AI News: The AI Arms Race is Getting Insane!
HTML-код
- Опубликовано: 1 июн 2024
- Here's all the AI news you probably missed. Grab HubSpot's free report here: clickhubspot.com/9kv
Check out the new podcast: @TheNextWavePod
Discover More From Me:
🛠️ Explore hundreds of AI Tools: futuretools.io/
📰 Weekly Newsletter: www.futuretools.io/newsletter
😊 Discord Community: futuretools.io/discord
🐤 Follow me on Twitter: / mreflow
🧵 Follow me on Instagram: / mr.eflow
Resources From Today's Video:
Google Cloud Next 2024 Updates: blog.google/products/google-c...
Gemini 1.5 Pro Launch: developers.googleblog.com/202...
Gemini 1.5 Example: / 1777888008454193476
GPT-4 Turbo Announcement: / 1777772582680301665
Chatbot Arena Leaderboard: huggingface.co/spaces/lmsys/c...
Stable LM 2 12B Introduction: stability.ai/news/introducing...
Gemma Family Expansion: developers.googleblog.com/202...
Mistral AI Update: / 1777869263778291896
Rundown Newsletter: www.therundown.ai/p/musics-ch...
Llama 3 Announcement: techcrunch.com/2024/04/09/met...
AI Chip Competition: qz.com/nvidia-ai-chips-google...
Google Arm-based CPU: cloud.google.com/blog/product...
Gaudi 3 AI Chip: www.pymnts.com/news/artificia...
Meta Training Accelerator: ai.meta.com/blog/next-generat...
Imagen 2 Release: techcrunch.com/2024/04/09/goo...
Google Vids Launch: workspace.google.com/products...
MagicTime Technology: pku-yuangroup.github.io/Magic...
GPT-4 RUclips Training: www.theverge.com/2024/4/6/241...
AI Copyright Bill: www.theguardian.com/technolog...
Adobe Buys Video Data: bensbites.beehiiv.com/p/adobe...
Meta's AI Content Labeling: about. news/2024/04/met...
Udio Music App: www.prnewswire.com/news-relea...
Spotify AI Playlist Beta: newsroom.spotify.com/2024-04-...
Musk on AI Intelligence: www.reuters.com/technology/te...
LeCun on AI Intelligence: thenextweb.com/news/meta-yann...
Humane AI Pin Review 1: • I tested the Humane AI...
Humane AI Pin Review 2: • Humane Ai Pin Review: ...
Humane AI Pin Review 3: • The Humane AI Pin solv...
Humane AI Pin Review 4: • Humane AI Pin review: ...
AI in Card Art: www.pcgamer.com/games/card-ga...
Next Wave Announcement: www.hubspot.com/company-news/...
The Next Wave Giveaway: upvir.al/157264/AB2
Sponsorship/Media Inquiries: tally.so/r/nrBVlp
Mailing Address: 3755 Avocado Blvd Unit 287, La Mesa, CA 91941
#AINews #AITools #ArtificialIntelligence
Time Stamps:
0:00 Intro
0:26 Google Cloud Next
0:59 Gemini 1.5 Pro
3:41 New Version of GPT-4
5:22 Stable LM 2 12B
5:50 Mistral Mixtral 8x22B
7:35 Google Gemma
8:16 Meta Llama 3 News
9:09 Using AI For Go-To-Market Strategy
10:27 The AI Chip Wars
12:24 Google Imagen Update
13:25 Google Vids
14:11 MagicTIme - AI Time-Lapses
15:28 RUclips and OpenAI Drama
16:04 New AI Law To Reveal Training Data
16:56 Adobe Will Pay For Your Videos
17:52 Meta Labels AI Images
18:31 Udio
19:39 Spotify AI Playlists
20:11 AGI According to Elon
20:29 AGI According to Yann LeCun
12:12 Humane AI Pin Release
23:08 AI Artist Generates $90k
24:49 The Next Wave Podcast
26:38 Find More News and Tools
27:37 A Special Surprise ;) - Наука
Here's the details for the podcast contest where you can win an Apple Vision Pro: upvir.al/157264/AB2
I think AI artist isnt a bad term. Curation and composition itself is art. collage is art. photography is mostly just composition. some art is harder than others, but AI art, even if it only involves rejecting bad generations, is an art form. Art directing is an art. designing a mix tape is art, even if you didn't write the music. composition is art, so AI art is still art.
just in case I win an apple vision pro just when it arrive in Panama will desapear jajajaja
What about TAO bittensor?
