Here’s this week’s AI news you might have missed. Please note that I record these on Thursdays, so any news that came out late Thursday or Friday is not included, but will be in next week’s video. Discover more from me: 🛠️ Explore thousands of AI tools: 📰 Weekly Newsletter: 🎙️ The Next Wave Podcast: @TheNextWavePod 😊 Discord Community: ❌ Follow me on X: 🧵 Follow me on Instagram: Today’s Video Resources: Runway Dashboard: Runway Assets : Luma Creations: ElevenLabs Voices: Voice Isolator: Suno App:…
My workplace blocked chatgpt and anything that is not Microsoft or copilot. Do you have any info on copilot?
Thanks for getting the word out about Bring Your Own Model! 😃
Here's a ChatGPT summary:
– Gen 3 Access was made publicly available, but only for pro users of Runway.
– Gen 3 is currently the best text-to-video generator, while Luma AI is preferred for image-to-video.
– 11 Labs added famous voices like Judy Garland, James Dean, Bert Reynolds, and Sir Lawrence Olivier to their reader app, with permission from the estates.
– 11 Labs released a new voice isolator feature to clean up background noise in audio.
– Suno released an app for iOS that mirrors the functionality of their web app for music creation.
– Meta released research on 3D Gin, a text-to-3D image generator.
– Kiyotai released an open-source voice model to compete with GPT-4O's advanced voice.
– Intern LM2.5, an open-source large language model with a 1 million context window, was made available on Hugging Face.
– Brave Browser now allows users to bring their own model into the browser's AI, Leo AI.
– Proplexity updated their Pro Search with multi-step reasoning and better math and programming capabilities via Wolfram Alpha.
– Apple is getting an observer role on OpenAI's board.
– The Center for Investigative Reporting is suing OpenAI and Microsoft for copyright infringement.
– Mustafa Soliman made controversial statements about the fair use of content on the open web.
– Cloudflare introduced a feature to prevent AI scrapers from accessing your site.
– Figma announced it would use user content to train models but will offer an opt-out option.
– YouTube now allows users to request the removal of AI-generated content that simulates their voice or likeness.
– Instagram changed its AI labeling from "made with AI" to "AI info."
– Grok 2, an improved version of the AI model, is expected to be released in August.
– Rumors suggest Apple might partner with Google Gemini for chat features.
– WhatsApp is reportedly adding a feature to generate cartoon versions of user images.
– A new company is developing smart glasses similar to Meta Ray-Bans but with ChatGPT 40 integration.
– Open Television demonstrated immersive robot operation from 3000 miles away using Apple Vision Pro.
– Main message: Despite a slow week due to the Fourth of July, significant advancements and updates in AI technology were made, including new tools, features, and research across various platforms and applications.
Actually insane how much is going on in the AI space all the time.
Matt waiting for you next video, it is has been 6 days still no new video. Is everything fine?
This channel is really awesome, never thought I would wait and care for any youtube video this much
For the algorithm!
🥰
I scrape and even pirate any data I want that I can get. I don't care about copyright laws, at all. If you put it out there, or even let it get out there, it's fair game as far as I care. Are libraries evil because they didn't seek permission from every author of every book they provide to the public for free? Especially if you deliberately publish information, you only have the right to ask politely for it to be used/not used in any particular way, but no one is under any moral obligation to oblige you in that. This includes copying/distributing… even selling IMO. What is criminal is punishing people for "copyright violation". If you don't want people using/modifying/distributing your data in some way, don't publish it, protect it. Simple.
Like always, very helpful information.
nice video good job Matt
Thanks
Calude is amazing, but so limited. You get timed out so fast while Open AI you can go all day and never time out
I love heading about AI and robotics
i'm so depressed
What would you prefer? An AI as your friend? Or someone to have a cup of tea with on Saturday afternoon?
