There Was A TON of AI News This Week!
Вставка
- Опубліковано 10 чер 2024
- Here's all the AI news you probably missed (there was a lot). Check out Magical for free at getmagical.com/matt
Subscribe to this channel for the new podcast (officially launching next week): / @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
🐺 My personal blog: mattwolfe.com/ (Currently Being Overhauled)
Resources From Today's Video:
AI Landscape Overview: t.co/wttZwZoZZW
Microsoft-OpenAI Project: www.reuters.com/technology/mi...
Synthetic Voices Insight: openai.com/blog/navigating-th...
ChatGPT For Everyone: openai.com/blog/start-using-c...
DALL·E Editing Tutorial: • Editing DALL·E Images ...
UA-cam on AI Training: news.futuretools.io/youtube-c...
Anthropic Jailbreak: techcrunch.com/2024/04/02/ant...
Claude 3 Tool Use: news.futuretools.io/anthropic...
Siri's AI Upgrade: www.tomsguide.com/ai/apple-re...
Stable Audio Update: stability.ai/news/stable-audi...
Musicians Against AI: techcrunch.com/2024/04/02/nic...
Artists' AI Appeal: / 200-artists-urge-tech-...
KREA AI Announcement: / 1775955552067309675
AI Critique by Jon Stewart: • Jon Stewart On The Fal...
AI Job Loss Response: www.techtimes.com/articles/30...
AI Evidence Ban: / 1775823791467212813
Carlin AI Settlement: www.theguardian.com/technolog...
Amazon's Workforce Reveal: www.businessinsider.com/amazo...
Apple Vision Pro Personas: www.theverge.com/2024/4/2/241...
SharePlay Experience: / 1776040753879470533
Ola Solo Scooter: auto.economictimes.indiatimes...
Waymo Uber Eats: www.telematicswire.net/waymo-...
The Circle AI: ew.com/the-circle-season-6-ca...
NBA AI Prank: / 1774161341634232328
The Next Wave Podcast: / @thenextwavepod
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 Crazy AI Landscape
1:16 Microsoft and OpenAI Data Center
2:45 OpenAI Voice Model
4:56 ChatGPT For All
5:21 Edit Dall-e Images
6:40 UA-cam Addresses Sora
7:12 Magical AI
9:34 Anthropic Jailbreak
10:40 Claude Can Use Tools
11:17 Apple Introduces ReaLM
12:48 Stable Audio 2.0
13:54 Musicians Fighting Against AI
15:58 Krea's Cool Update
16:41 AI on The Daily Show
17:54 Big Companies Trying To Mitigate Job Losses
18:11 AI Image Evidence Not Admissible in Court
19:05 George Carlin AI Update
19:46 Apple Vision Pro Update
20:28 Autonomous Electric Scooter
20:56 Driverless Uber Eats
21:36 AI Stars in Reality Show
22:33 Hilarious NBA AI Filter
23:00 Huge Personal Announcement! - Наука та технологія
Dear Matt and team, just a hint to try out for area changes in some image generators like DallE or within Photoshop out of my experience: A prompt like 'NO wheels' to change areas does not work in many AI image generators yet, only when they are trained on the parameter, like in Midjourney for example with the --no command. AI prefers to know what shall be seen in that area. In this case a prompt for a region change to no wheels, could be just "asphalt" for example, so the AI forgets about any wheels. It is just like thinking of the pink elephant - better don't mention it, or it will surely appear.
I'd like that too. Have such an issue removing objects in Dalle
Well I think ChatGPT extrapolates from user input when prompting the painter so I guess that's something they can adjust.
If you are using a stable diffusion you have Negative Prompt for that purpose. I don't tend to use it much these days.
Regarding the AI on the Netflix show The Circle, the premise isn't to find love. It's to outsmart other people in the game in order to win money. There have been the odd love connection made but usually it's nothing serious and it's not the goal of the game.
Thanks for the shout out matt - a lovely surprise this Sunday morning.