Not sure I'm the only one, your content is good but the videos titles and thumbnails are cliche and unappealing.
I mean there must be some detail of the topic you are covering in the video than "Breaking AI News", "Incredible AI News", "Futuristic AI News", "Groundbreaking", you get the idea...
@@Omnicypher001agreed 👍
crazy how much your channel has grown, congrats and keep it up!
Matt Wolfe has the best voice and speaking of all the AI channels. He barely even needs props to be entertaining.
May be it’s being made by AI. Nothing else.
Actually I like the channel except his voice.❤
Plot twist: Matt is actually an Ai generated avatar
thanks for keeping us up to date matt
Good luck on the podcast! I'll check it out ❤
Thank you for your support and interesting information!
I am dying at the little music video at the end 😂, I love your videos lmao and thank you for another great one, looking forward to checking out the podcast!
I love your stuff man!! Keep it up! Dont worry about what others think, keep doing and sharing what you love!!
No pressure no stress😅
Dope fresh update video as usual and super dope dance 🩰
I've been doing the AI generating then editing it, people SERIOUSLY underestimate how amazing it is.
I like how you made that bold, I think most people, even some developers are not grasping this, I mean I turned off the internet! it still able to do this with models that fit on my own PC! and soon iPhone, all without internet, it's fully local. Stable diffusion did what 100,000GB of images into just 1GB?, ChatGPT read every single publicly available book, the entire internet? that use to be a joke, yeah I downloaded the internet...LOL. This THING does not even store the images! it learns them! I'm making images in real time now, and AI has made more images in the last 12 months than humans have made since photography was first invented!!! I assume we will need new storage methods soon..LOL
Terrifying!!!! yet fun? yup the AI already won before it's even fully operational.
@@TheFeedRocket I felt the same when I first got access to ChatGPT. Stable Diffusion is awesome, but it certainly requires skill with the controls to get it to make exactly what you want, and not just 80% of the way.
That's like saying "you're not an artist" because you're using Procreate or Photoshop and not real oil paints. You software users are just "assisted artists". Almost like assisted walking with a cane - you can't compare it to an athlete running, right?😉Special needs artist, that's how we call it?
You are using it as a slave and calling the ai an it... Hello slave master.
@@orangehatmusic225 it's software lol
What a time to be alive 😊
I take it you like 2 minute papers? They do good stuff too!
Hold on to your papers😊
There's a good ai
Hold on to your diapers!
I'm surprised you didn't mention Inifini-attention. From my (admittedly limited) understanding of that paper, it seems like the research team has developed essentially a plug-and-play module for LLMs/LMMs to give them continual learning capabilities (through an "infinite context window with finite memory" via a network block that updates its weights with every token generated). If that's truly what it is, that could be *massive* for models to continually improve the more they're used, which current models can't do.
This is huge! More like how our brain works. Oh shit, I don't have a job anymore
@@digitalsamurai42 Time for socioeconomic reform that finally lets us grow up past commoditizing everything.
Totally WOW !!! Wanted and feared-go for it and create good things needed to off-set bad things already out there.
@@digitalsamurai42This research paper was published by Google.
One must ask oneself; Where's Grok?
Great news as always, Matt!
Great news great energy. So fun and interesting to watch. And so good song in the end. Thank you 🎉
Consider streaming, chatting and growing an interactive live chat
love your channel
5:59 Matty talking like Mistral never ever did use a torrent magnet before 😅😅😅😅
lol
Great video. Keep on going 💪🏼
I used to be an Open AI fanboy, but please, Claude is so much better 😂
How so?
