Latest / In the Loop AI / Meta Muse, Grok Bot, OpenAI Dots: Which AI Agent Should You Use?

Transcript

Meta Muse, Grok Bot, OpenAI Dots: Which AI Agent Should You Use? · In the Loop AI

0:00
Ejaaz: For the last three years, AI has been something that we talk to.
0:03
Ejaaz: We would type to it and then it types something back to us. And then we would
0:06
Ejaaz: take that information and go do something useful with it.
0:09
Ejaaz: But over the last eight weeks, that's changed pretty dramatically.
0:12
Ejaaz: And it comes in the form of this thing called an AI agent.
0:16
Ejaaz: Now, this is a term that has been floated across many different headlines,
0:19
Ejaaz: but no one quite understands what an AI agent is.
0:23
Ejaaz: The best way to think about it is it's an AI model that has the means to do things for you.
0:28
Ejaaz: It's like giving an AI model arms and the legs, and it's able to kind of run
0:31
Ejaaz: around and do various different tasks for you.
0:33
Ejaaz: Examples could be download or order that book that's been on your list for ages
0:38
Ejaaz: so that you can read it, filing your tax forms or organizing dinners between
0:41
Ejaaz: friends in your text chat. And there's so much more that these AI agents can do.
0:45
Ejaaz: So you might expect that the most popular AI agents come from the likes of Anthropic
0:49
Ejaaz: and maybe even OpenAI, but that's not actually true.
0:52
Ejaaz: The most popular agents come from the likes of Meta in the form of MetaMuse.
0:56
Ejaaz: You've got Elon Musk with GrokBot, you've got OpenAI that released Dots just
1:01
Ejaaz: last week, and I've been playing around with it, testing it out.
1:04
Ejaaz: And finally, you have this app called Instinct, which weirdly isn't actually
1:08
Ejaaz: an app. It just kind of lives everywhere and anywhere that you touch a digital surface.
1:13
Ejaaz: It was created by this 23-year-old and it's worth billions of dollars.
1:16
Ejaaz: So I've been playing around with these things for the last couple of weeks.
1:20
Ejaaz: Millions of other people have downloaded it and been doing the same.
1:23
Ejaaz: And I've been trying to answer two very important questions.
1:26
Ejaaz: One, what are the differences between these AI agents? Which one should I use?
1:30
Ejaaz: And two, what's the point of even using an AI agent in the first place?
1:33
Ejaaz: Should I materially move from using an LLM or a chatbot?
1:36
Ejaaz: Does it actually add anything to my life? Do I even care about productivity that much?
1:40
Ejaaz: Why should I use an AI agent? So we're going to dig into all of that and more
1:44
Ejaaz: on today's episode of In The Loop brought to you by Qualcomm.
1:51
Ejaaz: But before I do that, I want to address the elephant in the room.
1:54
Ejaaz: For all of you limitless listeners out there, you might be wondering what on
1:59
Ejaaz: earth is In The Loop and where are you, Ejaz?
2:02
Ejaaz: Well, let me start with the first question. In The Loop is the rebranded title of this show.
2:07
Ejaaz: It is obviously missing a very important co-host.
2:11
Ejaaz: I miss you, Josh, if you're watching this, but it now currently involves me
2:14
Ejaaz: and it's kind of like a rebrand, a refresh of what we're trying to do here.
2:18
Ejaaz: I'm deeply passionate, as you guys already know, about all the stuff that is
2:22
Ejaaz: happening on AI and Frontier Tech.
2:24
Ejaaz: And with this new show, we're going to bring it to you five times a week,
2:27
Ejaaz: Monday to Friday, so we don't skip any weekdays.
2:29
Ejaaz: You'll be kept up to date with anything and everything that's going along.
2:32
Ejaaz: Bite-sized chunks, 20 to 25 minutes so that you can listen to it on your commute
2:35
Ejaaz: to work or if you're cooking at home or whatever, if you just like or take pleasure
2:40
Ejaaz: in listening to our episodes or our kind of content.
2:42
Ejaaz: And yeah, it's just the rebranding of the show, I guess. And so I don't want that to be a mystery.
2:47
Ejaaz: If you're still wondering what on earth has happened, whether Josh and I have
2:50
Ejaaz: broken up, I can happily say that we have not. Josh has moved on to bigger and
2:54
Ejaaz: better things. And we actually put out a whole episode about this yesterday.
2:58
Ejaaz: So definitely go check it out on our channel. And to answer the second question.
3:02
Ejaaz: I'm in a pretty new swanky studio, and it feels pretty awesome.
3:05
Ejaaz: This is a major upgrade from the bedroom that we started recording the show
3:10
Ejaaz: around like 234 episodes ago, or whatever that might be.
3:14
Ejaaz: But I did bring a piece of home with me. This piece of art has been with me
3:18
Ejaaz: for a while. And so it's lovely to bring it here. So yeah, new studio,
3:22
Ejaaz: new show, excited to get going.
3:24
Ejaaz: Okay, so firstly, what the hell is even an AI agent?
3:27
Ejaaz: And I created this fun little animation with Claude to kind of describe this to you.
3:31
Ejaaz: So typically, when we're interacting with the web, we have to navigate to individual
3:36
Ejaaz: different websites, every single app, we need to kind of like scroll navigate
3:40
Ejaaz: through buttons, try and figure out where the product is or where the services
3:44
Ejaaz: or where that button is, click on it.
3:46
Ejaaz: And it's kind of an archaic whole mess. What I'm trying to demonstrate with
3:50
Ejaaz: this animation or what AI agents do specifically is it kind of coagulates all of that for you.
