Like every week, there has been a ton of AI news that's come out since I published my last news video.

Most of it has been fairly marginal updates showing that we are making progress on all of this stuff.

Some of it has huge future implications, and a couple of things I'm going to show you at the end are just downright weird and hilarious.

I'm going to do my best to give you a rapid breakdown and quickly run through everything I think you should know from this past week.

Let's start with this infographic from FirstMark Capital.

This infographic shows the absolute craziness that is the AI landscape right now.

They call this the 2024 MAD landscape.

MAD stands for machine learning, artificial intelligence, and data.

You can see how overwhelming this looks.

This shows just how many companies are involved right now, and these are just the sort of bigger companies.

A lot of the smaller AI companies that are out there didn't even make this graphic.

But really, this image just represents how monstrous of a space this is and how many companies have jumped on board.

I will make sure it's linked up in the description if you want to sort of dissect it and zoom in and look at more of the tools.

But I'm really just looking at this at the 30,000-foot view of holy crap there's a lot going on.

At the end of last week, we got a report published on Reuters here claiming that Microsoft and OpenAI are planning a hundred billion dollar data center project.

This is still yet to be confirmed by Microsoft or OpenAI, but seeing that Sam Altman said that he needed seven trillion dollars to build a data center and the fact that OpenAI and Microsoft obviously work very closely together, there's probably some truth to the rumors.

With this data center, they're also planning on building an artificial intelligence supercomputer that they're calling Stargate.

If this data center does happen, it is expected to be 100 times more costly than some of the biggest existing data centers already.

Why is this important?

Well, if you remember the articles that came out around tools like Sora, one of the things that they showed off in the Sora research was that the video quality got better And better and better by just using more GPU, more compute power to generate the videos, and so far the same seems to apply with almost all AI models: the more compute we Throw at it, the more powerful, the more realistic, the more helpful they can become.

Again, this isn't 100% confirmed, but if this happens and Microsoft and OpenAI do team up to build this monstrous data center, I find it hard to believe that there's going To be really any companies or open-source models that will ever be able to catch up with what OpenAI and Microsoft are doing.

And while we're on the topic of OpenAI, OpenAI put out a blog post at the end of last week called navigating the challenges and opportunities of synthetic voices, and Basically OpenAI has created their own voice model similar to what we hear from tools like ElevenLabs that can generate realistic voices off of a single 15-second sample.

The audio that's generated from this is even a step above ElevenLabs in quality; it is really good.

Here we have a reference audio: the initial 15-second sample training force is a push or pull that can make an object move, stop, or change direction.

And then we have the generated audio, the brand new audio that was generated from that voice.

Some of the most amazing habitats on earth are found in the rainforest.

A rainforest is a place with a lot of precipitation, and it has many kinds of animals, trees, and other plants.

That was all AI generated.

It didn't sound much different than the original sample that was given.

Here's another reference audio: friendship is a universal treasure.

It brings joy, support, and laughter into our lives no matter where we are in the world.

And then down here, here's some generated audio in various languages using the exact same voice.

And there are tons of other examples if you want to listen to more.

Again, I'll make sure it's linked up in the description for you.

But here's the thing: OpenAI put out this blog post and said look how cool this is, but then essentially said we don't trust it in the wrong hands yet, so we're not making It publicly available and you can't use it.

And because technology like this exists, they are now encouraging steps like phasing out voice-based authentication as a security measure, exploring policies to protect the Use of individuals' voices, educating the public around the possibility of deceptive AI content, and accelerating the development and adoption of techniques for tracking the Origin of audio visual content.

Again, we already have tools that can do this kind of thing like ElevenLabs.

However, OpenAI's version just feels a little bit better right now.

Too bad we can't use it ourselves.

Also, in OpenAI news this week, you can now use ChatGPT without actually having to log in.

If you're logged out and you go to chatGPT.com, you can see that we get our familiar chat box and we can use it right away without having to log in.

It's not going to save any of your chats, you can't come back to them for later, and of course, it's missing a ton of other features.

But if you just want to ask a question real quick, you can jump over and use it without ever having to log in.

Staying on the topic of OpenAI, OpenAI just made it so we have essentially inpainting inside of DALL·E.