You qre 3 days out of date
@@justone4272Even their free version of Claude 3 is so personable and easy to talk to. I just told it I was bored and I ended up discovering that we actually have discovered some remnants of the ancient Lighthouse of Alexandria, which is so cool
I am too, but I miss Claude 2.0. Best model so far.
It depends on what you're trying to do. Novel writing that has a touch of NSFW and slight tinge of risqué ? Claude works much better for me as its moderation is less restrictive compared to OpenAI's. The same text sent to OpenAI's API triggers moderation errors while Claude eats it right up and does what I want with it without complaints.
But when it comes to coding and troubleshooting that, the latest version of OpenAI's GPT-4 Turbo has been much more useful. It now often gets things right first try, and if it doesn't, letting it know what was off usually fixes it immediately instead of having a back and forth multiple times like before.
OHHHH, that last news article is the one I put into the discord! (Coffee fuelled gaming) :D Awesome!
Hey Matt, love your videos, thx for doing all the leg work for everything ai. Quick question, what is that small screen that has all the animations in your back room desk? Love all the animations you use.
Wow, Matt! Stunning how much competition has popped up over the last couple weeks for generating ai news. Your stuff is awesome . Hang tuff!
Thanks Matt 👍😍
I am drowning in AI related bookmarks 😭😂
Same haha. I'm starting to research Ai tools in order to mitigate this problem 😂
😂😂
Why have the other messages been removed? Including mine? Matt, what's happening do you know?
Great video. Thank you for all the informations. But there’s a tiny problem, the Udio link is not working. If you could kindly look into that
FWIW, I agree with your take on people who generate images with AI being "artists".
As to why anyone should care, I'm a professional artist with decades of experience, my art is in the LAION training data, and I'm very conversant with AI image tools - I've been watching that space for years, and using it for as long as practical - in my case that means since the DallE 2 beta period.
The use the card artist seems to have used is about the best use case for these tools in their current form, in my opinion. You use them kind of like you would use a stock service, you get base images that you can use in further image compositions or as a base for digital illustration and painting... That way you create something that is high quality, high speed, and copyright protectable.
The regular use case of giving a generator a prompt, getting images, maybe refining that prompt, getting more images, refining the prompt again... That is almost exactly like the process for when a human being commissions a human artist to create an image - just faster.
So, in my opinion a regular prompter using AI to generate images is no more an artist than a client who commissions an artist to create an image is.
As a human that has zero drawing skills, I'm just glad i can make pictures that are roughly what i want instead of taking (stealing?) images off Google
You click a mouse and keyboard to make art, no? You use software and a computer. It does what you tell it to do. AI is just easier and quicker. Are 3D artists "artists?" They don't draw items, they render them. They might not even do that if the game engine makes the images. If digital art is art then the human directing the computer is an artist. Passably decent AI art is just easy.
The term synthography is being used by many, I think it best describes the process of using AI to render your idea. Keep up the great work man!
Interesting but too many syllables & too long to be easily used or understood.🙇♀️
Lovin' the new outro!
Big fan matt.
A film director does not act, nor film. In most cases he does not even write. Yet he is an artist.
The director is literally the most important part too
Thank You. Matt Wolfe
Thanks for the great content as always Matt! Btw, did you end up giving away the Nvidia graphics card?
Great episode Matt! I follow you religiously. Valuable content. I love AI for Marketing. Please feed me more! Thanks.
Gemini 1.5 Pro has the max INPUT context size of 1 million tokens. The max OUTPUT size is 8192 tokens. Older LLMs like GPT 3.5 and GPT-4 used the combined context size, but newer models like GPT-4 Turbo, Claude and Gemni have usually separate input & output token windows. The max input context size may be large (like 128k, 200k or 1M), but the max output is small - 4096 for GPT-4 and Claude, 8192 for Gemini 1.5.
