe Polish and give it the prompt, upscale the quality of this image, and make it nighttime. Also, give me a baseball cap and submit it.
We'll see. It's going to give me two options.
It's generating it blindly with two different models and we have no idea
07:40
which models it's going to use. It gave me two images back.
This one, which doesn't look great, and this one, which is okay. Doesn't really look much higher quality, but it's okay.
This one is definitely better. We don't know which model is which.
We'll click right is better. And we can see this was seated
07:56
3.0 and this was Gemini 2.0. We did not get nano banana.
So, I have to just do it again until it gives me nano banana. And on my second attempt, we can see that it actually gave me nano banana as one of the options.
And this was the
08:11
image that it outputed. And it did actually make it slightly higher quality.
It made it nighttime in the background. It left everything the same.
And it put a hat on Joe Polish, which makes sense cuz I said, "Give me a hat." The model didn't know which one was me. So, it's very obvious that if I would
08:27
have said, "Put a hat on the guy on the right," it would have done a better job of putting it on me. But, it did do a pretty good job here.
And that's how you can currently use Nano Banana. You can use it directly inside of LM Marina.
You just might need to prompt it a couple
08:42
times until you get the actual model, but there's no limit to how many times you can do it. So, you can keep on generating it and generating it and generating it until you get the exact output that you want.
And well, once you get really good at this and these models get better and better, why are you going to need to go into Photoshop and sort of
08:58
crop around things and composite elements into the background, it's going to get to a point where you just either give it an image or multiple images, tell it what you want it to look like, and it's going to essentially do the photoshopping for you. Pretty sweet.
You know how the startup world works. Move
09:14
fast, validate your idea, raise capital. But here's the thing.
The first AI companies building on Lovable are already getting into Y Combinator and raising a half a million dollar VC checks before most people even have a working prototype. And that's not luck.
09:29
It's because the future of work is shifting. Now it's up to AI to find the right problem to solve.
AI to build and validate your MVP and AI to keep supporting your product as it grows. And that's exactly what Lovable does.
Whether you're a product manager with a game-changing idea, a designer stuck at
09:45
the handoff to dev stage, or a non-technical founder who just wants to get to market before the window closes. Lovable takes your vision and turns it into a fully functional product in minutes.
Landing pages, dashboards, SAS apps, even complex integrations done.
10:01
You skip the dev bottleneck, wow your team or investors, and focus on what matters. Shipping, testing, and scaling.
No coding, no hiring headaches, no long delays. Just describe it and Lovable builds it, keeping your vision intact while freeing you up to move 10 times
10:17
faster. So, if you've been sitting on an idea that can change your career or your life, don't wait for someone else to beat you to it.
Check out the link in the description, start building with Lovable, and join the wave of founders who are proving that in the AI era, the only limit is how fast you're willing to
10:34
launch. Thanks so much to Lovable for sponsoring today's video.
Now, let's get back to it. If you want to use a platform where you can do this kind of editing, Higsfield AI also rolled out this feature.
Now, this video makes it look like they're doing this with videos, but what they're really doing is
10:49
they're changing the starting image and that starting frame is becoming the beginning of a video. So, it's the same thing.
They're editing images by giving them products and things like that and then converting it into videos using things like V3. So, a feature that's available in Higsfield if you want to
11:05
try it there as well. I don't know exactly which model they're using.
If I had to guess, it's probably the QuinnEdit model, but I'm not 100% sure on that. Runway ML rolled out some new features this week, including this Runway Game Worlds beta.
So, it says game worlds employs novel uses of AI for
11:22
nonlinear narrative experiences, meaning every game session you play is generated in real time with personalized stories, characters, and multimodal media generation. So, it appears to create this like textbased game for you.
But as it's doing this textbased game, it's
11:37
generating images to go along with the game. So it's not just reading text.
It's like reading a book but with pictures. And they're making it available right now over at play.runwayml.com.
And I'm not 100% sure if it's fully available or if it's just creative
11:54
partner program that it's available for. I believe it's for everybody cuz it doesn't have any sort of specifics around who can use it.
And when I clicked on this link, it let me into mess around with it. So I can create a game world.
What's the name of your game world? I called mine Apocalyptic
12:10
Wasteland. And then I just clicked autocomplete and it built out the description for me just based on the name of the game.