If we generate an image with DALL·E, we can now select a certain area of that image and prompt it to make changes to that image.

Like this dog here, they're highlighting areas of the dog.

Then, if I fast forward the video a little bit, they gave it the prompt add bows.

And then it regenerates the image here but now bows on the dog.

This feature seems to work directly inside of ChatGPT or if you're using Bing image creator.

If I generate an image like this inside of ChatGPT, I click on the image.

You can see I've got a new bar here that says select area of the image to edit, highlight the area and describe changes in chat.

My goal with this image was to make a guy on a hoverboard, but it kept on putting the wheels here.

Let's see if we can do it.

I'll select just where the wheels are down here and right here, maybe even on the shadows too.

See if it fixes that.

And then in the prompt, let me type remove the wheels.

We have a new image, but it kind of made the wheel squat here.

It removed the wheels on the shadow, but it just moved the wheels on the skateboard.

Anyway, it's a new feature in DALL·E 3.

You can play around with it.

It might take a few prompts and a few tries to get exactly what you're trying to do.

And in the final bit of OpenAI news for this week, the YouTube CEO Neil Mohan was asked about Sora being trained on YouTube videos.

He didn't know if Sora was trained on YouTube videos, but he said that if it was, it would be against YouTube's policies.

If YouTube ever found out that Sora was trained on its videos, OpenAI could be facing more issues.

But again, we don't know.

Mira was asked, she didn't know.

And when the creators of Sora themselves were asked on MKBHD's podcast, they basically said we're not willing to talk about that.

Today's video is brought to you by Magical, and this is a really cool time-saving Chrome extension.

If you head on over to getmagical.com/matt and then click on the button that says get started for free, we can create an account and then install the extension.

It'll take us over to the Chrome Web Store.

We just simply click add to Chrome, add extension.

Because my Chrome browser automatically hides the extensions, I'm going to click on this little puzzle icon here, scroll down to Magical, and then click on the little Thumbtack icon here to add the Magical button to my browser.

This tool has a handful of time-saving features.

My favorite feature is the ability to automatically reply to emails in my inbox with just a click of a button.

Here is an email where I'm being invited to an event.

If I feel I can go, I can simply click yes, and it automatically writes an email reply for me that says, Thanks for the invitation.

I would be interested in attending the AI-focused news briefing.

If I don't think I can go, I can click no, and it writes that email for me as well, using the context of the initial email.

I appreciate the offer, but I won't be able to attend the AI-focused news briefing.

And we also have a custom option.

I can click custom, and let's say I don't know if I can go yet or not.

I can simply type, I'm not sure if I'll be able to attend yet, and it generates an email that basically says, I will let you know if I'm able to attend closer to the date.

It saves you a ton of time if you just need to quickly go through a whole bunch of emails and give a yes, no, or maybe reply.

It's also got a really handy auto-complete feature where you can create a whole bunch of templates and easily use these templates in any message to somebody.

For example, I have a little address template here where I type -address, and it will automatically paste in my address.

Whoever I'm replying to, let's pretend for a second that email they were asking for my address.

I can click reply, type -address, and just like that, it automatically filled in my address into the email.

They also have a cool feature called transfers, which makes it really easy to sort of scrape content quickly from any page.

Magical will save you 10 hours per week, and it works on every single website.

And as of right now, you can use it completely for free by going to getmagical.com/matt.

I'll also put the link in the description, so definitely check it out.

It will save you so much time.

And thank you once again to Magical for sponsoring this video.

Moving on to some Anthropic news.

Anthropic researchers wear down AI ethics with repeated questions.

Basically what happened here was that Anthropic researchers figured out that if they ask the AI bot to do something harmful, it will tell them no on the first try.

However, if the researchers ask them a whole bunch of harmless questions over and over and over again and then eventually ask a harmful question.

It will actually answer the harmful question.

As of right now, nobody really knows why this is happening, but it seems to be happening on the Large Language Models that have the large context windows like Claude's 200,000 Token context window.

When they tried to mitigate this from happening, they found that although limiting the context window helps, it also has a negative effect on the model's performance.

Right now, AI researchers are trying to find that balance of how do we keep on raising the context windows without these Large Language Models being willing to provide Potentially harmful information at those larger context windows.