3:56
Ejaaz: It kind of like becomes your center nexus or your medium of interaction with
4:01
Ejaaz: the rest of the digital world.
4:02
Ejaaz: And so it ends up becoming, I guess, a lot less messy and much easier to interact
4:08
Ejaaz: with instead of having to find a website or navigating an app to download it or finding that button.
4:13
Ejaaz: You just speak to an AI agent, ask for what you want, and it magically appears
4:18
Ejaaz: or it magically gets done for you. The other cool thing about this that isn't
4:21
Ejaaz: shown on this animation is the AI agent can intuitively predict what you might want to get done.
4:27
Ejaaz: And I think this solves a very big problem in AI, which is.
4:32
Ejaaz: 99% of people don't know what to use their AI model for.
4:35
Ejaaz: I'm certainly in that category where I see all these new models getting released
4:39
Ejaaz: and I'm wondering, hmm, what can this do for me?
4:43
Ejaaz: What's the net new thing? If I don't code, if I am not a designer at heart,
4:47
Ejaaz: how can I use this as a normal person and create a more productive lifestyle
4:50
Ejaaz: for me or just kind of free up more time for me?
4:52
Ejaaz: The idea of a personal agent or these new AI agents that we're about to get
4:56
Ejaaz: into is that it's not just focused on productivity. Some of them are in the
5:00
Ejaaz: likes of GrokBot and OpenAI's Dots.
5:03
Ejaaz: But when you look at MetaMuse, when you look at Instinct, it becomes a sort
5:06
Ejaaz: of lifestyle product. And that's something that I think a lot of people would
5:08
Ejaaz: benefit from. But enough of this. Let's get into the actual agents, pros and cons.
5:13
Ejaaz: So the first AI agent also happens to be my current favorite personal assistant.
5:17
Ejaaz: It's the number one agent in all app stores and it's called MetaMuse. It's by Meta.
5:22
Ejaaz: And Muse is this kind of very intuitive, very human feeling ai agent you kind
5:28
Ejaaz: of download an app and the setup takes around five seconds it's as easy as adding
5:32
Ejaaz: your number stating your name and then you're introduced to this weird
5:37
Ejaaz: wonderful looking blob like character i'm trying to find a demo here that demonstrates
5:42
Ejaaz: it oh yeah over here where we got zuck kind of like casually leaning against
5:46
Ejaaz: a shelf and we're greeted by this kind of weird blob.
5:49
Ejaaz: Alien slash teddy bear looking creature and it talks to you very normally now
5:53
Ejaaz: a few things instantly hit me when I downloaded this AI agent.
5:56
Ejaaz: One, setup was super easy.
5:58
Ejaaz: But two, it wasn't bombarding me with text. It was almost just kind of short
6:04
Ejaaz: form, kind of, hey, what's up? How can I help you?
6:08
Ejaaz: Is there anything that's on your mind? And what I liked about it is it could
6:12
Ejaaz: proactively kind of predict what I might want.
6:15
Ejaaz: Now, one way that it was able to do this is it got access to my text.
6:20
Ejaaz: Now, it's very important to point out that, listen, you don't want AI agents
6:23
Ejaaz: randomly going out to text your friends or family or whatever that might be.
6:27
Ejaaz: Metamuse can't do that. It can't do a lot of things without your required approval.
6:32
Ejaaz: And you can determine when and where that makes the most sense.
6:35
Ejaaz: But because it was able to have the context of that it knew that I was discussing
6:39
Ejaaz: with my fiance something that I wanted to buy on Amazon and it came to me and it was like hey like.
6:44
Ejaaz: I saw that you were looking for some things, I actually found a better option
6:47
Ejaaz: for you. That was literally the first message that it sent me.
6:49
Ejaaz: And I thought, okay, that's weird.
6:51
Ejaaz: It kind of feels a little bit invasive.
6:55
Ejaaz: But I'm willing to trust this a little bit more because it comes from a reputable
7:00
Ejaaz: firm such as Meta. Now I get that Meta has had privacy issues in the past.
7:05
Ejaaz: And one thing that Zuck has made very clear, especially in this clip that I'm
7:08
Ejaaz: showing you here, when they announced this at Meta Connect, is that privacy
7:12
Ejaaz: is at the core of this entire product. And I'll get into where exactly that
7:15
Ejaaz: applies in a second. But first, the actual stats behind this AI agent, it's been downloaded.
7:20
Ejaaz: I think this figure is actually outdated. It's been downloaded over 3.5 million times.
7:24
Ejaaz: And as of two days ago, I saw a stat that basically said that they had 2 million
7:29
Ejaaz: daily active users, which is pretty insane.
7:32
Ejaaz: Now, granted, Meta's user magnitude is much larger than that.
7:36
Ejaaz: I think they have about 3.5 billion users across Instagram, WhatsApp, and Facebook.
7:41
Ejaaz: But the fact that they're able to get this much usage on a new novel AI product is pretty cool.
7:48
Ejaaz: They're also constrained by compute. That's partly why it's only rolled out
7:50
Ejaaz: in America. But if you're living in America right now, you get access to this
7:53
Ejaaz: thing and you can play around with it.
7:55
Ejaaz: So what exactly can this do for you? Well, I'll speak personally from my own
7:59
Ejaaz: anecdote on this tweet that actually went viral.
8:02
Ejaaz: I was basically on my way to a destination and my flight, my second flight,
8:07
Ejaaz: got delayed seven hours. So I was stranded in this airport, and it was the most
8:11
Ejaaz: frustrating experience ever.