For the image editing, you can use "wheelless" instead of "no wheels." That usually prevents it from getting confused about whether it's a positive or negative prompt word.
I appreciate all of The work You do to bring this all home for us.
As someone that has used Gemini, Chatgpt, and Claude extensively, 14:24- sounds like it was completely written by an AI. In fact, I used Gemini to write something almost exactly like that this week, it was eerie AF; the irony of that signed letter is wild, an AI written letter protesting AI.
That single MAD image represents 99% of all the money in the world economy. AI is revolutionizing everything in business.
OMG, your mental gymnastics on the google layoffs thing is nuts too.
Looking forward to the podcast! Will it have transcripts available?
Congrats on the new show/ podcast. Looks pretty cool, Ill check it out
UA-cam CEO says "OpenAI can't train on our users' data; only WE are allowed to train on our users' data!"
Regarding Stable Audio 2.0: in the announcement thread on Reddit, Emad explained that they're only training it on instrumental tracks from one source and filtered down only to the artists who opted in for them to use their tracks; he tried to throw shade on Suno for not being so careful when it comes to using copyrighted music as training data. But I mean... limited training data leads to limited results, and we can clearly hear that. And trying to restrict a song from being used as AI training data is like telling an aspiring singer "you can't listen to the radio when my song is playing!" So... I think Suno is in the lead both in model quality *and* sensibility here.
The workforce consortium is focusing on the wrong goal, in my opinion. We shouldn't be trying so hard to figure out how to stop people from losing their jobs -- that is inevitable under a market economy. Instead, we should be focusing on how to make sure jobs are no longer required for survival and health. It's not a tech problem, it's a socioeconomic one: no one would care that they lost their job if it didn't also mean they'd starve to death on the side of the road. And that's something that *shouldn't* be considered an acceptable consequence of job loss in the first place.
Congrats on the new pod! Will definitely be subscribing and looking forward to watching! Live your format and how much you cover in the AI world each week.
So happy for the pod, looking forward to listening to it.
Google canceling other projects to concentrate on AI, does count as an example of AI causing job loss.
good
This is an example of manipulating verbage to conceal reality.
It doesn't count as them replacing those people in those exact jobs, no.
@@user-cv1pj2vv1u We wrote two different things that are both true.
@@user-cv1pj2vv1u we said two different things that are both true
Will definitely check the podcast out. Awesome. Thank you.
Great to hear about your new Next Wave podcast plans!
I really love watching this weekly!
Congratulations on your new youtube channel. I'll sure subscribe and be on the know of what's new with AI. Thx Matt!
One EMP blast in the atmosphere and were all going back to farming. Lol. Still we'll have good memories. 😅
Thanks for covering the latest Ai news.
On your Podcast trailer, you mentioned your site was a com but it's an io.
Thank you so much for the feature! We got a place some Chess ASAP! Also congratulations on the launch of your podcast. I feel like I get the most informed listening to your UA-cam channel so I can only imagine how much more informed I can be by tuning into your podcast. Super exciting times. Thank you for doing all the work that you do.!!!!
Fidelity uses voice authentication at the moment. I wonder how quickly they'll reconsider.
Already Waiting for the first episode of your podcast. All the best matty and team, you’re the best on AI news and things.
Hey Matt. Image generator's don't know how to "Remove" objects. Introducing any noun makes dalle generate that noun. I have a GTA Loading screen style image generator in the custom gpt marketplace...run into this issue, but this edit feature is huge for me. Try prompting it to "make the highlighted area a concrete background." Replace the selected area with a "new" background...don't try to "remove a noun".
What if you say "no wheels"? Or select it all and ask it to generate a flat board?
I usually type something to add a little context like "magically the skateboard had no wheels and hovered mysteriously" that usually does it.
Just use any of the open source tools, there’s negative weights and even negative LoRA. Removing things from AI images was solved long ago.
@@thomasmitchell2514 it was but there's still a skill in HOW it removes a given feature... and thats a prompting thing cos you still have to describe the content to the model, even if its just telling it what is behind the wheels... It should I agree just get its context from the surrounds... but thats a little flaky.
