Perplexity AI Masterclass: Search Will Never Be the Same

Summary

This tutorial provides a comprehensive overview of Perplexity AI, an answer engine that uses large language models to deliver accurate and reliable responses. It outlines key features like search modes, collections for organization, file uploads, and image generation. The video focuses on real-world use cases and prompting tips to get quality answers, positioning Perplexity as a time-saving tool for research, academics, writing, coding, and more. Overall, it makes a compelling case for Perplexity AI as an efficient way to get factual information across various professional settings.

Takeaways

• 😀 Perplexity AI is an answer engine that provides accurate and reliable responses to questions by using large language models
• 📚 It has 6 different search modes - All, Academic, Writing, Wolfram Alpha, YouTube and Reddit - each catered to specific needs
• 📝 The Writing mode allows you to generate text and code without accessing the internet
• 🔢 Wolfram Alpha mode is for mathematical and computational queries
• 📺 YouTube mode gives targeted video results from YouTube
• 🤔 Reddit mode summarizes Reddit discussions and opinions
• 🗂 Collections help you stay organized by grouping related threads together
• 🖼 It can accurately describe images and summarize CSV files uploaded to it
• ❓ Well-structured prompts with clear goals and keywords produce the best results
• 💡 Overall, Perplexity AI saves time by delivering fast, accurate answers from reliable sources

Q & A

What is the main purpose of Perplexity AI?

Perplexity AI aims to provide accurate and reliable answers to complex questions whenever people are searching for information.

What are the key features that make Perplexity AI unique?

Key features of Perplexity AI include the ability to rewrite answers using different AI models, view the sources used to generate answers, customize searches with different focus modes, and organize information into shareable collections.

What are some of the advanced search focus modes in Perplexity AI?

Some of the advanced search focus modes in Perplexity AI include Academic, Writing, Wolfram Alpha, YouTube, and Reddit. Each mode tailors the search to specific types of information.

How can you stay organized when using Perplexity AI?

Perplexity AI has a Library feature where users can create collections to organize relevant threads of information. Collections can be customized with descriptions, images, privacy settings, and AI prompts.

What file types can Perplexity AI read and analyze?

Perplexity AI can read and analyze images, CSV files, and other documents. It can describe images, summarize CSV file data for insights, generate images, and more.

How does the Playground feature in Perplexity AI work?

The Playground allows users to have conversations with different AI models from Perplexity and other providers. It shows how different models can provide varied responses.

What are some key elements of crafting a good prompt for Perplexity AI?

Key elements include blending background context and goals into one sentence, using relevant keywords to guide the AI, formatting the prompt clearly, and indicating the user's role where applicable.

What are some ideal use cases for Perplexity AI?

Ideal use cases include academic research, market analysis, data analysis, writing assistance, coding, and other information-intensive professional tasks that require accurate answers.

How can Perplexity AI save users time compared to other search tools?

By providing direct and accurate answers instead of links to search through, the ability to rewrite and refine answers, and advanced organization features, Perplexity AI can save hours of searching.

What future capabilities could Perplexity AI add?

Possible future capabilities include integration with other workplace apps, enhanced conversational abilities, voice search, and more customizable AI model options.

Outlines

00:00

😲 Introduction to Perplexity AI

This paragraph introduces Perplexity AI as an answer engine that uses large language models to provide accurate answers to complex questions. It outlines that the tutorial will demonstrate how to use Perplexity Pro to get the most value out of the tool, especially for professional/academic use cases.

05:02

📚 In-Depth Feature Overview

This paragraph provides an in-depth overview of Perplexity's key features, including the sources, images, videos, share button, rewrite models, view sources, copy answer, edit query, and related questions functionality. It also introduces the search focus modes and the library/collections for organization.

10:03

🔎 Diving into the Search Focus Modes

This paragraph analyzes each of the search focus modes (All, Academic, Writing, Wolfram Alpha, YouTube, Reddit) through relevant examples, highlighting the differentiated value offered by each one. It demonstrates how Copilot guides the search process and how the modes cater to specific use cases.

15:06

🗂️ Organizing Research with Library & Collections

This paragraph focuses on the Library page and Collections feature for organizing threads of research. It provides examples of creating a collection with AI instructions, adding threads, asking questions, and generating images. It also covers adding threads to collections and reusing collections for related projects.

20:08

📷 Image and File Analysis Capabilities

This paragraph demonstrates Perplexity's ability to analyze images and files, including describing images in depth and summarizing CSV data for trends/patterns. It aims to showcase the AI's comprehensive understanding and insight generation from various media beyond just text.

⚙️ Customization Options for Personalization

This final paragraph touches on customization options, including default AI models, user profile instructions, and the Playground for conversational practice. It ends by linking to Perplexity's blog on prompt best practices to help users frame questions optimally.