8:13
Ejaaz: So I was bored and I brought up Muse and I said, listen.
8:18
Ejaaz: My flight's been delayed seven hours. Can you file for compensation for me?
8:22
Ejaaz: I had no idea whether I qualified for compensation. I had no idea about my airline's
8:27
Ejaaz: policies or guidelines.
8:29
Ejaaz: I just told Muse, hey, my second flight's been delayed. Is there anything you
8:33
Ejaaz: can do to help me out and figure out getting compensation?
8:36
Ejaaz: And it did a miraculous thing.
8:38
Ejaaz: In exactly five minutes, so five minutes later from writing that prompt,
8:43
Ejaaz: I had $250 worth of Delta credits for future flights in my account.
8:49
Ejaaz: And it even helped me rebook a new flight, which got me out to my destination
8:53
Ejaaz: much faster than I would have if I had to wait for my delayed flight.
8:57
Ejaaz: Now a few things to point out here. One, I gave it no context about my flights.
9:03
Ejaaz: But because it was able to read my text, it saw that I had a message or notification
9:09
Ejaaz: from Delta an hour earlier about this particular delay.
9:13
Ejaaz: So it knew exactly what flight details I was on, which flight or airline carriage
9:19
Ejaaz: got delayed and where I was routed to or where I was trying to get to.
9:23
Ejaaz: And it didn't regurgitate all of that back to me. It just kind of knew and responded
9:27
Ejaaz: intuitively to me, said, listen, I got you, I'll figure this out.
9:31
Ejaaz: Number two, it was able to navigate to the Delta website. So it didn't connect
9:36
Ejaaz: to any of my Delta logins on my Delta apps.
9:38
Ejaaz: It just went to the main website. It found the right contact details. It spun up an email.
9:43
Ejaaz: It added my own email address. And then it wrote out the entire thing and filed the entire form.
9:48
Ejaaz: It added my SkyMiles number. It did a whole load of things that would have basically
9:52
Ejaaz: taken me, I want to say 30 to 40 minutes. It would have taken me 30 minutes
9:56
Ejaaz: alone to find the actual page. It's like hidden under like several different pages.
10:00
Ejaaz: Muse did it all autonomously by spinning up its own desktop browser.
10:04
Ejaaz: So this leads me to how this thing actually works.
10:08
Ejaaz: When you think of an AI model, typically you just talk to it and it gives you
10:12
Ejaaz: back a response built on the knowledge that it already has.
10:15
Ejaaz: With an AI agent like Muse, it is given its own computer. It's known as a virtual
10:20
Ejaaz: machine or a virtual computer.
10:22
Ejaaz: Think of this as like your laptop or your desktop that you see in front of your
10:25
Ejaaz: screen or on your phone right now, but it's given to this AI LLM that can basically
10:31
Ejaaz: do the exact same things that you can, but on its own instance.
10:34
Ejaaz: Now it's important to note here that it doesn't share the same desktop as you.
10:38
Ejaaz: It can't see your personal files. It can't see anything that you have on your
10:41
Ejaaz: own computer. This is a completely separate thing, but it can do all the things
10:45
Ejaaz: that you can do. It can download various different plugins. It can navigate on a browser.
10:49
Ejaaz: It can open up tabs in the same way that you do and figure out all of this autonomously
10:53
Ejaaz: at lightning speed because it's AI and it's software.
10:57
Ejaaz: So it just kind of like repeatedly does that and it never gets bored, which is amazing.
11:00
Ejaaz: And this is just one of very many examples that i've had i've seen a guy take
11:05
Ejaaz: a picture on his walk every morning of telephone wires that were dangling down
11:11
Ejaaz: from trees or from exposed wires.
11:14
Ejaaz: And ask his Muse agent to file a complaint to Verizon to send a technician out there to go and fix it.
11:21
Ejaaz: And it just did it completely autonomously. And he got an email an hour later
11:25
Ejaaz: saying, it's been scheduled, you don't need to be there, technician will fix
11:28
Ejaaz: this over the next 48 hours.
11:31
Ejaaz: I have another personal anecdote where I met this exec, Qualcomm recently,
11:35
Ejaaz: and she was like, I don't get the fuss around AI agents. I don't think this
11:38
Ejaaz: is something that I might, you know, use in my personal life.
11:41
Ejaaz: And I managed to convince her to play around with Muse because,
11:44
Ejaaz: you know, she had a bunch of photos scattered around Facebook,
11:47
Ejaaz: WhatsApp, her iPhotos and stuff like that.
11:50
Ejaaz: And she wrote me a very lengthy email a few days later, basically saying,
11:54
Ejaaz: this has completely changed my life. And I'm using it for so many different things.
11:58
Ejaaz: She has a book club, she gets it to organize ordering books for everyone,
12:01
Ejaaz: sending it to their various different addresses, she does the same thing for
12:04
Ejaaz: movie recommendations.
12:05
Ejaaz: So the point here is, I would describe Muse as a lifestyle change,
12:10
Ejaaz: a lifestyle product that is powered by AI.
12:14
Ejaaz: AI isn't front and center. It's just a really good product. And actually,
12:18
Ejaaz: when you uncover the hood, it's not even a frontier model that is doing all
12:22
Ejaaz: of this for you, which is the next interesting thing, I guess.
12:25
Ejaaz: It is Muse Spark 1.2 or 1.3, whatever the latest meta model is. So for any of.
12:33
Ejaaz: Meta's models aren't as smart as Claude Fable 5.1 or GPT 6.1 Astra,
12:38
Ejaaz: which are the frontier leading models.
