We have so many AI news stories to cover today, let's jump right in! The first story is Elon Musk's prediction for AGI. Elon Musk hasn't been known for being the most accurate when it comes to timelines and predictions, but a lot of

people seem to think this is the general timeline for AGI, say 2026 at the latest for AGI. So what do you think? You think that's accurate, 2026 at the latest? Let me know in the comments. So moving on from Elon's prediction, we're going to be

moving to humanoid robots now. Brett Adcock, the CEO of Figure Robotics, has released a new video update, and it is stunning. These are Figure O2 robots working in the BMW Factory, and they have sped up their movement and accuracy greatly, according to Brett. Figure 2 is

now an autonomous fleet, 400% faster with seven times higher success rate. So check this out. This is the Figure robot working in an actual factory, in a BMW factory, that is. So here we can see the exact movements between all three of these robots. I think they look really

really cool. Obviously this one is sped up, but these are robots in the field working right now, and it's clear that they've really been able to speed up the robots. There's not only a physical limit in how fast these robots can run based on the hardware that they're using, the

actuators and so on, but there's also a safety limit because if they're moving too fast, any mistake is going to be very costly. I think this is so cool. I've said it before, I cannot wait to have a humanoid robot in my house doing

different chores for me, helping me throughout the house. I am ready for that future quickly. I just want to mention that we're doing another giveaway. We are giving away this Dell 34-inch UltraSharp Monitor and it is phenomenal. Dell and Nvidia have provided this to us and now

I'm going to give it away to you. All you have to do is subscribe to the Forward Future newsletter, go through the giveaway link down below, subscribe, and you have a chance to win. It is North America only for this giveaway, but we may change that in the future for future

giveaways. So if you want to stay up to date on all of the latest AI news, if you want to get deep dives on AI topics and really just stay ahead of everything trending, definitely check out the Forward Future newsletter and now you have an even better reason to subscribe.

Next, from robots now to text-to-image models. Black Forest Lab has released some new tools for us. Here's Black Forest Lab's post: "We are excited to release Flux One, a suite of models designed to add control and steerability to our base text-to-image

model. Flux One enables the modification and recreation of real and generated images. So this means you're going to be able to edit images rather than just create images from scratch with their new toolset. And remember, Flux is open source, so you can download it, you can

play around with it on your computer and you can also find it powering a lot of different model providers' text-to-image capabilities, such as Midjourney. STW uses Flux and I believe Gro uses Flux as well. The first tool they're releasing is Flux One, a state-of-the-art inpainting and out-

painting models enabling editing and expansion of real and generated images given a text description and a binary mask. Here's another example of it, so you just have this small image of somebody's eye and then we can expand it out to get their whole face. Next, they have Flux One.

depth models trained to enable structural guidance based on a depth map extracted from an input image and a text prompt, so you can actually extract the depth as seen here from images. They have Flux One canny models trained to enable structural guidance based on canny edges

extracted from an input image and a text prompt, so you can see all of the edges of the cat's face, the outfit it's wearing, everything. And then last, Flux One Redux, an adapter that allows mixing and recreating input images and text prompts. So here we have this original

cartoon-like duck. Then we have it as a fine glass sculpture on a shelf, a simple 3D model, an actual duck on a desk, and an epic anime art flying through clouds. Really, really cool. I haven't really done a full Flux tutorial. Let me know if you

want to see that in the comments. All of this is open source and available on Hugging Face right now. Next, from text to image, let's move back to large language models. Quen has been on a tear recently, and now they have extended their context window to a million tokens, and not only

that, they have greatly increased the inference speed on those million tokens, and today they've announced Quen 2.5 Turbo. They've expanded from 128k to a million token context window, which is about a million English words, and it achieves a 100% accuracy on finding

information from those million tokens and inference speed on those million tokens dropped from about 5 minutes all the way down to 68 seconds, which is a 4.3x speed up and it's staying at the same cost, so you get all of that additional functionality at the same cost now.

from text to speech, 11 Labs now allows you to create conversational agents. As a reminder, 11 Labs allows you to clone voices, create voices, all with AI. According to this TechCrunch article, the company announced that users can now build complete conversational agents on

11 Labs developer platform with customizable variables such as tone of voice and response length. Users can add their own knowledge base, like a file URL or text block, to power the conversational bot. Plus, they can integrate their own custom LLM into the bot, so definitely something fun to play

around with. You can probably recreate advanced voice mode from ChatGPT pretty easily, powering it with an LLM plus the 11 Labs new bot, so definitely check that out next. It looks like the company behind Pokémon Go, the hyper-popular game for the last decade plus, is using all of

that data that it collected from its users and has published a geospatial AI model. Now, likely Pokémon Go players did not realize that they were going to be doing that, but that's kind of the cost you pay to play the game. So, according to this article in The

Verge, scans of the world from Pokémon Go and Ingress are the backbone of Niantic's AI model, which aims to navigate the world like ChatGPT spits out text. Now, there are only a few geospatial models, and Tesla is one of the leading

companies because they have all of that video data from all of their automobiles, but of course now that everybody's been taking pictures and video with Pokemon Go, the company behind that game, Niantic, has all of that data to train a model on. According to Niantic's blog post over

the past five years, Niantic has focused on building our visual positioning system, which uses a single image from a phone to determine its position and orientation using a 3D map built from people scanning interesting locations in our games and scanning. So what do you think

they're actually going to do with all of that data? I'm not sure, likely sell it, but who knows? Let me know in the comments what you think. Next, Suno V4 is here. So according to their blog post, V4 is here. We're super excited to introduce

