00:00:00
Hello, I'm Mike with this week's new
and trending AI tools.
00:00:02
You can't afford to miss.
00:00:04
This week's been huge because CES,
the Consumer Electronics Show,
00:00:08
has taken place in Las Vegas for 2025.
00:00:11
And I think the company that blew
everyone away was NVIDIA
00:00:15
with their talk of AI agents.
00:00:18
Jensen Huang gave an amazing keynote.
00:00:20
one of the most important things
00:00:22
that's happening
in the world of enterprise is agenetic AI.
00:00:25
So not only did
00:00:26
we learn about agenetic AI, but
we also saw a preview of those tasty GTX
00:00:31
50 series GPUs.
00:00:34
this is a good time to run local AI.
00:00:36
So a quick look at agentic AI.
00:00:38
It's like a team of AI models
all working together,
00:00:41
getting info and using tools to do things.
00:00:43
Now NVIDIA is making tools to build
these AI agents even easier.
00:00:48
So AI agents essentially are going
to become part of the digital workforce.
00:00:51
Or as Sam Altman would say,
they'll enter the workforce in 2025.
00:00:56
Let's take a look
at what NVIDIA have been building.
00:00:59
I'm pretty impressed with it.
00:01:01
If we go to try now on their AI website,
which of course I'll link down below,
00:01:05
you can create AI agents
and something really cool my AI,
00:01:09
in fact it's top at the moment
as I click in is the PDF to podcast agent.
00:01:15
So let's go in there
we get a broad overview.
00:01:18
We get the architecture diagram
here on what it's going to do.
00:01:21
You can see it does multiple things
to make this podcast.
00:01:25
It uses Llama LangChain
and a bunch of other cool stuff.
00:01:29
And it's kind of ready to go.
00:01:30
So let's deploy
and see if we can get it working.
00:01:33
Now I need to use an NVIDIA account.
00:01:35
Agree to the Terms of Service here.
00:01:37
As you can see right now,
it's deploying my GPU.
00:01:39
It's going to install the software
00:01:41
and give me access to my notebook
to run this agent.
00:01:43
This really is crazy
how easy it is to make things happen.
00:01:47
Okay.
00:01:47
Now, this is up and running,
and charging is $0.77 an hour.
00:01:50
We're going to go to the notebook
and follow the instructions,
00:01:53
which I'll fast forward through
so you can see how this works.
00:01:56
We'll start by cloning the repository.
00:01:58
Then we'll click here to get our NVIDIA
API key, which actually gives us
00:02:02
free credits to get started with.
00:02:04
Once we've got our NVIDIA
API key, and also one from ElevenLabs,
00:02:07
we'll go ahead and run the bash
script to configure it.
00:02:09
Then we'll install dependencies.
00:02:11
Next I'll make sure I've got Docker up
and running here on my desktop.
00:02:13
And I'll spin up the blueprint.
00:02:17
And there you go.
00:02:18
That took a few minutes
to get up and running.
00:02:19
But my AI agent is now active
using NVIDIA and ElevenLabs APIs.
00:02:24
Let's use another terminal window
to test it.
00:02:26
And yes, it's testing and we can see
things are up.
00:02:29
The agent is up.
00:02:31
It's ready to create a podcast.
00:02:33
Now we get on to the exciting stuff.
00:02:34
We can create a podcast using this Python
file they give and it will just run this.
00:02:39
I'm going to create a virtual Python
environment inside Cursor.
00:02:41
Now we can create that Python file.
00:02:45
Now I've edited the instructions here
to look for the create
00:02:47
a magic PDF to create a podcast
all about this YouTube channel.
00:02:51
Let's click play
00:02:52
and you can see it's submitted a job
and it's waiting to initialize.
00:02:55
And boom, all PDFs processed successfully.
00:02:58
It's summarizing again
using NVIDIA agents to do so.
00:03:02
Next agent
00:03:03
picked up the job and is generating
the initial outline of the show.
00:03:06
Creating the monologue. Transcript.
00:03:07
This is really cool stuff.
00:03:09
Formatting the final conversation
a bunch of agents all working together,
00:03:13
to make something amazing.
00:03:15
Now, the text to speech processing.
00:03:16
And the podcast generation is completed.
00:03:18
Let's have a listen.
00:03:20
Welcome to our analysis of Creator Magic,
a leading YouTube channel on AI powered
00:03:24
content creation with a consistent stream
of weekly updates.