12:40
Ejaaz: And yet it has a better product experience than the agents that those two companies have built.
12:45
Ejaaz: So it's this crazy thing where Meta has an advantage of having all the data
12:50
Ejaaz: built off of your Instagram, WhatsApp messages, and they kind of intuitively know what you want.
12:55
Ejaaz: They know how to speak to individuals. They know how to build a really good
12:59
Ejaaz: consumer product. And that's how I would describe Metamuse.
13:02
Ejaaz: If this is something that you haven't tried it, I would definitely recommend.
13:05
Ejaaz: Now, you might be wondering on that note, is this thing free?
13:08
Ejaaz: Do I need to pay for another subscription?
13:11
Ejaaz: Or is this something that I can just kind of flip on, it connects to any of
13:15
Ejaaz: my existing accounts, and I can just use it?
13:18
Ejaaz: It's the latter. It's that answer that I just gave you. Number one,
13:21
Ejaaz: it's completely free to use up till I think 100 million token cap per week.
13:26
Ejaaz: Now, let me tell you, if you're hitting 100 million tokens worth of usage on
13:30
Ejaaz: this AI agent, per week, you are doing something else.
13:33
Ejaaz: You are running multiple businesses, and there's no other reason that the average
13:37
Ejaaz: person needs to be hitting that cap.
13:38
Ejaaz: But if you end up do loving Muse so much and it hits every single aspect of
13:42
Ejaaz: your life, and I don't know, it's automating your grocery list,
13:45
Ejaaz: it's cooking for you, it's controlling your lights or whatever that might be,
13:48
Ejaaz: you might hit that 100 million cap.
13:51
Ejaaz: And if so, it only costs you $20 a month, which granted is the cost of an average
13:56
Ejaaz: AI subscription, but it gets you to, I think it's like 500 million tokens,
14:00
Ejaaz: which again is so plentiful. I cannot emphasize how much this is.
14:06
Ejaaz: You don't need to worry. I think the free version is all good.
14:08
Ejaaz: And the cool part is you can literally give it access to your ChatGPT subscription
14:12
Ejaaz: or your Claude subscription, and you can just get this agent to do anything and everything for you.
14:16
Ejaaz: And finally, a last point on the privacy thing that I mentioned earlier.
14:20
Ejaaz: Now, listen, Meta's had a bad rap when it comes to privacy.
14:24
Ejaaz: People haven't forgotten the Cambridge Analytica stuff and so they're kind of
14:28
Ejaaz: wary when they use Meta's products. They.
14:31
Ejaaz: Predict, or I would assume people predict that Meta would use this information
14:36
Ejaaz: to benefit their products so that they make it more lucrative or stickier to
14:40
Ejaaz: use, which gets you kind of addicted to their products.
14:42
Ejaaz: Now, one thing Zuck made very clear when he announced or launched Muse is that
14:46
Ejaaz: it's built around privacy.
14:48
Ejaaz: And they've done this in a very particular way. I'm not going to get into the
14:51
Ejaaz: architecture, but it basically can be thought of as follows.
14:54
Ejaaz: Imagine a hotel. It's a luxurious hotel and it has many rooms hundreds of thousands of rooms
15:01
Ejaaz: each room is dedicated to a particular agent so when you spin up a muse agent
15:06
Ejaaz: when you open up the app and you sign up you have your very own agent that has
15:10
Ejaaz: its own room now that room can't be accessed by any other individual.
15:15
Ejaaz: No other user of muse can get any kind of access to your particular room and
15:20
Ejaaz: in that room there's the equivalent of a
15:23
Ejaaz: hotel safe and in that hotel safe is where all your passwords and credentials
15:28
Ejaaz: get saved and encrypted so that not even meta can get access to it
15:32
Ejaaz: if you deem it so if you want to give your agent access to your email or if
15:36
Ejaaz: you want to give it access to i don't know
15:38
Ejaaz: some account that's secured by a password you can do so confidently without
15:42
Ejaaz: worrying that anyone could get access to it now as for the room itself
15:47
Ejaaz: you know it's a hotel so you have other agents wandering around the same kind
15:51
Ejaaz: of system meta has the equivalent of a concierge or a sentry guard which basically
15:57
Ejaaz: monitors anything and everything that comes in
16:00
Ejaaz: and comes out of your interaction with your muse agent
16:04
Ejaaz: and so if that flags any kind of issue
16:07
Ejaaz: or malicious spam or whatever that might be it instantly cuts it out and so
16:11
Ejaaz: far it's been an extremely robust method of defending against any kind of attacks
16:16
Ejaaz: i haven't come across anything and i actually haven't seen any other instance
16:19
Ejaaz: of people suffering from incidents like that.
16:21
Ejaaz: So the point is it's private and it can be used for some really cool things.
16:24
Ejaaz: The last thing actually I'll add about this is from a business perspective.
16:28
Ejaaz: It is crazy what you can do with this agent. I've seen examples of people who,
16:33
Ejaaz: run ads on Instagram or who have competitors that post social media content on Instagram.
16:39
Ejaaz: And they've asked their muse, hey, could you go check out this person's business?
16:44
Ejaaz: Could you kind of like look at the style of their content and maybe recreate the same for me?
16:48
Ejaaz: Or actually, could you check out their follower list and DM all of them and
16:52
Ejaaz: say, hey, we're offering you a 50% discount, go use our product.
16:55
Ejaaz: And muse agent will just go out and do that. So it's this weird mixture of like
17:00
Ejaaz: a lifestyle thing and a really good business application, but it's not too intimidating as an AI agent.