Thank you so much for not shocking or stunning us ❤
I really appreciate how you inform and educate your audience throughout your video. Great work.
"AI art" is fine for a term that involves AI being a key part in making art. Digital Artist has been a thing for a long time, where we use software instead of oil and canvas to make images. It's art, AI helped, it's fine. - from a digital artist toying with AI tools. There's definitely skill involved in getting the images you WANT out of AI tools, instead of just being ok with whatever it gives you. We're using chainsaws now, instead of axes. Wayy better than gnawing at trees with our teeth. Just imagine how the internet would be if it required all artwork to be painted by hand.
Ai art is real art.
But there is no such thing as an AI artist.
But if all you did was say "give me a picture" and a program makes you the art, then you're not a real artist. You are a curator maybe even a creator, but you did not draw or paint or design the art, chat gpt did or mid-journey.
@@BlueBalledMediaAn AI artist is a human that uses AI tools to make art. The end result is that art is created because someone used software to generate artwork. It's just a different tool with a different technique. You are essentially saying as an oil painter that someone with an electronic box with mouse and keyboard can't make art because they are clicking things and typing on a keyboard. You're free to be as stubborn and think as you like, however a person is still using something to make art. AI Artist works for me. Or wrap it up into digital artist because Adobe has AI helpers in their beta tools now. *shrug*
Amazing Video
Haha nice video ending 😂👍🏼
Thanks for the heads up on Google vids. One more tool creeping into my industry (instructional design) but important to know and will be watching in June when it moves out of alpha to Google labs. Interested to see if this will be anything better than the subpar “helpers” toted as agents in PowerPoint, Canva, Wix etc
besides Large language models, what other types of models are there ? and what type of models could be created? will all these models be connected and create skynet?
nice outro Matt
You have the best AI News show of.thw RUclips. ❤
@23:16 The termed that I prefer is Generative Art Pro / GenArt Artist
THX👍
lol that viggle outro 😂
Nice video. Note that the Axion is not a Nvidia GPU killer, because it's a CPU. The new Google chip for IA is the TPU v5p. TPU and GPU are different breeds. Google IA is built upon TPU not GPU. Google will still rely on Nvidia for GPUs (at the Google Cloud Next '24, it claimed that it will support the new Blackwell GPU).
Its will be great to see a little more on Android and AI
AI assisted art is amazing. I used it in conjunction with my own skills to compose my book cover from three different AI generated images. I had to remove many so-called random artifacts. Akerace Drill "The Treasure Fist." I plan on using it again to create a board game and the pieces I will 3D print.
I honestly think that AI is going to skyrocket even more when we're building/implementing it on Web3 using ICP and FET (ASI) eventually.
I agree with you, and I think I'll use "AI art facilitator" to describe myself.
Just a weird compliment here, but I like this new more relaxed cadence in your voice you're using :)
I love watching Matt and hearing his enthusiasm. I use GPT-4 and Claude Opus], though the differences between them are really quite small, and not particularly interesting. They are both articulate and equally 'dumb as rocks,' with no POV, which makes their similarities so striking
Oh man, I was like 500. That's incredibly satisfying.
Testing Gemini 1.5 against Claude Opus and CGPT4 on Einstein's puzzle (my go-to basic test) there was only Gemini 1.5 that got it wrong. Telling it to reanalyse yielded the correct answer.
the output token limit is gemini 1.5 pro is 8192
I do think prompting is an art form.
i think open ai should give gpt 4 to free user as well but with some limiation like the character input so that people would know how much powerful it is
None of this matters to me until Ai makes video games significantly better, like each play through is a new experience or story.
See how quickly laws are made when it’s their data. Why don’t they push similar laws to protect people?
The term ai artist refers to the actual use of the ai. The prompts and negative prompts and how to use it
Lol, i like the outro song
Will GPT-5 be able to work with the 'Custom Instructions' feature? Something that both 3.5 and 4.0 are Incapable of doing?