00:03:28
Mike Russell shares his hands
on experience with the latest AI tools.
00:03:32
From video
editing to content summarization,
00:03:34
this broad scope and emphasis
on practical applications
00:03:37
make the channel a valuable resource
for content creators.
00:03:41
As the field of AI powered tools rapidly
grows, Creator
00:03:44
Magic is well positioned to adapt
and cover new developments.
00:03:48
and there you go.
00:03:48
Using NVIDIA's AI agent tools,
I've created my own set of agents
00:03:53
that find out information, scripts,
produce, and finally voice a podcast.
00:03:58
For me, this is the future.
00:04:00
I've basically built a Notebook
LM podcast creation clone
00:04:04
using NVIDIA agents, and I can't wait
to see what you're going to do with this.
00:04:09
Do let me know in the comments down below
00:04:10
because I'm there helping and join
my community as well.
00:04:13
If you want to get involved because
we're creating from the ground level
00:04:17
in the Creator Magic community,
and that link is also down below.
00:04:20
If you like what you see right
now, throw like
00:04:22
and subscribe to my channel for more stuff
just like this in the future.
00:04:25
Okay. Next up. This is insane.
00:04:27
Somebody built an open source
clone of Perplexity,
00:04:31
and it's actually available on GitHub.
00:04:34
So you can clone this
and you can run this yourself locally,
00:04:37
but they're actually hosting it here
at this web address.
00:04:40
Then of course I'll post down below
00:04:41
and you can ask questions
just like you can Perplexity.
00:04:44
Not only can it search across the web,
00:04:46
but it can also look at X
and YouTube as well.
00:04:49
So if we go in search X and we type in,
what did Jensen Huang
00:04:52
say about AI agents during his keynote
at CES 2025?
00:04:57
straight away.
00:04:58
And it's
searching for the latest news as well.
00:05:00
So it popped up
quickly, gave me an answer.
00:05:02
Then it's pulling in latest from X
and it's
00:05:05
saying that he shared his vision
for the future of AI agents.
00:05:09
It goes on it references Let's do a
new search and look for YouTube.
00:05:12
And again, we'll ask the same question
00:05:14
and see if it'll search YouTube videos
to find out.
00:05:17
And again, yeah, here we go.
00:05:18
It's searching through YouTube
to find the answer.
00:05:21
And there we go.
00:05:22
It's directly referencing transcripts
from YouTube videos to give me
00:05:26
a comprehensive breakdown
on everything that Jensen said
00:05:29
about AI agents at CES 2025.
00:05:32
That is incredible.
00:05:34
So you can test
this really cool tool out at
00:05:37
mplx.run, or you can go to GitHub again.
00:05:39
I'll post that link down below
00:05:40
so you can clone this repo
and maybe try it out
00:05:43
and maybe even alter it for yourself.
00:05:44
But it can do so much like web search.
00:05:46
You can do weather search,
it can look for maps, YouTube search.
00:05:50
I don't know how this developer
has managed to create
00:05:53
a Perplexity clone,
but the future is really exciting
00:05:57
because if Perplexity can be cloned, well,
so can every single tool out there.
00:06:02
Also worth to mention, it's
not the only game in town.
00:06:04
Perplexica is also a very busy GitHub
repo, which is an AI powered
00:06:10
search engine, again, very much
based on Perplexity, and you can clone it.
00:06:14
It's got 18,000 stars at the moment.
00:06:16
At the time of making this,
I round up so pretty cool.
00:06:19
And again,
another alternative to look into
00:06:22
if you're interested
in running your own Perplexity clone.
00:06:25
So seeing as this video
is all about AI agents.
00:06:28
Yes, AI agents that use NVIDIA
GPUs to create podcasts from your contents
00:06:33
to AI agents that go across the web
and crawl it for answers.
00:06:36
In the style of Perplexity,
I signed up again.
00:06:39
Yes, for Google Gemini.
00:06:42
The Google Gemini Advanced
is pretty incredible,
00:06:45
and they added a tool actually
at the end of last year,
00:06:48
which is Google 1.5
Pro with deep research.
00:06:52
Again,
you can get in-depth answers on anything.
00:06:55
And Google is using its own special
mix of AI agents.
00:06:59
I want to show you how cool this tool is.
00:07:01
So I can ask Gemini
when I'm in deep research mode.
00:07:04
Find out everything you can about
MiniPerplx, who's behind it, what it does,
00:07:08
and anything else. That's relevant.