17:05
Ejaaz: I don't know, it's very cool. Okay, so the next one is from our lord and savior,
17:08
Ejaaz: our favorite man ever, Elon Musk. So we're gonna dig into all of that and more
17:13
Ejaaz: on today's episode of In The Loop, brought to you by Qualcomm.
17:20
Ejaaz: Now, Elon's been busy. He's been busy acquiring all the compute that he can
17:25
Ejaaz: for SpaceX AI. I actually heard that it's called, it's being rebranded to SpaceX.
17:30
Ejaaz: SI, super intelligence and rebranding that Trump did. It's a whole thing.
17:35
Ejaaz: But he's been raising money to buy compute. He's IPO'd one of the biggest AI
17:39
Ejaaz: IPOs to date prior to Anthropics. That's reportedly coming in a few weeks.
17:43
Ejaaz: And SpaceX has, for all intents and purposes, still fallen behind on frontier AI models.
17:50
Ejaaz: Grok 4.7 got released. And listen, it's a great model.
17:53
Ejaaz: It's very cheap per unit cost of intelligence, but it's by no means frontier.
17:58
Ejaaz: And Elon Musk has been very upfront about this. He said that Grok 5 will eventually
18:02
Ejaaz: come out by the end of the year, and that will be competitive to something that
18:05
Ejaaz: we're seeing with Fable right now.
18:07
Ejaaz: But when it comes to consumer products, I would say that SpaceX,
18:11
Ejaaz: or X specifically, has a real advantage because they have millions and millions
18:15
Ejaaz: of users all over the world that uses X to transact money, to store money like a bank.
18:20
Ejaaz: XMoney launched, I think, a few months ago, and to share the world's most important
18:25
Ejaaz: news and information on the actual social media site itself.
18:28
Ejaaz: So you could argue that it could build something similar to what Meta has with Muse.
18:33
Ejaaz: And that's actually what they ended up doing. Now, GrokBot is unique in a few different ways.
18:38
Ejaaz: GrokBot can be thought of as not just a single AI agent, but a single agent
18:44
Ejaaz: collectively part of a larger swarm of agents.
18:48
Ejaaz: Now, there's a lot of technical jargon there, but what that basically means
18:51
Ejaaz: is, instead of one agent working for you, there are hundreds,
18:54
Ejaaz: maybe even thousands of agents working for you.
18:56
Ejaaz: But they all get access to a computer in the very same way that Metamuse is able to do that.
19:01
Ejaaz: And the focus of GrokBot isn't necessarily around lifestyle,
19:05
Ejaaz: although it's very good at doing that in my personal experience.
19:08
Ejaaz: But it's more focused around productivity. Even in this demo that you see in
19:12
Ejaaz: front of you, GrokBot basically gets access to any high-powered tools that you
19:17
Ejaaz: might use on personal life or productive enterprise work kind of life.
19:22
Ejaaz: So think about Slack, think about CRMs to manage your customer or your business,
19:27
Ejaaz: think about Notion, Jira, Atlassian, all the work productivity apps that you
19:32
Ejaaz: might use on a day-to-day basis, you can give GrokBot access to.
19:36
Ejaaz: It stores your password safely encrypted it so that it doesn't leak any of that stuff.
19:40
Ejaaz: And it is able to actually conduct really good knowledge work.
19:43
Ejaaz: Now, in my experience, when I'm doing sorts of research around particular episodes
19:49
Ejaaz: that we're going to do, creating demos, coming up with ideas of what to talk
19:52
Ejaaz: to you guys about, this has been a really good colleague.
19:56
Ejaaz: And I mean that term intentionally, it's can be thought of as like a work partner
19:59
Ejaaz: and less of something that can be like a friend, a mentor, or something that
20:04
Ejaaz: can kind of free up your time like Metamuse.
20:06
Ejaaz: Now, the coolest feature by far with GrokBot is that it can monitor tasks that
20:10
Ejaaz: you do autonomously and then eventually replicate you doing those exact tasks on its own separately.
20:17
Ejaaz: Let me repeat that. Let's say that an AI has no idea about a task that you know
20:23
Ejaaz: very well to do, that you've studied years trying to figure out,
20:26
Ejaaz: or that you just do naturally because you've been on the internet for decades at this point, right?
20:30
Ejaaz: GrokBot, or most AI models in general, usually lack the context.
20:34
Ejaaz: You usually have to spoon feed it. You need to explain, hey,
20:36
Ejaaz: this is what I'm doing and this is why I'm doing it.
20:38
Ejaaz: GrokBot sort of watches you. It sort of records your screen and kind of understands
20:43
Ejaaz: why you're doing what you're meant to be doing.
20:45
Ejaaz: And so the idea here is that eventually it becomes the perfect emulation or
20:49
Ejaaz: predictor of tasks that you might want to do on a daily basis.
20:52
Ejaaz: So eventually it frees you up to go and do other stuff and it can just kind
20:56
Ejaaz: of take all of this off your plate.
20:58
Ejaaz: I think that's really cool. Now, in terms of pricing, it comes bundled with
21:02
Ejaaz: Cursor. Cursor is the AI coding platform that SpaceX acquired shortly after
21:07
Ejaaz: their IPO for $60 billion.
21:09
Ejaaz: It's not free to access. You have to pay at minimum, I think it's $20 a month
21:14
Ejaaz: for very limited or not so limited weekly usage, but it's capped.
21:18
Ejaaz: It's not as free as MetaMuse with 100 million tokens. I think it's a little less than that.