And then under advanced options here, you can see it plugged in the genre, art style, game mechanics, how do you win, how do you lose, all of that kind of stuff. And now
12:26
we can just click next. And it's generating an image for our game.
And here's what it generated. So, let's click create game world, and we're in.
You must journey through a world reclaimed by colossal otherworldly flora and the lingering emotional echoes of a cataclysm, seeking scattered ancient
12:42
relics to mend humanity's spiritual wounds. All right.
What is the name of your character? Let's go, Mattius.
Nice to meet you, Mattius. We can now generate your avatar.
Upload an image for reference, or feel free to skip this step if you prefer not to. Yeah, sure.
Let's upload a reference image. Give it
12:57
my image here, and let's go ahead generate with image. Here's my game avatar that looks like me but in this world.
Let's click continue. And now it's creating your world.
My main quest is unite the heart shards. Seek and activate the ancient relics to mend humanity's spiritual wounds.
And so we can see as we play through the game, it
13:13
almost looks like a comic book. Your throat burns.
Each breath a dry rasp against the still humid air. I start with plus 100 vitality, plus one water supply.
The giant twisted roots of a verdant scar loom above, offering no shade. A distant shimmer catches your
13:29
eye. A mirage of clear water in the dust choke distance.
So, you can see how this is like basically the same concept as a textbased like adventure or fantasy game, except we get images to go along with it, so we don't have to like try to visualize it all in our head. A faint melancholic hum resonates from a crack
13:45
in a nearby root, revealing a glimmering crystalline fragment. The shard pulses with a strange energy tugging at your spirit while your body craves water.
Will you pursue the possible water or investigate the glowing shard? And then it gives me some options here or I can type in my own.
Let's examine the shard.
14:02
You move towards the pulsing light. The thirst in your throat temporarily forgotten.
Now my water supply is minus one. As your fingers brush the shard, a wave of profound sorrow, then fleeting joy washes over you.
I mean, this is really actually kind of fun, especially since it's like injecting me into the
14:17
game and I'm seeing the images as I go. The shard vibrates.
A faint whisper of laughter and song echo in your mind. Brief glimpses of a world that once was.
It also seems like it could be a colossal time waster, but it's kind of fun. So, you get the idea.
I'm not going to keep on playing here on this
14:33
recording cuz I will probably just get lost in this world and get further and further sucked in, but this is super cool. Not like the most useful productive thing ever, but if you just want to kill some time and get lost in this fantasy game, this seems like a great way to do it.
Since we're talking
14:48
about Runway, they also rolled out some like quality of life features this week. Like they're adding third-party models directly inside of their chat mode.
You'll be able to use VO3 inside of Runway. They've updated the way the chat looks so your chat stays on the left and all your generations on the right so you
15:04
can, you know, see them both side by side as you have a conversation and improve your videos. And they're also bringing the Gen 4 image turbo mode to all users, which is a really, really fast model.
11 Labs dropped another feature update this week as well. This
15:19
week they launched a new video to music flow in 11 Lab Studio. In one click, 11 Music model generates a custom soundtrack based on your video's context.
After adding music, you can layer in voiceovers and sound effects directly inside of the studio. So, to
15:34
me, this seems like it could potentially replace a lot of those like stock music websites. No, no, no, no, no, no, no.
Wait, wait, wait, wait, wait, wait, wait, wait, wait. You give it a video and it figures out based on what's going on on the video
15:50
what kind of background music will go with that video. Here's some examples like the music they put over this like raining on a leaf here.
[Music] I believe they also added the like rain sound effects using 11 Labs. Here's
16:07
another one. a sort of synth wave cyberpunk song for that video.
Grand awinspiring outdoor scene
16:22
[Music] and it figures all that out by just giving it a video. It's available to try over at 11labs.io/studio/videotomusic.
But here you could just upload a video. Here's an AI generated video of me
16:38
dancing. I wonder what it's going to do for this one.
I'll just drop it in here. It's watching my video.
And here's what it gave us. So, it knew it was kind of like a 70s vibe and it gave it sort of 70s vibe
16:54
music. Pretty cool.
What do you think? Are stock music websites in trouble?