Keywords

💡answer engine

Perplexity AI refers to itself as an 'answer engine' that uses large language models to provide accurate answers to complex questions. This means it focuses on delivering comprehensive, factual responses to queries rather than open-ended discussions like a chatbot.

💡sources

Perplexity shows the sources it used to generate an answer. Clicking on a source link allows the user to view additional reference information to verify facts or learn more in-depth. Using credible sources improves answer accuracy.

💡rewrite

The 'rewrite' feature allows selecting a different AI model to re-generate the answer to a question with an alternate interpretation, formatting or style. This showcases how various large language models have their own unique way of responding.

💡view sources

The 'view sources' option shows all references used for an answer. The user can exclude unwanted sources to remake the answer without them. Controlling information sources improves answer relevancy.

💡related

The related section suggests follow-up questions to explore tangents from the original question. It aids deeper investigation by chaining logical inquiries in a gradual, focused manner.

💡collections

Collections help organize search threads into custom groups. The AI prompt feature allows defining special instructions for entire collections to standardize answers. Staying organized this way improves efficiency.

💡generate image

The 'generate image' tool creates artwork representing elements from an answer thread. While basic now, it adds helpful visual context and will likely improve over time as AI artistry advances further.

💡read files

Perplexity can ingest files, summarize data in them and extract insights. This leverages its analytical capabilities beyond just Q&A to automate tasks like data processing. Compatibility with external assets boosts utility.

💡focus modes

The specialized search focus modes (Academic, Writing, YouTube etc.) filter results to credible or user-specified sources. Choosing the right focus for goals maximizes relevance and minimizes noise.

💡prompting tips

Prompting tips guide effective query formulation based on intended goals. A well-structured prompt with relevant keywords channels the query for optimal answers. Following prompt best practices improves outcomes.

Highlights

Perplexity AI delivers accurate answers to complex questions using large language models

Perplexity Pro is best suited for professionals and academics who need in-depth, specialized information

Perplexity creates shareable threads for each search query that get saved to your personal library

You can rewrite an answer using different AI models like GPT-4, Claude 2.1, or Gemini Pro

The academic search focuses on scholarly articles and publications to provide specialized, in-depth information

The writing mode generates original text and code without accessing outside information

Wolfram Alpha focus solves complex mathematical and computational problems

YouTube focus delivers targeted video results to complement the written answers

Reddit focus summarizes discussions and opinions from Reddit posts

Collections help organize your library into shareable folders based on topics or projects

The AI prompt gives custom instructions for all threads in a collection

Perplexity accurately describes images and summarizes datasets

You can set a default AI model to use for all your searches

The profile section allows you to give Perplexity custom instructions to tailor its answers

Use keywords, background context, and clear formatting in your prompts to get better answers

Transcripts

Perplexity AI is on a mission to redefine how you get accurate

and reliable answers whenever you're searching for information.

Perplexity AI dubs itself as an answer engine

that uses large language models to deliver accurate answers

to complex questions.

And in today's comprehensive tutorial, I'm going to show you everything

you need to know to make use of this pretty exciting

and useful AI tool.

Prepare to level up and let's get right into it.

Head over to perplexity.ai.

Just a note for this tutorial, I'm going to be using Perplexity Pro.

As from testing out the tool, it became clear that the people

that will benefit most from Perplexity would likely be doing so

in a professional or academic setting, as we'll uncover throughout the video.

On the homepage, you'll find a collapsible nav bar

that contains additional links to the Discover page and your library,

all features I'll be going through today.

But for now, let's learn by doing and ask perplexity a question.

When you click into the search box you'll be greeted

with suggested searches that sometimes make for fun browsing

on the internet and these are updated daily.

But for now let's ask it how x-rays work.

And just like that you're greeted with a pretty comprehensive

and seemingly factual answer.

Despite the simple interface, there is a lot going on on this page.

And let's take some time now to dive into the many features

that really make Perplexity shine as an answer engine.

At the very top, you have the sources Perplexity

used to give you back your answer.

And if you'd like even more in-depth reading,

you can actually click into a source and read additional information there.

To the right of our answer, you notice it also gave us images

that help answer our question.

And there's also a search for videos feature

and a generate image feature, which we'll take a look at further on

in the video.

Now, beneath our answer actually lies some pretty powerful features.

First, clicking the share button copies the link to this exact same thread.

What are threads you may ask?

Every time you run a search in perplexity it creates a new thread

which then gets added to your library as you can see on the left.

I'll dive more into library in the next little bit.

Supposing you didn't like the answer it first gave you,

you can actually just simply click rewrite and then select a different model.

This is a truly powerful feature as you're able to select models such

as GPT-4 used in the latest chat GPT, Claude 2.1 by Anthropic or Gemini Pro

which was recently released by Google and still in beta.