21:23
Ejaaz: But you could increase that limit if you forked out $60 a month with Cursor
21:29
Ejaaz: Pro Plus. So it's more of a bundle thing. It's certainly not free.
21:33
Ejaaz: And if you're going to end up using this platform, I think it's something that
21:36
Ejaaz: you should trial out for your work stuff.
21:38
Ejaaz: That's my personal take. I think if you want to play around with an AI agent
21:42
Ejaaz: and really understand the magic behind it and you don't have an explicit use
21:45
Ejaaz: case, you kind of are confused about what you might want to use it for,
21:48
Ejaaz: go out and use Metamuse. That's the best thing that's available right now.
21:51
Ejaaz: But the final thing I'll say about GrokBot is spinning up multiple agents to
21:56
Ejaaz: do your work for you was way more productive than I thought it would be,
21:59
Ejaaz: but it's also a little confusing. Now, what I like about,
22:04
Ejaaz: I guess, an app like Muse or Instinct that I'm just about to talk about is that
22:09
Ejaaz: it's just a single interface. It's a single agent. So you could build a sort
22:13
Ejaaz: of relationship with that agent. You can't really do that.
22:17
Ejaaz: I don't get that sense from using Crock-Bot. It is more like working with a workforce.
22:22
Ejaaz: They're a little more robotic and mechanical, but extremely productive.
22:25
Ejaaz: So again, if you're an enterprise that's trying to figure out whether using
22:28
Ejaaz: AI agents versus models are more productive and cheaper for you.
22:32
Ejaaz: Go try it out. And now the third AI agent is one that weirdly doesn't have a
22:37
Ejaaz: home. It's called Instinct. And it was this stealth startup,
22:41
Ejaaz: which I think came out of stealth a few months ago, but it's already been downloaded over 500,000 times.
22:47
Ejaaz: And technically, it's still in private beta. It was started by a 23 year old
22:52
Ejaaz: that has a lot of history in building AI type products.
22:55
Ejaaz: His name is Noah Shin, and he's an absolute genius. He used to work for this
22:59
Ejaaz: company called Sierra AI, which was a customer service focused AI.
23:03
Ejaaz: But what I'm really impressed by is this paper called Reflexion that he wrote, I think back in 2023.
23:10
Ejaaz: And it is the most cited research paper for AI agents ever.
23:14
Ejaaz: And so I looked at this paper and the core idea behind it is if you can instruct
23:19
Ejaaz: an AI or an AI agent to write down on a document.
23:24
Ejaaz: The core reason why it made a mistake or why it can complete the task and then
23:29
Ejaaz: get it to rerun that task, it does a much better job.
23:33
Ejaaz: And this was famously used and seen in GPT-4, which had a coding pass rate back then of 80%.
23:40
Ejaaz: And because of this paper, because of Noah Shin, it got pushed to 91%.
23:44
Ejaaz: So Noah has a lot of background and credibility in two of the most important
23:49
Ejaaz: things, building bleeding edge frontier AI, and also customer service.
23:52
Ejaaz: So building a product that people actually want to use.
23:55
Ejaaz: And that's really the manifestation of this product called instinct.
23:59
Ejaaz: Now, firstly, I mentioned earlier, it doesn't have a home because it literally doesn't have a home.
24:04
Ejaaz: When I got an invite link to this thing, I just tapped the link,
24:07
Ejaaz: I gave it my number, and then suddenly it just kind of.
24:11
Ejaaz: Appeared. It texted me on my iMessage. And so I opened up my iMessage and I
24:16
Ejaaz: responded saying hello.
24:18
Ejaaz: And it said, hey, I'm Instinct. What do you need a hand with?
24:21
Ejaaz: And it then started to be a lot more proactive in the same way that Meta Muse was.
24:26
Ejaaz: Now, what's cool about this compared to Muse is that there is no separate app
24:30
Ejaaz: that I need to kind of keep track of or open.
24:33
Ejaaz: It is just kind of this very imposing AI agent that is completely invisible.
24:38
Ejaaz: And I understand that that's completely an oxymoron, but it really is that.
24:42
Ejaaz: It is anywhere and everywhere that I am without being in my face and imposing
24:47
Ejaaz: and trying to control everything that I'm trying to do.
24:49
Ejaaz: It's just this kind of ambient AI, very much like a Jarvis from Tony Stark,
24:54
Ejaaz: any of you Iron Man fans out there.
24:56
Ejaaz: And it's really easy to use. So actually, it's one of my favorite products to
24:59
Ejaaz: use, but it is in private beta.
25:01
Ejaaz: And I do have a few comments on this, the main one being around the terms and
25:06
Ejaaz: agreement and privacy. Now, I just explained to you why Metamuse has been built
25:11
Ejaaz: around privacy, and it's taken a lot of feedback from previous past incidents. So.
25:15
Ejaaz: This hasn't been the case with Instinct. When you sign the terms and agreements
25:19
Ejaaz: and you give it access to your phone, and let's say you delete your account.
25:23
Ejaaz: It still has all the data that it had access to, that it's indexed,
25:27
Ejaaz: that it's able to train on and build a better model.
25:30
Ejaaz: And that's been a key reason why a lot of very prominent figures in AI themselves
25:34
Ejaaz: haven't signed on to this thing, because it's quite invasive compared to any
25:37
Ejaaz: other product that you could go out and use.
25:40
Ejaaz: The other thing is, I don't know what model is powering it behind the hood.