Google also updated their AI mode in search this week. Now it's got more agentic features.
So the example they give here is you can now ask about getting a dinner reservation with
17:10
friends that includes multiple constraints and preferences like party size, date, time, location, and preferred cuisine. Now this is a feature that needs to be turned on.
So if you have access to Google Labs, you'd click on this little search labs button and then make sure you have AI mode turned
17:27
on and then you can try AI mode. But let's try something like find me a nice restaurant in downtown San Diego for Saturday night that serves sushi and can seat up to 8:00.
We want to eat sometime between 6:00 p.m. and 7:00 p.m.
and it should be somewhat close to the ballpark. So, we can walk to the game.
17:42
So, if I give it all of that information and send it, let's see what happens. And we can see that it found me a bunch of sushi restaurants with high ratings downtown.
It looked for them near Petco Park. And then down at the bottom, to ensure seating for a group of eight,
17:58
it's recommended to make a reservation, especially since you plan to dine on a Saturday night. So, I mean, it did find all the information for me, but it didn't like take a ton of actions on my behalf.
It just sort of pulled all that information together for me and gave me a good response. There's also a new like quality of life feature inside of Google
18:15
Docs where you can actually have it read the doc to you with a AI voice. So, if I open up a Google Doc here and I come up to my tools menu, there's a new option for audio and I can say listen to this tab.
Now, I believe this is only rolled
18:32
out on the pro and ultra plans. Like, I think you need to have one of Google's paid plans to be able to use this.
I'm not 100% sure, but I believe I remember reading that it's only on the paid plans for right now. But, we can go ahead and click listen to this tab.
The practicality index, a blueprint for benchmarking the real world utility of
18:49
large language models. Introduction.
the practicality gap in LLM evaluation. So, we can see that it's actually a pretty human sounding voice.
It doesn't sound like an old robotic AI voice. It sounds like a real human reading it.
And pretty good quality. We also have the
19:05
ability to change the voice to a narrator, an educator, a teacher, a persuader, an explainer, or a coach. So, if I switch it to coach for instance, select that.
The practicality index, a blueprint for we get audio that is a slightly different voice. So, a pretty handy
19:21
feature if you want your documents read to you. There's also some new AI features inside of Adobe Acrobat, which can help you do cool AI stuff with your PDFs.
I do believe you have to have like a Google Cloud membership to be able to use this, but they have a new feature
19:37
called PDF spaces, which allows you to have conversations with your PDFs. It'll allow you to easily pull data and information from PDFs and things like that.
So, if I bust open my Adobe Acrobat here, you can see over on the left there's a PDF spaces link. If I
19:52
click into here, this is the new PDF spaces feature. I'll pull in this ebook called Mastering LLM as a judge.
Just drag it right in here. Opens it up in Adobe.
I'm going to click add to PDF space. And we can see that similar to like notebook LM, it pulled in a bunch
20:08
of information that it grabbed from this PDF here, including even making like a little table of details from the PDF. And then there's various assistants.
We can see choose an assistant. Down on the bottom, you have an analyst, an entertainer, and an instructor.
And I could do things like put it on
20:23
entertainer. And it seems to reconfigure all of this content here for a more entertaining read.
LLMs can be like a mysterious magician. Impressive, but you're left wondering how they pulled off the trick.
So, just another way for you to interact with your PDFs directly
20:39
with inside of Adobe and quickly pull out details. especially good for longer PDFs or more complicated technical PDFs.
This week was also Google's annual made by Google event. This is like Google's
20:55
annual hardware event where they announce all the latest things they've done with like their Pixel phones and things like that. And of course, since it's Google, everything rolled out with AI.
Now, I'm not going to get super deep into everything they announced at this made by Google event because there was a
21:10
lot and I don't really think we need to get into the details about what all of these various phones and watches do. But I will give you a quick TLDDR of the various like AI stuff that they've added into these gadgets.
And the people that got to demo this stuff are absolutely
21:26
raving about how good the AI has gotten inside of these various Google Pixel products. So, they showed off their new Pixel 10 phone.
It's got a new chip in it, a Tensor G5 chip, and it was co-designed with Google Deep Mind with the Gemini Nano model sort of at the
21:44
forefront. So, these phones are basically built from the ground up for ondevice AI, which is something that I think a lot of people really, really want.