Also, this week Anthropic announced that it can now use tools inside of Claude3.

According to an email that went out from Anthropic, tool use includes retrieving documents, accessing public APIs, and orchestrating Claude subagents for tasks like scheduling.

Also, at the moment, it doesn't appear that this works inside of the Claude bot that you use by just going to Claude AI, but it's in the API.

Right now, I would imagine over time, if you're one of the Claude pro users, they'll probably roll out the ability to use tools and connect to other APIs in there at some Point.

But right now, this is more for developers who are developing with the Claude API.

Apple appears to be getting deeper into the AI game as well.

This week, they revealed ReALM, which stands for a reference resolution as language modeling.

This language model is designed to run on a phone and make voice assistants like Siri smarter by helping it understand context and ambiguous references.

A few weeks ago, there was speculation coming out that Apple was going to use gemini 1.5 and maybe that was going to be the language model behind Siri.

But now that we're seeing this ReALM model and the research behind it, it's probably more likely they're going to use this model, which was developed in-house.

This model is different because it's not as big as the other Large Language Models.

It's actually very tiny because it's designed to work on a phone and just help you with phone stuff.

It doesn't need to do everything, it just needs to know what you're trying to accomplish on your phone like scheduling things in your calendar, sending an email, finding a Photo from your photo albums, things like that that are unique to just your phone.

It's also very good at understanding what's going on on the phone, so it's got a visual model inside of it that can see what is happening on your phone and interpret it to Help make decisions within the language model.

We know Apple's WWDC, their Worldwide Developer Conference, is happening in June.

Last year, when they did WWDC, was when they announced the Apple Vision Pro.

The speculation is that this year we're going to hear all about the AI stuff Apple's been working on, including possibly a new version of Siri with much, much stronger AI Built into it.

Stability AI this week, who last week lost their CEO Amad Mastak, just announced stable audio 2.0.

If you remember, stable audio was a music generation platform similar to Google's music effects.

You give it a prompt, it generates a song in that style.

In the past, it was only generating roughly 30-second songs.

This new version can actually generate three-minute songs.

Here's an example song called lo-fi funk.

It's not horrible, but I don't feel like it's anywhere near what we're getting out of tools like Suno yet.

The other new feature is audio-to-audio generation where you can kind of hum or make the instrument sounds with your mouth.

And then it'll try to replicate that with ai.

It's pretty interesting, but in my opinion, it's not really generating any music that I get super excited by or would want to actually use in one of my videos yet.

And since we're talking about ai music this week, a whole bunch of musicians including Nicki Minaj, Billy Eilish, Katie Perry, and other musicians signed a letter against Irresponsible ai.

I'll link to the full letter below if you want to read what it actually says, but in my opinion, it's sort of unclear what they're looking for.

The very first sentence says, We the undersigned members of the artist and songwriting communities call on ai developers, technology companies, platforms, and digital music Services to cease the use of artificial intelligence to infringe upon and devalue the rights of human artists.

But then the signed letter goes on to say, Make no mistake, we believe that when used responsibly ai has enormous potential to advance human creativity and in a manner that Enables the development and growth of new and exciting experiences for music fans everywhere.

At the very end it says, The assault of human creativity must be stopped.

We must protect against the predatory use of ai to steal professional artists' voices and likenesses, violate creators' rights, and destroy the music ecosystem.

To me, it feels like a very vague letter of like ai is ruining music and we need to stop it, but at the same time, we also see that ai has the ability to do really Interesting creative things, but we should use it responsibly and not irresponsibly.

Like I want musicians to make a living and to get good at their craft and to put out music that everybody wants to hear, but I don't think it's always cool for people to go And just use other people's voices and music styles to create new songs in the style of that person.

I get that, I just honestly do not understand what this open letter is asking for.

Like it's just not super clear to me.

But this letter, I don't think it's gonna go anywhere.

It's not like because these 200 artists signed this all of the researchers working on this tech are gonna go, Okay, 200 musicians didn't like this so we're just gonna stop.

I don't see that happening, so I'm not quite sure what this is accomplishing other than people just saying, Hey, we don't like ai music, and I'm raising my hand saying I'm One of those people who doesn't like ai music.

Okay, cool, we now know.

What?