25:43
Ejaaz: And maybe you don't care at the end of the day, but I do like to know how my
25:46
Ejaaz: data is being used for whatever model behind the hood. And if it's a Chinese
25:50
Ejaaz: open source model, I might have different thoughts about it,
25:52
Ejaaz: whether it's that or versus a Western frontier model.
25:55
Ejaaz: But all that might still be worth it for a few different features.
25:59
Ejaaz: Number one, the price. It is completely free and being subsidized to high heaven.
26:04
Ejaaz: Now, when Instinct was pushed to comment on that business model,
26:07
Ejaaz: their response was, our aim is to basically take a percentage of every transaction
26:12
Ejaaz: that you might facilitate using Instinct.
26:15
Ejaaz: So the idea I get from there is there's going to be a heavy push on e-commerce
26:20
Ejaaz: via the Instinct agent. So it should be very easy for you to go out and buy
26:23
Ejaaz: stuff or engage with a product or service.
26:26
Ejaaz: And they'll just take a percentage cut from that, which makes sense.
26:30
Ejaaz: But I have a feeling that they can't subsidize this forever and people aren't
26:33
Ejaaz: gonna be fully trusting of an AI agent to go and do its shopping.
26:37
Ejaaz: It can't even come up with basic ideas of what to buy in the first place.
26:40
Ejaaz: So I feel like that's a bit more of a stretch when it comes to this.
26:43
Ejaaz: And so I don't think it'll be free forever, but certainly cool to try out for now.
26:46
Ejaaz: The second thing is I have watched Instinct literally call up.
26:51
Ejaaz: A service, in this case, a restaurant, and book a reservation for me.
26:55
Ejaaz: I was super curious about how this would work. And he got access to 11 labs,
26:59
Ejaaz: voice API, and was able to just do the whole thing completely autonomously,
27:02
Ejaaz: which was very weird and surreal to watch, but cool nonetheless.
27:06
Ejaaz: And the funniest meme-worthy example that I saw was this post by Aaron where
27:10
Ejaaz: he goes, I asked my instinct to ask my sister's instinct for her Netflix code.
27:15
Ejaaz: So AI agents speaking to other AI agents.
27:17
Ejaaz: Now I never need to talk to my family again, which I found hilarious.
27:20
Ejaaz: And I sent straight to my sister.
27:21
Ejaaz: Okay, and now, if you've stuck along this far, the final agent comes from none other than OpenAI.
27:28
Ejaaz: And this is more of a recent addition. This literally got released last week
27:31
Ejaaz: at OpenAI Dev Day, which is an annual conference where they release a bunch
27:35
Ejaaz: of things they've been cooking up all year.
27:37
Ejaaz: And one of their flagship announcements was this product called Dots,
27:40
Ejaaz: D-O-T-S, which is a cute and affectionate name to this weird little avatar thing
27:45
Ejaaz: that you're seeing on your screen right now.
27:47
Ejaaz: And in the promo video, which I think is actually the most interesting thing to comment on.
27:52
Ejaaz: No one touches a keyboard or a screen. They just kind of look at a screen and they speak.
27:57
Ejaaz: And these AI models or these AI cute blob agent-like things can run around and go do a bunch of things.
28:03
Ejaaz: Now, the examples that they give were mainly pushed on the work-oriented side
28:09
Ejaaz: of things. So knowledge work, enterprise work, things that you would do at your
28:12
Ejaaz: day job, less of a lifestyle thing.
28:14
Ejaaz: The second thing I'll point out is it's very intuitive when it comes to using
28:19
Ejaaz: tools. It can kind of like connect to any tools, Slack, work interface,
28:22
Ejaaz: whatever that might be, in the same way that GrokBot does that for SpaceX AI.
28:27
Ejaaz: And it can go out and do stuff whilst you sleep. It can work 24-7 around the clock.
28:32
Ejaaz: Now, this thing can connect to over 4,000 tools, apps, and services,
28:37
Ejaaz: which I believe is the largest amount for all the AI agents that we've covered
28:41
Ejaaz: so far in this episode, which is cool and productive if, I don't know,
28:44
Ejaaz: you're a large enterprise.
28:46
Ejaaz: But for the average person, I have a feeling you don't care.
28:48
Ejaaz: I have a feeling you can't come up with even three tools that it needs to connect to.
28:51
Ejaaz: So I would err on the side of, again, using something like Instinct or Metamuse
28:56
Ejaaz: to kind of trial out this whole thing unless you're an enterprise customer.
28:59
Ejaaz: Now, the other thing is the pricing. You cannot get access to this thing unless
29:04
Ejaaz: you have a pro account, which I think runs you between $150 to $200.
29:10
Ejaaz: If you're an enterprise account, but you pay more, you should also get access
29:14
Ejaaz: to this thing. So it's very expensive. It's actually the steepest price upfront
29:17
Ejaaz: that you need to pay to get access to an AI agent. So again,
29:20
Ejaaz: I implore to you that if you're just a regular human being that doesn't need to run a.
29:25
Ejaaz: Business making 10 to 100 million dollars a year, I don't think this is something
29:29
Ejaaz: that you necessarily need to care about.
29:31
Ejaaz: The other thing I'll say is, it's still unclear to me why I should use this agent.
29:36
Ejaaz: And I say that independently of all the other agents that we've covered so far.
29:39
Ejaaz: Firstly, the live demos that they did completely flopped, like two out of three
29:43
Ejaaz: didn't work. And the third one was cool. But like, I was still like, okay, this looks bad.
29:48
Ejaaz: And so I still think they're kind of figuring out a few kinks and trying to
29:51
Ejaaz: figure out how this actually applies to the majority of people.