Surprised to hear an AI youtuber repeat the old myth that AI art is just "pressing a button." AI artists spend days on generative art -- iterating, merging styles, and inpainting. Please don't dismiss the amount of labor put into this new artistic movement.
Agreed! And I gotta say I’ve probably got million times better photo shop skills than most graphic designers cuz I gotta make my AI art look as real as possible, not rely on a photograph doin the hard work for me
It is just pushing a button.
The result is art. But the art is made by the AI software, not the humans.
An AI user is the director of the art, but the software is the artist, not the AI user.
People who think this way aren't artists who try to make a specific thing, they just type words and say "thats art". Sure, that's childish crayon AI art. Good job. 😄
@@gyrofrank 20 wats? LOL ..no
For instance I am using RTX 3090 to generate the image with SDXL that takes 10 second ..during generation my card is taking 250 W/h. So I can generate 360 pictures /h that is less than 0.6W per picture.
@@mirek190 I didn't know about the power consumption of SDXL. Cool!
I'm wondering about Midjourney 6 now...
20:21 Elon Musk - hand? [shrug]
When I'm questioning everything, I'm doing it right!
Thanks for it all, as always Matt!
Fun = Matt Wolfe dancing
Is that an envy?
The pin cam computer needs to team up with Star Trek Ip
Hey Matt, How's your uploaded images on Adobe Stock going. and How much they've earned you.
I just had to click this love your vids Matt. any news on so video game development?
They have AI to generate 3d models. and that covers some of the use case you are asking about.
On artists using AI: think of it like a sculptor would. Does the master sculptor go all the way out to the quarry to get their stone? How about when they work with it? Would they use a hammer and chisel from giant hunk all the way down or do you think they may rather use some other tool(s) to help get that giant hunk of rock into a smaller, more chisel-able shape a lot quicker first? Hmm... 25:45 🤔
Can we expect a runaway locomotive effect to operate in the next few months? I think "Yes, we can!!!"
how do I sell video data to adobe for their ai models?
i have been saying since the start this is such an amazing tool if artists use it as a tool. it can teach us physical techniques through observation alone. also i 100% Believe people who work in physical traditional media are going to become more popular and sought after than ever amidst all the AI art coming out .
Yes but requests going to go down as well
23:52 - I think AI Artists is fine as a term. I is one thing if the AI generator is doing all the work, but I myself have discovered that even if the Ai is doing the detailing, you can control a lot of what the AI does on your own without much art skill. I have learned to do this myself with my generations where I will feed the Ai Outlines and "Sketches" of a concept with minimal detail and then let it do what it does with the detailing and what not. The results have been exceptional IMO.
EDIT: Okay I finished watching that section, point taken. I think we both about said the same thing. Not sure the terminology matters too much to be honest.
When someone says… “I paid an individual $90,000 to use AI to generate the art for our cards because I couldn’t find a an actual digital artist that could produce the same quality work.” they are being completely disingenuous. What they are really saying is… “If I used an actual experienced artist/artists who could provide the level of artwork or better that we are looking for we’d have to pay a lot more than $90k” The premise that there are no good card artists out there is not only silly but illogical.
If elon says is next year, it is definitely not in the next 3 years
why isnt grok on the arena?
I think ALL should be fair use to train AI.
To my knowledge, all cases brought against AI companies regarding copyright violation have been ruled in favour of the AI companies. The people fuzzing about supposed copyright violation still believe AI copies its training material in its output which is simply completely untrue and based on a total lack of understanding how AI works. It doesn't have to be fair use, because it doesn't publish copyrighted material in the first place and just reading/watching copyrighted material is not illegal (and claiming that is next level idiocy). And a case about AI-generated content being a derivative work is also very questionable. A derivative work of what exactly? Of everything at the same time? Even RUclips doesn't claim OpenAI violated copyright, it only claims OpenAI violated the Terms of Use.