Let's move on to another form of ai art.

And this time let's talk about Crea AI.

If you don't remember Crea AI, it is a tool where you can type a prompt like a man on a towel at the beach and as you type it actually generates new variations of the image.

Well, they just rolled out this cool new image to image feature where you can upload multiple images, change the weight of the various images that you upload.

And then as you change the weights of the various images, you upload it sort of blends all those things into a single image.

This example, they uploaded three images with the prompt fish made out of porcelain, and as they change the weight on the various images, we can see the image adjust in real Time.

This is just crazy, like this is really cool.

I'm a huge fan of the daily show, and this week John Stewart actually did a whole piece on AI.

Basically, all of the big companies, the Googles, the Microsofts, the OpenAI's, all of these companies are basically saying that AI is going to create this utopia and we're All going to have our own AI assistants, and it's all going to be this amazing world and it's going to get better and better for everybody.

While at the same time also talking about how AI is taking away jobs from companies.

I'm not going to play this clip, but I highly recommend watching it because some good points are made here, and well, the daily show is just hilarious.

But one of the points that's being made here is that a lot of people are going to lose jobs.

In fact, there was a report that last year Google laid off 12,000 employees, and a lot of the reports that were going around about this was because of AI.

Google got rid of 12,000 employees, but that's not the case.

It wasn't these employees being replaced by AI, it was that Google wanted to move further into AI.

To build teams and sort of readjust the company around building for AI, they made old positions and old sort of departments in Google obsolete.

That's why the layoffs happened, not because AI replaced them.

But speaking of AI replacing jobs, a lot of the big tech companies are coming together to create the ICT Workforce Consortium.

Companies involved include Cisco, Google, Microsoft, and IBM.

And the whole point of this consortium is to Figure out how to keep people from losing their jobs as a result of AI.

Shout out to Let Me Be Frank over on X for pointing this one to my attention.

That a court in Washington banned the use of AI-enhanced video evidence.

What they mean by AI-enhanced evidence was there was video footage of an incident, and somebody took that video and ran it through Topaz Labs to upscale it, thinking that it Was going to improve the quality of the video so that you can see more details in the video.

That's not really how this upscaling works though.

Really, the AI on these upscaling tools is sort of trying to guess the pixels in between.

When you take the image and you double the size, it's inserting new pixels and trying to guess what the rest of the image is.

It's not making things that were there more visible.

In some cases, it's adding things that were never actually there in the image.

It totally makes sense that you shouldn't be able to use this in court.

A few weeks ago, I reported on the George Carlin stand-up comedy.

It was a completely AI-generated stand-up set, which I think later came out was not actually written by AI.

They just used George Carlin's voice to speak it out.

That case has come to a conclusion.

The company behind the stand-up comedy that was in the style of George Carlin agreed to take down all of the audio and video and anything that existed online of this Stand-up set.

There was some sort of settlement behind the scenes, and it wasn't disclosed what the actual settlement was.

We just know that George Carlin's estate and a podcast team that put out this stand-up comedy set all agreed to some sort of terms, and it's no longer online.

Not AI-related, but just kind of fun.

Apple Vision Pro rolled out a feature where you can actually interact with other people now.

One of the biggest downsides of the Apple Vision Pro is you strap it to your head and you're just kind of alone.

There was no real interaction other than like a FaceTime with somebody.

You can actually play games, you can watch movies with people, you can do virtual presentations, and you appear as sort of a floating disembodied head with hands while in The conversation.

Here's another example from my buddy Don Allen Stevenson here where it looks like they're playing like virtual board games.

They're playing virtual chess against each other.

I have an Apple Vision Pro.

I haven't actually played around with these features yet, but looking forward to doing that.

We're getting into the kind of weirder, more interesting stuff.

A company out in India created this little scooter called the Ola Solo, and this claims to be the first fully autonomous electric scooter.

You jump on this electric scooter and it, I guess, takes you where you need to go.

I don't know if I'd feel comfortable riding an autonomous scooter, at least not yet.

I don't know if I'd want to be testing out version one on the roads yet, but this will be something to follow along to and see if it ever comes to fruition because this, to Me, is fascinating.

Out in Phoenix, Waymo vehicles are now delivering Uber Eats.