They don't want to send their prompts off to the cloud and wait for a response. If we can get more of this on device, that's obviously better.
And
22:00
these phones have a feature called Magic Q. And this isn't an app.
It's something that's just always running on your phone. And it's basically letting the AI see what's going on on your phone so it can give you advice and feedback and
22:16
help you jump around to different things on the phone. So, one of the examples they gave in their demo was somebody texted them and said, you know, where was the location of that concert we're going last week?
Well, in their text message app, the AI read the text
22:31
message, went and figured out where the location of whatever concert it was was, and then put it into the send box so they can just press one button and send the location without the user actually having to go and look around in their calendar and, you know, find locations and stuff. Google just sort of connected
22:47
the dots for them. So they can just one-click reply with the answer because the information was already on the phone and the AI was able to sort of locate it and pull it forward for them.
And this feature is sort of crossplatform, cross app and is just understanding the
23:03
context of what's going on in your phone. And this is something I wish iPhones have cuz my main daily driver phone is an iPhone and I really want this feature.
It looks really good. They also show off this feature in the camera app where it can look at your photo and actually give you advice on how to get a
23:19
better photo while you're actually in the process of taking the photo. You can see ask the person to look at the camera to create a more engaging and personal photo.
So, it's giving you advice on what to tell the person to do. Move closer to the person and the sculpture.
Lower the camera slightly. And it just gives you instructions until it dials in
23:37
on what it thinks will be the best photo for you to take. And here was the resulting photo, which looks like a pretty decent composition.
They also showed off that the phones can have realtime voice translation. So you could talk to anybody in any other language and it will automatically translate it
23:52
on the fly in real time. That's insane.
Voice translate works when translating from English with Spanish, German, Japanese, French, Hindi, Italian, Portuguese, Swedish, Russian, and Indonesian. That is just wild to me that I can just call somebody in Russia and
24:07
have a conversation with them. They'll hear me in my voice speaking Russian and I'll hear them and their voice sounding like they're speaking English.
Crazy. They also showed off that you can edit images with just text prompts.
So you can see remove the glare right in the photo, etc. And then it goes and fixes
24:25
the photos based on their text prompt. And a lot of people believe that this is using that new nano banana technology from Google, but that hasn't been confirmed.
It just seems like that's most likely. It can restore old photos and colorize them just like what we were
24:42
looking at Nano Banana. And it gives those like Photoshop style editing features to change the image however you want directly inside of your phone.
Like a lot of the features they showed off could cause me to switch to the Pixel phone as my full-time daily driver cuz
24:58
it's all really cool stuff. They also rolled out the Pixel Buds 2A which allow you to talk with Gemini even without taking out your phone.
These to me look just like the little earbuds they have in the movie Her, which is probably kind of what they're going for. And all you
25:15
have to do is say and you're talking with Gemini. They showed off the Pixel Watch 4 and it now has a personal AI health coach.
It provides proactive fitness and sleep coaching and ondemand guidance. And of course, it also has Gemini, your on- wrist AI assistant.
So,
25:31
a lot of the stuff you can do with Gemini, you'll be able to do directly from these watches. and the same AI features and health coach that went to that watch, they're also coming to Fitbit.
So, if you don't want to get the full watch and the full watch experience, but you do want that fitness and health coach, well, you'll be able
25:48
to get it inside of the new Fitbits pretty soon. They also announced a Gemini for Home.
So, your Google Home devices are going to have Gemini built into them. And they made some updates to Gemini Live where it will actually give you on-screen cues.
Like, they took a picture here and said, "What seasoning
26:03
do I use?" and it actually circled the spot on the picture for him. They showed the camera some like wrenches and asked which one to use and it highlighted the wrench to use.
They're also making it connect to more Google apps, Google Calendar, Keep and Tasks are both available today and soon. Messages,
26:20
phone and clock apps. So, a lot of the apps that are on your phone that are in the Google ecosystem are all going to start to tie together a little bit better using AI.
And that's pretty much the quick TLDDR of all of the cool sort of AI gadgetry that came out of Made by
26:35
Google. I know I've shared a lot already, but there's a few smaller items that I want to put on your radar and share with you.