V4, enabling you to make any song you can imagine with better audio, sharper lyrics, and more dynamic song structures. Try it now on sun.com. Here are some new features powered by V4: Remaster, upgrade your tracks to V4 quality, Lyrics, create higher quality lyrics for your

songwriting, Cover Art, fresh designs to complement your music's vibes. Let's take a look at this video. Father, how have you been? Did you love his name? What is it? Two decades? I [Music] just honestly, that could be any current pop song, that kind of sounds like a

Billie Eilish song. So really cool. I know on this channel I don't really play around with music generation much, it's just not something I'm super into, but I do find it really interesting nonetheless. I know Matt VidPro AI did a whole test of it, so

definitely check that out on his channel all right next Gemini is getting an update one thing that was desperately missing from the Gemini model was its ability to remember things now if you think about ChatGPT when you're having a conversation with it it picks out

important things to remember about you that will inform future conversations and make them more relevant to you but now Gemini can do the same thing so as you're having a conversation with it it will pick out memories to create based on your context and based on what it

thinks will be relevant in the future according to their post rolling out starting today you can ask Gemini Advanced to remember your interests and preferences for more helpful relevant responses easily view edit or delete any information you've shared and see when it's used and as a quick reminder we

just covered Gemini because they topped the leaderboards on the LLM leaderboards so apparently they they are making huge strides in quality and feature sets so great job to Google with Gemini and the next one's a quick update ChatGPT can't be left out here's

another feature update from OpenAI if you're using ChatGPT on desktop you now have advanced voice mode which small update but very welcome I don't use advanced voice mode that much I've actually found it to be quite buggy and it repeats itself but it's fun to use

every once in a while but don't get me wrong I use ChatGPT and Perplexity every day all day so small update there good job next we finally have an open-source thinking model and by thinking I mean test time compute that means that this

model can think and take its time and do reflection and Chain of Thought and then give you the best answer possible and again like I said it's all open source it's all free and it's by Deep Seek I already made a full video about it so

I'm not going to go too deep into this but an open-source thinking model yes all day yes please you could play around with it for free right now but you can't quite download it yet they said it's coming soon next moving on from LLMs we

now have a major update from Google DeepMind they have released AlphaCubitt which is a way to predict errors in Quantum Computing which is kind of insane to think about I actually wrote an entire tweet thread about this I'll drop the link down below the gist is one

of the biggest hurdles in using Quantum Computing for real world applications is that they're very very sensitive even the slightest bit of noise will make them error but if we can predict those errors with accuracy we can obviously avoid them or at least account for them

and that is what AlphaCubitt promises to do it is using artificial intelligence to predict the errors inside of quantum computers so this might be that last unlock necessary to see quantum computers in everyday life and of course there are a lot of consequences that

come from unlocking quantum computers and actually being able to use them in real world applications I haven't touched much on Quantum in this channel because it doesn't always overlap with artificial intelligence but it is fascinating to me and in fact I'm going to be doing an interview with the head

of the quantum department at IBM and so IBM is one of those companies that has been on the forefront of quantum computers with Google so that interview should drop in the next few weeks definitely check it out and now from Quantum to GPT-4, a major update this is

very welcome I think people don't appreciate how good this is going to be GPT-4 got a big update now it is going to be much better at writing according to their post the model's creative writing ability has leveled up more natural engaging and tailored writing to

improve relevance and readability it's also better at working with uploaded files providing deeper insights and more thorough responses one of the biggest gripes I've always had about large language models is they don't write very well it's usually very boring and they

use a lot of fluff words and actually Matt Schitt responded with excited to delve in because seemingly every single large language model used the word delve and that's kind of what we're talking about here they are very repetitive very boring and now I'm excited to test this

out so that's it for today if you enjoyed this video please consider giving a like and subscribe and I'll see you in the next one

AI News: Musk Says AGI 2026, Open-Source Q*, Flux.1 Updates, Quantum AI, and more!

Artificial Intelligence (AI), Open Source, Generative Art, AI Art, Futurism, ChatGPT, Large Language Models (LLM), Machine Learning, Technology, Coding, Tutorials, AI News, and more

** Exclusive Pinecone Partner **

Matthew Berman

Elon Musk's AGI prediction and humanoid robot advancements

Elon Musk predicts AGI arrival by 2026 at the latest.

Discussion on the accuracy of Musk's prediction.

Figure O2 robots in BMW factory show significant speed and accuracy improvements.

Video showcasing Figure robots' movements in a real-world factory setting.

Discussion on physical and safety limitations of fast-moving robots.

Announcement of a giveaway for a Dell monitor.

Black Forest Lab releases Flux One, a suite of models for image editing and creation.

Flux One allows modification and recreation of real and generated images.

Description of Flux One's inpainting and outpainting capabilities.

Flux One's depth and canny models enable structural guidance based on depth maps and edges.

Flux One Redux allows mixing and recreating images based on text prompts.

Open-source availability of Flux One on Hugging Face.

Quen expands context window to a million tokens and improves inference speed.

11 Labs allows creation of conversational agents with customizable variables.

Niantic's geospatial AI model and music generation advancements

Niantic uses data from Pokémon Go and Ingress to create a geospatial AI model.

Suno V4 is released, offering improved audio, lyrics, and song structures.

Suno V4 features include remastering, lyric creation, and cover art generation.

Gemini update includes improved memory capabilities for more relevant conversations.

ChatGPT desktop gets advanced voice mode update.

Release of an open-source thinking model by Deep Seek.

Google DeepMind releases AlphaCubitt to predict errors in quantum computing.

GPT-4 receives a major update, improving creative writing abilities.

GPT-4 update focuses on more natural, engaging, and relevant writing.

Improved handling of uploaded files with deeper insights and thorough responses.