If you're not familiar with Waymo, Waymo is a self-driving car company.

I reported in a video a couple months back that in Phoenix, you can now order an Uber Waymo, and a self-driving car will come pick you up and take you to your destination Without a driver even in the car.

Well, now they're applying it to Uber Eats, where you can place an order for food.

I guess the restaurant will go take it out to the car.

And then the car shows up at your house and you go and get the food from the car.

I'm guessing, I don't know, I'm not in Phoenix.

I haven't tried this yet, but you do have the option to opt out during checkout if you prefer a human to deliver it.

We're about to get AI as a cast member of a reality show on Netflix, The Circle of season six now includes an AI catfish.

Basically, this is like a dating show where there's a bunch of people that are in apartments and then they message with each other and have conversations with each other Without ever actually seeing each other.

And I think the goal is to eventually get them to hook up and date.

I don't know, I've never seen the show, but I think that's the goal.

There's some catfish in there as well, like this branded dude pretending to be someone named Olivia and Caress here pretending to be Paul.

Let's scroll all the way down here and we have AI bot pretending to be Max on this season.

It would be really, really interesting if a bunch of the cast members end up falling in love with the AI and not any of the real humans.

I might actually watch this just to see how well the AI performs and if the AI gets dates or, I don't know, if the end result is, we'll see.

Then finally, this is hilarious.

At an NBA Pacers game, they used a Snapchat filter to make it look like Lakers fans were crying.

You can see in this video, the camera is showing these people, they all look like they're crying, but that is an AI face replacing their mouths with a crying face instead of A happy face.

I just found this to be like just super, super funny and I want to see more of this at sporting events.

Before I wrap up, I have one last super exciting announcement.

We've been working on this for months and months and months now.

I think this is going on almost nine months from the idea to finally launching this and I couldn't be more excited.

Me and Nathan Lands of lore.com are launching a podcast called The Next Wave where we're going to interview CEOs and founders and engineers and people that are building the AI space along with conversations just between me and Nathan and no guest about the current landscape of AI.

And we've had a blast.

We've already recorded a bunch of episodes and it is launching next week, the week after this episode goes live.

And I've got the first sneak peek little trailer for you that I'm going to show you right now.

Hey, welcome to The Next Wave podcast.

Consider us your chief AI officer in your business.

My name is Matt Wolf.

I have the number one YouTube channel in the AI space.

I also run futuretools.com and I'm joined by my co-host Nathan Lands, founder of lore.com and one of the most connected people in Silicon Valley.

We want to bring you the latest AI news and trends, show you how you can use AI in your business and personal life, and help make it super easy for you to understand and Execute.

We're going to equip you with the knowledge to thrive in this upcoming wave of change.

The world is changing and like the current leaders, like don't understand how this technology works and like it's really difficult to adapt to that change if you don't Understand how the technology works.

Dude, if you don't change it dramatically, no one's going to realize you're different from Google.

They're just going to think you're Google with some add-ons, even if you're worse than Google, even if you suck at navigational queries that people go back to Google.

It's going to register in their mind why you're different and better.

On this podcast, we're going to be having some amazing guests like people from Microsoft, Google, OpenAI, Perplexity, and more to give you the inside scoop on what's Happening in AI.

The interesting thing about AI is everyone from your barber to Jeff Bezos is interested in AI.

These models that are coming out from these not OpenAI companies are getting really, really good.

Yeah, and you know when you're saying 4X, I think that's a very conservative number, like 100 to 1,000X territory.

A lot of things have to be rethought, so that's super exciting.

I can't be more excited to share those episodes that we've been working on with you.

We have a brand new YouTube channel just for the show.

I will make sure it's linked up below, but you can find it over at youtube.com at the next wave pod, or you can find it wherever you listen to podcasts, Apple Podcasts, Spotify, any of those places.

It's going to be everywhere.

I hope you tune in with me for these longer, forb discussions and deep dives into the world of AI because I can't wait to share them with you.

And that is all I got for you today.

Thank you so much for tuning into this episode.

I hope you enjoyed it.

I hope you learned something new and are keeping your finger on the pulse of AI news with me.

Thank you once again to Magical for sponsoring this video.

You guys rock and I'll see everyone in the next video.

Bye bye.