00:07:10
And rather than give me an answer
straight away like most AI models will,
00:07:14
this will go across the web and use agents
to find out everything it can.
00:07:19
And in fact, right now it's creating me
a little plan of what it's going to do.
00:07:23
And it's told me it's going to find
the website, it's going to find news
00:07:27
articles, it's going to find the history,
find the founders, and a lot more.
00:07:30
So a bunch of AI agents are going to go
out, crawl the web and give me the answers
00:07:35
I need. I'm
loving deep research right now.
00:07:38
Let's start the research.
00:07:39
Now. As you'll see, this is not instant.
00:07:42
It actually goes out
and a genetically crawls the web and does
00:07:46
basically what I might spend all morning
doing in just a couple of minutes.
00:07:50
You'll see straight away
it's pulled up the site, the GitHub
00:07:53
repo of MiniPerplx,
and a bunch of other articles
00:07:57
all about this and related to MiniPerplx,
even YouTube videos.
00:08:01
So far it's looked at 23 websites
that would have taken me at least 10
00:08:05
or 15 minutes, and very shortly
will get the answers back.
00:08:08
Now, while it's doing that research,
I think it's really important to do this.
00:08:11
Let's look at Notebook LM.
00:08:13
I love Notebook LM, and you've seen other
videos where I spoken about it in depth.
00:08:17
Recording your thoughts, adding sources,
and creating AI generated podcasts.
00:08:22
Well, I realized when I signed up to my
Google One AI plan to get Gemini Advanced.
00:08:27
It also gave me better
limits for my notebooks.
00:08:30
So you'll see here.
00:08:31
I've been jamming about the Creator
Magic channel,
00:08:34
where I've essentially added
all of my videos to a notebook now.
00:08:37
Previously on the free plan from Google
50 sources was the maximum.
00:08:42
When I paid for an account with Google,
my source limit is 300,
00:08:46
so I can add a maximum of 300
YouTube videos to one notebook in Notebook
00:08:51
Elm and ask a bunch of questions
and get answers based on
00:08:55
all of the content on my channel.
00:08:57
Okay, with that said,
it seems to have researched 42 websites.
00:09:00
It's now analyzing the results, and
very soon it's going to build me a report.
00:09:04
And now you can see.
The ring is nearly closed here.
00:09:07
And it's creating my report.
00:09:09
Very soon I'll have something to look at.
00:09:11
And here it is, a deep dive
into the minimalist AI search engine.
00:09:15
It says that it's getting a significant
buzz right now in the tech community.
00:09:19
As an open source AI search engine.
00:09:21
As I scroll down, it'll show you
all of the features and functionality,
00:09:25
including that YouTube
00:09:27
search, academic search
that I've mentioned and X search as well.
00:09:30
It's all here in the summary.
00:09:32
now? Who's behind MiniPerplx?
00:09:34
Well, again Google
the research has found out.
00:09:36
It's the brainchild of Zaid Mukaddam,
an indie developer
00:09:40
with a background in software
engineering and cybersecurity.
00:09:43
Pretty cool AI research there.
00:09:46
now scrolling along here.
00:09:47
Media coverage.
00:09:48
It said.
00:09:48
During our research, we explored various
news sources, however, found limited news
00:09:52
coverage suggests the project is still in
its early stages, which indeed it is.
00:09:57
So Google Gemini
Deep Research has done an incredible job.
00:10:01
It provides a conclusion here
at the bottom, and it lists
00:10:04
all of the sources it crawled and all
of the websites to find the information.
00:10:09
And if I want to then save this for later
reference, I can open it in docs.
00:10:13
And boom!
00:10:14
Now I have a full report from Google
Gemini Deep Research that I can scour over
00:10:19
however I wish or share
with someone at the click of a button.
00:10:23
So there you go
00:10:24
NVIDIA talking about the AI agent future
that we're all heading towards.
00:10:27
MiniPerplx
for searching the web using free and open
00:10:31
source tools to look at YouTube
and other sources as well.
00:10:35
And Google Gemini advanced
with deep research to a genetically scour
00:10:40
the web and build you complex reports on
anything you can think of.
00:10:45
The world in 2025
is getting better and better by the day,
00:10:49
and I'm here every week
to bring you all of the AI tools
00:10:53
and ideas that you can use yourself
to make life incredible.
00:10:58
I hope this has helped you.
00:10:59
And by the way, if you enjoyed this video,
throw a like subscribe
00:11:02
and there's a video showing on your screen
now that you should probably watch next.