So, let's jump into a quick rapid fire and I will fire off the rest for you real quick.
26:51
We got a few new large language models that you can use this week. Deepseek has a new version 3.1 with 685 billion parameters.
So, it's a big big model. You can make it think or turn off the thinking mode.
It's apparently really
27:06
really fast. And you can use it over at chat.deseek.com.
I haven't personally tested this one yet, but I've heard very good things. I've heard it's a very very impressive model.
It's got 128,000 token context window and seems to score pretty well on all of the various benchmarks. So, looks
27:24
like another solid model. Another model came out of China this week as well from Bite Dance, the company behind Tik Tok called Seed OSS 36B.
So, it's a 36 billion parameter model, but it has a 512,000
27:39
token context window, which is really, really impressive for an open source openweight model. Again, I haven't tested this model myself, but I want to make you aware of all of the newest large language models that are dropping.
And finally, the other model that dropped this week was from Nvidia called
27:56
Neotron Nano 9B version 2, which also has the ability to toggle on and off reasoning. The model also appears to be fairly open.
So, it says here, "The Nano 9B V2 model is released under the Nvidia open model license agreement. It's
28:12
designed to be permissive and enterprisefriendly. NVIDIA explicitly states that the models are commercially usable out of the box and that developers are free to create and distribute derivative models.
So you can use this basically as if it was an open- source openweight model. We got a few
28:29
updates out of OpenAI this week. Apparently they made GPT5 warmer and friendlier based on feedback.
So there's small genuine touches like good question and great start. Not flattery but just a little bit friendlier I guess.
And the comments on this thread are, "Nobody
28:45
wants that. Just let us pick between our models again." It's kind of funny.
OpenAI just can't win. Everybody hated the fact that there was so many models and they had to pick which model they needed to use and nobody knew which model to use.
So, they rolled out GPT5 which basically picked the model for you
29:01
and everybody hated the fact that their old models that they loved went away. OpenAI brought back the old models and then started to tweak GPT5 to be more friendly like the old models and everybody went, "No, don't do that.
Just give us GPT5 the way it is, but also let
29:17
us have our old models." So, it's like OpenAI just can't seem to win. Interestingly, even though GBT5 is only a couple weeks old at this point, we're already talking about GPT6.
Sam Alman is claiming that it's coming faster than most people realize. And it seems like a
29:33
big focus of this one is enhanced memory. Sam Alman didn't give a release date for his company's next artificial intelligence model, but he made clear that GPT6 will be different and that it will arrive faster than the gap between GPT4 and GPT 5.
It won't just respond to
29:49
users, but will adapt to them and allow people to create chat bots that mirror personal tastes. He sees memory as the key for making chat GPT truly personal.
It needs to remember who you are, your preferences, routines, quirks, and adapt accordingly. So, the future models of
30:04
chat GPT, we actually talked about this when I made my videos about GPT5. The future is everybody's going to have like their own version of chat GPT.
when I log into my chat GPT account and have conversations with it, it's going to respond to me completely differently than how it would respond to you inside
30:20
of your chat GPT account because it's going to remember what you like, your preferences, the way you want it to respond to you and things like that. And they're just going to get more and more and more tailored to exactly what you want chat GBT to be for you.
And it
30:35
sounds like that's the big goal with GPT6. Take this with a grain of salt, but Sam Alman also believes that we're currently in a bubble in the AI world.
He said, "If you look at most of the bubbles in history, like the tech bubble, there was a real thing. Tech was really important.
The internet was a
30:51
really big deal. People got over excited.
Are we in a phase where investors as a whole are over excited about AI?" My opinion is yes. Is AI the most important thing to happen in a very long time?
My opinion is also yes. So, he's kind of comparing it to the internet bubble.
does think that it will
31:07
probably crash and valuations will probably go down. At least that's my take on this.
However, even though there was the internet bubble and then it crashed, it's not like the internet went away and there weren't still some big winners to come out of it. And well, this will likely be the same.
Again, that's my take on what Sam's saying.
31:24
Microsoft is rolling their co-pilot directly into Excel. I have the newest beta version of Excel and I don't have it yet.