29:54
Ejaaz: And that's been the general reaction that I've seen from the public.
29:57
Ejaaz: The other thing I'll say is OpenAI has spent hundreds of billions of dollars
30:00
Ejaaz: at this point trying to acquire as much compute to serve products like this.
30:05
Ejaaz: So why then is it being rolled out in the slowest manner possible?
30:09
Ejaaz: I know many people that have pro accounts that still don't have access to their own dots.
30:15
Ejaaz: And so when it comes to actually wanting to engage with this product post-announcement,
30:19
Ejaaz: it's been like a week and a half now, a lot of people still don't have access to this thing.
30:22
Ejaaz: And so they can't get feedback. So in my opinion, it's been a bit of a slow
30:25
Ejaaz: rollout, maybe wait a bit before you try a product like this out.
30:28
Ejaaz: Okay, so where do we stand collectively across all the agents that we covered
30:31
Ejaaz: today, I created this table to kind of like break down the kind of takeaways.
30:35
Ejaaz: But basically, if you want something that is more helpful towards your life,
30:40
Ejaaz: that you don't want to spend too much time thinking about that you just want
30:43
Ejaaz: to experience the magic of AI, and you don't care about the productivity or the KPIs that.
30:49
Ejaaz: Use metamuse that is like the best thing that you can use right now it's completely
30:53
Ejaaz: free it's cheap as hell even if you somehow
30:56
Ejaaz: surpassed 100 million per week token cap and it's very easy to use it's like
31:01
Ejaaz: talking to a friend it's the closest thing that i've come to interacting with,
31:05
Ejaaz: an artificial human being typically i can tell when lllm is speaking to me it's
31:10
Ejaaz: getting harder to do that with muse so that's the muse use case then moving
31:14
Ejaaz: next up i would say instinct instinct is also incredibly intuitive.
31:18
Ejaaz: It is more invisible than Metamuse, which is fantastic. And it is deeply integrated
31:23
Ejaaz: into all of your digital life, it can do a lot of things.
31:26
Ejaaz: But that comes at the major trade off of privacy and signing over your passwords,
31:30
Ejaaz: logins, and data that they can use to train God knows what in the future.
31:33
Ejaaz: So I would err on the side of caution when it comes to privacy.
31:37
Ejaaz: It's also private beta access. So maybe let's just wait until that becomes more
31:40
Ejaaz: public. And then you have GrokBot, which is kind of a weird hybrid between helping
31:44
Ejaaz: out with your personal life, but mainly helping out on your work life.
31:47
Ejaaz: And it does so in a very unique way where you can kind of like spin up a bunch
31:50
Ejaaz: of worker agents and it gets incredibly productive and connect to whatever tool you give it access to.
31:55
Ejaaz: And it's also relatively cheap.
31:56
Ejaaz: It's not exactly free, but for 20 bucks a month, you can get access to it.
32:00
Ejaaz: Then at the final end of that spectrum, you have OpenAI Dots,
32:03
Ejaaz: which is still rolling out very slowly. It is the most expensive tier,
32:07
Ejaaz: but it also is powered by the most powerful models in the world,
32:11
Ejaaz: which is GPT 6.1 Astra. So.
32:14
Ejaaz: Various different agents, various different options, various different reasons why you might use it.
32:18
Ejaaz: If you are someone that can relate to any of these, I would like to hear from
32:21
Ejaaz: you. If there's a different use case that I might have missed that applies to
32:25
Ejaaz: one of these agents, please let me know.
32:27
Ejaaz: Also, let me know how I did on my first episode without Josh.
32:30
Ejaaz: Josh, man, it is tough trying to figure this out without you,
32:33
Ejaaz: mate. Like it's weirdly with the co-host, I'm able to kind of like bounce back
32:37
Ejaaz: and forth and take a breath. I'm still adjusting to this. So please have a bit of lenience with me.
32:43
Ejaaz: I'm always welcome to feedback, by the way. So let me know in the comments.
32:46
Ejaaz: I know you guys are very good with comments and providing feedback.
32:49
Ejaaz: I read all of them and I'll respond to as many as I can.
32:51
Ejaaz: If there's anything that I missed in this breakdown, if there's an agent that's
32:54
Ejaaz: out there that maybe got released recently that I missed and that I should try
32:57
Ejaaz: out, please send it to me.
32:58
Ejaaz: I love to test out as many of these things as I can. And if you are listening
33:02
Ejaaz: to this and you aren't subscribed or if you're new here, please subscribe.
33:05
Ejaaz: It helps us out massively. Turn on notifications.
33:07
Ejaaz: We post five times a week. They're five episodes a week, 20 to 25 minutes on
33:11
Ejaaz: anything and everything AI.
33:13
Ejaaz: And we go way, way beyond the headline. And we try and give you some thoughtful,
33:16
Ejaaz: useful information about how you can actually use these things and how it's
33:19
Ejaaz: going to affect your life going forwards from an investment angle,
33:21
Ejaaz: as well as just a general productivity and vibes based thing.
33:25
Ejaaz: So I'm going to be here pretty much every day.
33:27
Ejaaz: And your feedback is very welcome. Share it with a friend if it's interesting.
33:30
Ejaaz: We're also on Spotify or on Apple Music.
33:33
Ejaaz: We're anywhere and everywhere that you can listen and watch a podcast show.
33:36
Ejaaz: And that's it. Thank you so much for listening to In The Loop,
33:39
Ejaaz: brought to you by Qualcomm. See you on the next one.