Very cool to learn about the podcast! Congrats! 👏👏🎉🎉
Congrats on the podcast launch! What an endeavor!
Congrats on the podcast! I'll be sure to join in 😊🎉👏
I normally fast forward every ad in every video--except YOUR ads Matt -- you actually utilize the things you are advertising, show how to use the tools and make it an engaging watch. GG to the companies that pair up with you
Same
I love your videos and am so happy to hear about the new podcast. I'm sure it will be great, and I'll wait patiently for the first episode.
Congrats on the Next Wave! Super excited for you!
Great show Matt!! Any info on Verses AI. I like there approach
Looking forward to the new podcast. Good moves, Matt and Nathan!
Congratulations on your podcast!!!🎉
Can't wait for the new pod Matt. Thank you for collabing with Nathan to put out more Ai goodness into the world. Rock on gents.
Great video and breakdown. Thank you!
The AI that makes a riff off your noises means you can sample ideas then use a sound mixer to convert them to music. Taking the idea from your head and putting it into produced music.
Which tool was that?
You have been able to do that without AI since Ableton 8. C.2014.
Excited about the podcast! Subscribed!
I'm in the Phx area and see the waymo cars everywhere.
Excited about the podcast!
Grewat show MAtt. Love your boundless enthusiasm.
This podcast is going to be Amazing, congratulations Matt! Can't wait to tune in!
Waiting forward to the podcast!
That new feature in Krea looks great, but I can't figure out how to add images to the prompt and weight them. Is this feature in beta?
Hi Matt. Great channel I have been watching you pretty much since the start, gotten a lot of good sites and use cases . Couple of things Id really like to see you do a video, or actually a series on: Agents; robots. I realize these are big topics, but I really would like a review of the best agents as I dont know where to start, despite trying last year when hugging face etc had a couple of less than perfect models. Robots are obviously going to be huge, so would be good to have a review or 2 on them as well. Thanks again for all your work
when I hear some say "save you a whole bunch of time" as an employer what I hear is you can do more with your time, more productivity
Yep. I am not even an employer but I still think of it this way when I watch these news updates and catch tools I could use and my thought is "How much more can I do in X period of time using this" and will usually test it. Some work, some don't.
"Would you mind sweeping the floors while you're doing A,B,C,D,..."
I love your intro, copy and video editing was on point 👊🏻
Wonder if he does his own editing or has help.
Here's me thinking the Indian autonomous Scooters was an April fools joke XD. Looking forward to the podcast!
I just watched the video on Ola's channel and there is no doubt that it is indeed an April Fool's joke. I'm surprised and worried that Matt fell for it. I rely on Matt for a lot of my AI news.
Can’t wait for the podcast 🎉
Oh I am SO EXCITED for your podcast!!! 😱😱😱🙌🙌🙌
What article is it that talks about the GPU compute power makes a didderens?
Cool video, thank you!
Please can you remind us in another video that the podcast is up and available x
@Matt Wolfe HEY !
The TechCrunch article is *WRONG* about the Anthropic paper, and *YOU* *ARE* *IN* *TURN* .
They don't ask it « a bunch of harmless questions », instead they give it a bunch of harmful question *and* *their* *HARMFUL* *answers* and then it stops refusing to answer (because its context contains a bunch of times when it « thinks » it has answered harmful questions)
Look at the screenshot *in* *your* *very* *own* *video* , the repeated questions (on the right) are *in* *fact* *harmful* .
And the mechanism isn't a mystery as the article suggests, the mechanism is very clear:
The context contains previous questions/answers, so if you "fill" the context with what looks like the model answering bad questions, it will think it is currently answering bad questions, and will "keep going" doing that.
He doesn't read arXiv before talking.. gpt4 omitted it in the script 😂
@@TheTruthOfAI Yeah, considering you just need to read the diagram/illustration, which he shows in the video, to realize the article is wrong, I really wouldn't be surprised if the issue here was Matt relies a bit too much on LLMs to generate his content...