Even though the report here says that I should have it in my account by now, but apparently you'll be able to type in a command like equals copilot and then have it actually do a bunch of
31:41
stuff for you. Copilot list airport codes from major airports in E3.
E3 is USA. And you can see it automatically filled in a bunch of airport codes.
They change it to Australia. Fills in a bunch of airport codes from Australia.
They built that all using Copilot. Here's
31:56
another one where they put a bunch of feedback in a list and then they gave it a prompt of what is the sentiment of the comment over on the left and then it figured out if it was positive or negative and they were able to drag and drop that sort of code all the way down
32:12
and it automatically, you know, figured out the sentiment of all of these comments. So, seems like it'd be a pretty handy feature once we can actually use it.
Their article here does say it's rolling out to beta channel users with a 365 Copilot license, which
32:29
I do have. I was trying to test it on the Mac version, and I do have 16.101, and it didn't work for me yet.
So, I don't know if I was doing something wrong or if it isn't fully rolled out yet, but it looks like it will be pretty cool once it's available. Grammarly is
32:45
rolling out some new AI features which will help with teachers grading and students writing. Students today need AI that enhances their capabilities without undermining their learning.
Grammarly's new agents fill this gap, acting as real partners that guide students to produce better work while ensuring they develop
33:01
real skills that will serve them throughout their careers by teaching students how to work effectively with AI. Now, we're preparing them for a workplace where AI literacy will be essential.
So, we can see in the demo here, it added like a sidebar and I believe this is like helping them grade
33:17
this paper. see predict a grade and it's reading the paper and I guess helping a teacher grade the paper.
So, some pretty interesting features out of Grammarly. Not something I've tested myself yet.
Meta is expanding their AI translation feature and making it so your Instagram
33:32
and you know Facebook reels and things will automatically be translated. YouTube has already rolled out a similar feature, but it's now coming to the meta platforms.
It says, "Meta translations uses the sound and tone of your own voice so it feels authentically you. And the lip-s syncing feature matches your
33:48
mouse movements to the translated audio so that it looks like you're really speaking the language in the real. And from what I gather from their article, it looks like it's already rolled out.
Not something I've tested yet, but I am actually spending more time on Facebook. So if you're a Facebook person, you
34:04
could feel free to follow me over there as well, cuz I have been posting there a little bit more often lately. Just a random interesting factoid for you.
And then finally, since we're talking about Meta, it's expected that they're going to show off their new smart glasses at Meta Connect this year. And the rumor is
34:20
they're going to have a little display in one of the eyes, so you can actually have like a heads up display inside of your glasses. Now, whether they're going to be able to keep them in the same form factor and keep them just as light, I don't know.
but they're expected to be released somewhere around 800 bucks and
34:36
have a screen for apps and alerts on one lens and a wrist accessory that can help you control the glasses. Unfortunately, I'm not making it out to Meta Connect this year.
I'm going to be at New York City at a YouTube event instead. But I am really, really excited to see these
34:53
in action. I saw Google's version of them, you know, six or seven months ago in person and they blew my mind.
If these come anywhere near Google's versions, they're going to be a hit. Everybody's going to want them.
So, really excited to see what they do with this. But that's what I got for you today.
Again, it was a lot of news this
35:08
week. A lot of interesting stuff, a lot of things out of Google, some cool image editors.
Hopefully, you enjoyed this video. You feel more looped in on what's going on with the AI world and what you're capable of doing with AI as of right now today.
That's my goal with this channel. And if you like videos like this, make sure you like this one
35:24
and subscribe to this channel. and I will make sure that more updates like this one as well as tutorials and other interesting news from the AI world shows up in your YouTube feed.
But once again, thank you so much for hanging out with me and nerding out with me in this video. I really, really appreciate you tuning in and uh trying to stay on the
35:41
bleeding edge of AI with me. It's been really fun and I hope you're having fun as well.
So, thanks again and hopefully I'll see you in the next one. Bye-bye.
Thank you so much for nerding out with me today. If you like videos like this, make sure to give it a thumbs up and subscribe to this channel.
I'll make sure more videos like this show up in
35:57
your YouTube feed. And if you haven't already, check out futuretools.io where I share all the coolest AI tools and all the latest AI news.
And there's an awesome free newsletter. Thanks again.
Really appreciate you. See you in the next one.