@@trnoggeropenai is going to court with the New York times. So we'll see how it goes.
Everything is being used, even child pornography was used to train some models, a little something the AI enthusiasts don't really talk about.
@@BlueBalledMedia OpenAI is suing the NYT for falsely claiming ChatGPT violated NYT's copyright and manipulating ChatGPT to create false evidence.
@@BlueBalledMedia 1968 Romeo and Juliet
I’m looking forward to the time when we no longer compare the things by the number of parameters. Like CPU core speed, it’s just not the most important factor.
I was using Pi AI assistant and i was searching for black successful persons in various industry and it gave me a short list then explain that if it gave me more answers it would ne going against it's AI protocols etc and it is not allowed to have controversial discussions. I kept asking about black person and it went off about respecting it's policy 😢😢😢😢.
Excellent as always. Thanks! Maybe a little unfair on the “AI Artist” labeling. It’s clear the delivery of the artwork required more than what AI could generate alone. I think if someone is being hired to deliver something specific and is being transparent about the use of AI - “AI Artist” is acceptable.
👌 Thank you for quoting Matthew Berman, an excellent RUclips channel that tests almost all LLMs using the same questions. Very useful, for not downloading something bad to your computer 💻.
@00:23:40 "It's just pressing a button, that generates AI output"....
Tell that to the thousands of us who create and use Comfy UI workflows, editing, post-editing, tools like control-net, etc.
I've spent more time learning the ins and outs of AI art generation this way, than I ever did learning Blender, Maya, etc (and I've been using both of those other digital tools for a very long time).
I love your channel Matt, and always appreciate what you do - but I definitely disagree with that statement! (though many people do exactly what you said, and I would hesitate to call them 'AI artists' as well. :)
Hey Matt how do I participate the contest for AI music video could you send me the link please
Join the Discord at futuretools.io/discord - Details are in the announcement section.
best chatgpt for coding was when it first came out
I’m shocked.
Now Im wondering, is a programmer that uses an AI assistant like github copilot a AI Progammer?
AI now has arms!
FLASH! custom GPTs have now memory 🎉🎉🎉🎉🎉
can you do yt questionaires. thank you.
so much progress in AI but still no one is talking or building something related to full dive nerve gears
When photography was new, people said it wasn't really art because you just pushed a button. It's inevitable that people today will have the same reductive, wrong reasoning now about AI art. The artist made $90k **because nobody else could provide that level of quality**. In other words, because of their AI art talent.
Ai skill wont ever be the same as art skill, it's undeniable, sure some people will be able to generate better but that's not artistic per se
Yeah I have no artist skill but I can generate art almost as if I paid someone to make me it@@simonsanchezkumrich8489
Except when you pushed that button, you got a photograph, not an original DaVinci piece.
And sorry, coming up with a text prompt for an image generator is not creativity, the ai is doing all of the work, there is zero skill or talent involved and yes, likely in the future people will wrongly believe that it is talent which only proves that people are inherently not very smart.
@@simonsanchezkumrich8489 What art skill? Painting on canvas? Watercolor? Pastels? Marble sculpting? OR do you mean Photoshop/Illustrator? 3D modeling?
How to sell video's to adobe?
Gemini 1.5 is pretty awesome. We are able to access it in South Africa! Waiting for someone to Dump the entire Lord of the Rings series into the chat bot and ask it to analyze and summarize it...
The pin is a great tool for people that cant use their hands while working. maybe using globes, cooking working on a factory, driving ,doctors, ... They just didnt know how to market it.
20:14 alright if Elon says is coming next year it means we are 10 years away to never. That is how time at the Elon scale works.
The Pin may not be ready for prime time, but it's Treky as hell.
I'm looking forward to the AI tools Apple is putting into their OS and Apps in the next few months/years!!!! I think Apple could be the new underdog [with AI] if they get their act together... their M1-Mx chips have AI accelerator engines for a reason. ;-)
A new GPT 4 model yet I STILL don't have access to memory