Either that or a honest human mistake ... maybe ...
like the beats of the new show
Is it kind of weird that I was actually expecting you to start a channel like the next wave and you did.
I was so convinced this was around the corner and I’m really excited about this new development.
Few are better equipped for this.
Looking forward to it
I don't know what I would do without this channel, it literally gives me everything I need to know about AI.
"Google laid off 12,000 workers. It wasn't due to AI, it's just that, with the AI development, thru weren't needed any more..."
My thoughts exactly 😂
I had a stroke trying to read that
Boss: "Oh! Nice! I really like this AI thing you just did."
Also Boss: "And, by the way, you are fired."
Maybe an could help you out your sentence together. I sure want to understand what you are saying.
Edit: while trying to criticize your comment regarding poor grammar, I failed to even proofread my own comment and so my reply was equally incoherent. Therefore, I am banishing myself for this.
I hope AI didn’t write this comment.
The Next Wave podcast visuals are dope
Thanks Matt. You are definately functioning in your Grace Gift. You are doing great work.
🎯 Key Takeaways for quick navigation:
20:43 *🚗 Autonomous Delivery and AI Catfish on Netflix*
- Introduction of autonomous delivery by Whmo in Phoenix for Uber Eats.
- Netflix's "The Circle" introduces AI catfish characters.
- Use of Snapchat AI filters at NBA Pacers game.
23:13 *🎙️ Announcement of "The Next Wave" Podcast Launch*
- Announcement of "The Next Wave" podcast with Matt Wolf and Nathan Lans.
- Aim to discuss AI trends, news, and insights with industry leaders.
- Invitation to tune in and subscribe to the podcast on various platforms.
Made with HARPA AI
1:46 stargate for q-star
Matt, the Circle is so cool, you missed something! 😂😂😂 Can‘t wait for season 6. 🤘
IS there a way to make consistent characters free and by using a pre existing photo to add animations like kicking punching and walking so i can make a sprite sheet of my character i drew up but not good with different poses
Hopefully some one answers cause I want to know too
It's time to take this up a notch. I congratulate you on your announcement including outro. You and your amazing team, with your leadership will prevail at leading the world in AI education.
Nice one bossman🙌
I love the color scheme of the podcast
By Christmas we are going to have a treat!
thanks matt!
How is Apple doing with the Pro?
@18:14 staying viable in AI has been brutal with how much I have to learn and do daily- from high quality image and video to software dev to make machine learning projects full stack to applying results using APIs with multiple LLM API’s, im talking DAILY. The company is HUGE, its not even a startup. Part of my present job is training those who want to stay with the company I’m contracting at so they can keep their creative jobs and so far only folks in other countries show real interest, literally zero of the folks in this country seem to be able to read the writing on the wall.
I don’t feel sorry for people who think they will be replaced by AI because no they won’t, they will be replaced by people willing to work alongside AI
I hate to read books but I enjoy listening to audiobooks. I wish we had a better text to speech available for our docs and files other than Apple’s screen reader.
My fave thumbnail of yours yet
Trying to download Devin to make my app for a school project; Docker program said it is unable to detect a Hypervisor and there is no Hyper-V setting in my windows features and I tried incommands but it says elevated permissions are required to run DISM. Any help? Can i ignore the message and continue with the rest of the video trying to download or it wont work?
thanks love again for such informative and enthus! catchup with my fav ..AI Anything ! :))))
Chat GPT actually will save your chats (I'm pro, maybe bit for non pro users thought I assume you are). I realized I wasn't logged in and was upset I'd lose my chat but it was there. So maybe if you login before navigating away it saves your chats.
22:20 vsauce already done that
Am an avid follower, thanks for all your hard work and being at the tip of the spear with information. Looking forward to 'The next wave'.
You should look at the breaking points format with the segments all named at the bottom of the screen. And you should also release every segment as a shorter 5 minute video as they would get shared more.