OMG if they made Moshis eye red and made the voice male, it could say "I'm sorry dave, I'm affraid I can't do that"
We should, as much as possible, aim to negate copyright… Always…
When AI leaves the screen it will then be useful
regarding 3d AI generation: the deep learning chair at TUM (Technische Universität München) where I studied has been doing research into this for years, and though Meta has of course more resources, let's not forget that StableDiffusion came from the sister university of LMU. This is to say, what these researchers have is probably close to state of the art research. And let me tell you, as someone who works in gamedev: 3d AI generation is lightyears away from being used in games. a lot more goes into 3d modelling than just making realistic proportions. Any 3d artist has that in the bag. The biggest part of the 3d pipeline is retopologizing (turning a high poly model into a low poly model that is actually usable in a game and ensuring that the topology is good for animation and deformation). Proper topology is one of the hardest problems to solve (there exist many retopologizing algorithms, none of them are even remotely good enough), and the very nature of the problem is what is going to make it nearly impossible for AI to handle: Good topology is inherently specific to the rig and the use case of the model. There are literally no hard rules. This means, Ai wont be able to properly learn patterns bc those patterns simply dont exist. Prediction: there will be no AI to create game-ready 3d models used widely in the next 10 years.
You can't copyright information. Once you put it out there, it becomes public use. So AI models can take the information and use it to train ai systems, which generate new content and isn't a violation of copyright. Matt, you literally built your channels income off of compiling information that was created by other companies. AI does the same thing and they can take your videos to glean the most important AI news and redistribute that information how they see fit. This is not the same thing as reposting your video to make money for themselves. It's how the world has always worked but instead of humans doing it, a computer is doing it with much more efficiency.
Hey, Matt, you didn't give Moshi a chance. Did you ever hear Moshi play chess? It was hilarious.
Oh great. That last robot is a globalists dream. Outsource work to third world countries and pay 20c an hour. Have fun with that future.
well, Mustafa didn't say you can do everything you want with it, he said "fair use"
Good stuff as usual, Matt! If I may push back a bit on the idea of use of publicly available content (IP):
Here's the beginning of an article I wrote on that topic:
On the surface a person might believe, “If we didn’t have intellectual property laws, it would open the floodgates for copycats everywhere, reducing incentive for people to create. We would have Chinese phones that look like iPhones and poor Apple! Innovation would be lost because no one needs to innovate in order to get sales.”
I get and understand that some entities will copy instead of innovate. I get it that many believe the government is the only way to prevent this scenario.
Nine points that weigh in favor of reforming or even completely removing intellectual property laws
(1) How many of these “copycats” will do this and be successful at it? Not everyone will do this or even want to. The fact that we have anything new in the world is a testament to that.
(2) Reputation, brand, first to market, quality, and integrity. These factors all influence the market and have value.
(3) Consumers win (price, etc.). Example: If a Chinese company succeeds in making an iPhone copy equal in every way (price, perceived quality/looks, AND quality), then one would assume they would try to sell it cheaper to draw away Apple customers. This is a big “IF” because Apple has agreements with their suppliers, economies of scale, etc. where it might be difficult for most companies to exactly dupe the iPhone in all ways, including quality, and afford to sell it cheaper. BUT let’s say they can …
If you want to read more on this, pls google "can you own an idea by scott swain"
Hey brother how can I contact you?
Is no one going to point out that 7 532 * 7 is not 53 203, it's 52 724 from the Moshi demo @7:02 ?
She didn't say don't take it by force so…..
Hello great video. I hope you had a great 4th of July.
Amazing vid we need more like this
Too bad those apps are only available in the Apple app store. I got excited and went to download them for my Galaxy phone, and no love. 😢
I checked, and you don't mention that. You just said the apps were available on the "App Store". Disappointing.🫤
Why does it make me uncomfortable when 🍎 and ms are “cooperating” with OpenAI? I love them… sort of but…. ??? Hmm?🤔. 2020. Antitrust laws please? Last thing we need so more concocted narratives.
Super interesting video. Thanks! I have so many questions. Getting through content like this can take me hours!😂
I had no idea that Burt Reynolds died in 2018 before watching this video. 😢
Copyright is just make believe.
Check out Scite for research
If I can read it and learn from it, for free so can an AI. It's the same thing.
Hey Matt. Claude 3.5 system prompt got leaked. Maybe you should make a video about it
7.532 x 7 = 72.724 and not 52.203
Is Claude greatest for editing or helping with writing books?