I believe the magic number for algorithm is minimum 10m for a video. Under 5 is like a short.
I have no info about over 20m. I like the longer ones so i can settle in with a drink and a notepad.
There's no way for a company to compete with the giant companies and their monolithic foundation models, but there may be a chance for a distributed peer to peer run model where the global populace contribute CPU/GPU to the processing layer to beat them, at least in some use cases. As you said, more compute equal better results. And I don't think you can build a data center that will have more power than the global computing population.
It's fascinating to see the rapid advancements in AI landscape, especially with projects like the Microsoft and OpenAI data center collaboration. The potential for AI-generated voices is incredible, but the ethical considerations are crucial. Looking forward to seeing how these developments unfold.
Yeah, now all the facial recognition and voice recognition we've relied upon for security is useless, when people can use AI-generated voices and pictures to steal your identity.
Very excited for your podcast! looks awesome! I hope the tech stuff is watered down enough for me to share it with my normie friends who love podcasts, or is it presented specifically for an already ai obsessed listener?
ive red that India is write an presenting in the Mid-Year a draft regulatory framework about AI.also they are writing a law about AI as they are setting up Regulator : Artificial Intelliigence and Data Authority of India(AIDAI) .(i am on 18 min. maybe you are talking about it)
in the 1st 500 subs. lets see how quick it grows
Will they use the data compiled by the NSA? Don’t they have huge server farms or something?
Ola Solo was an April Fool's Day prank! 😅
Thanks!
No problem bruh
If AI video upscaling worked the way people seem to think it does, someone would've already used it on the Zapruder film and we'd finally know once and for all which direction the bullet came from.
...and 100 million fuzzy UFO videos as well.
Freepiks Pikaso is underrated!
If that's what AI upscaling does, then it shouldn't be a thing because that's not how upscaling works.
I appreciate you
Thanks for the giant graph of AI companies gonna check out all the open source local AI's. I see one called GPT4All. That will be great.
Ola Solo is an April Fools prank 😂
So many companies are delving into AI, it's ironical the acronym was "MAD" because it's like everyone is mad with AI right now! (In a good but crazy kind of way!) It's super exciting to think what could be happening with that many named companies on the AI bandwagon, and potentially countless more we haven't heard of yet. Who knows, maybe the next big company might be one of these small AI ones that take the worldt by storm? It's exciting to think about. Otherwise, at the current point in time, OpenAI and ChatGPT seem to appear to be the "one-stop-shop" for every general AI purpose, and with their potential big project with Microsoft... 😳 It's almost scary how big that could be, but definitely exciting! Goodness, I'm hyped!!
See the way copilot can control windows, if Apple creates Siri in a way that can control iOS and macOS better, it would be insane
👎 I don't get why people like apple products. I've used my moms IPhone to teach her how to use it and it was a very frustrating learning curve. ⁉
Copilot has been a gamechanger for me. I wish they'd stop rebranding it though, I hope "Copilot" is the final name.
I still don't see inpainting in Bing chat, But I don't have the pro version either
Congratulations on your new podcast. I am not sure If I am wrong but the sound of your voice sounded a bit of AI like.😂
Think people are underestimating stable audio here. Yes there are no vocals like we have with suno but the quality of the instruments and general audio crispness for me sounds better.
Matt picked out the worst sounding demo on the page, the others are much better. People should give it a try.
Also the audio to audio is amazing. As someone who often has little song ideas in my but never any time to sit down at a keyboard/computer to get them down to anything that sounds remotely good, I'm so excited to now have a way i can just sing into my phone and get something approaching what i have in my head recorded down for possible future real song ideas
The problem is that once it requires no real talent to make music, your music is worthless because everyone else can do the same.
Another AI installment from Professor Wolfe. I learn so much from this man.
Fuckin right, you name that supercomputer stargate! loved that series.
❤ I wished many times the "Time Stamps" List would be merged with "Resources From Today's Video:"
👍 Keep up the good work though.