So let's rewrite this answer using Claude 2.1

and see what we get.

And just like that we have a different answer.

So even though the facts about x-rays may still be the same

across models, each model has its own way of interpreting your question,

answering it, and formatting it in a way it thinks makes

the most sense to the user.

And just a quick note, to the right of your answer,

you will see what model is being used.

Next is a seemingly simple but pretty powerful button,

the view sources button.

When you click view sources, you once again are able to

see all the sources referenced to give you your answer.

However, in this case, supposing I didn't want YouTube

to be a part of the knowledge source use, I can simply delete YouTube from

my sources, click done, and then perplexity will regenerate

my answer, omitting the YouTube source.

Pretty incredible.

And moving on from sources, next is the copy button to simply copy

your answer to the clipboard.

And lastly, the edit query button, which simply lets you rewrite

your original ask.

I'll cancel out of that for now and scroll all the way to the bottom of

our answer to the related section.

As I mentioned earlier, the related section really is a fun place

to just go down rabbit holes of knowledge.

For example, you can then ask Perplexity what

are the different types of x-rays.

After which it goes on to give a good and well formatted answer

on the several types of x-rays.

Okay, that was a basic overview of Perplexity AI and now let's dive deep

into the search focus features to really start to unleash the power

of Perplexity AI and see where it shines against tools such as Google search

and chat GPT Back on the home page if you click focus You will notice there's

six different focus modes ranging from all to academic to writing,

Wolfram Alpha, YouTube and Reddit.

I am going to take the time to quickly go through each one of these

as each really comes with its own set of unique differentiators.

And it's where perplexity seems to provide the most value for your money.

I'm also going to activate Copilot, which essentially helps

you guide perplexity to the exact answers you're looking for,

And we'll get to see how it works throughout these examples.

So our first search about X-rays was using the all focus mode,

as that is the default mode.

In the all mode, Perplexity searches the internet

to provide you the best possible answer.

But now let's home into the academic focus mode,

which is a specialized search within scholarly articles

and academic publications.

I'll ask it, what are the most recent advancements

in artificial intelligence and finance and then hit enter.

And then Copilot steps in to guide us towards a more refined answer.

Of course, I can ask it to be broad in the optional text field.

But for this case, I'll tell it to zoom in on risk management.

And then it goes on to give us a pretty comprehensive answer

on how AI has significantly impacted the field of risk management

in finance,

covering areas such as AI integration in risk management processes,

explainable machine learning for risk management,

and so forth.

And you'll notice it even throws in a cautionary paragraph at the end,

letting us know that these advancements have the potential

to revolutionize risk management and finance, but they also come

with challenges such as data privacy concerns,

regulatory compliance and ethical considerations.

Really cool stuff.

Okay next let's test out the writing mode.

The writing mode enables you to have a conversation

and perform tasks with perplexity without it accessing the internet.

This is particularly useful when you're trying to generate code or pieces

of text.

For this example though, let's ask it to generate some code.

I'm going to ask it to write me a simple program in Python

that counts the letters in someone's name.

Copilot then steps in to ask us who'll be providing the name, us

or the program itself.

I'm gonna say it's gonna be user input, then hit send.

And in the blink of an eye, you have a pretty simple

but handy Python script that can prompt a user for their name

and then counts the number of letters in their name.

You also notice that it took the liberty to write the program

in such a way that ignores any spaces in the user's name,

so it'll correctly count the letters in both single word names

and multi word names.

Very handy indeed.

Okay, next is Wolfram Alpha.

The search focus mode is designed for computational data

and mathematical calculations.

I'm going to paste in an example I found from Perplexity AI.

And just like that, it provides us an answer

to help us solve the equation.

Moving on, let's try the YouTube search focus.

This focus mode allows for targeted searches within YouTube,

offering time steps and the options to watch videos

in the platform.

So, for example, we can ask Perplexity how to build a beginner gaming computer.

Let's be honest with prices these days, it's going to be over a thousand dollars.

Next, it asks us what specific games or game genres we want to play on

our computer.

I'm going to choose strategy games and simulation games and then hit send.

Before we dive into the answer, I'll take a moment to expand

the Copilot box just to give you a better sense

of what it's doing in the background.

As you can see, Copilot took our inputs and then used those

as keywords when it was searching YouTube for the best possible answer,

after which it then found 24 different sources to help provide

our answer.

I can't even imagine how long it would have taken to go

through each and every one of these videos

and yet here we are just 10 seconds later.

And I'm no computer building wiz but reading

through the answer it genuinely does look like it knows what it's talking about.

And supposing you need more context, you can simply click on any

of the little number icons, which will take you to the video

it referenced to provide the answer.

Really awesome stuff. Okay, let's quickly look

at the last focus mode, Reddit.

This mode focuses on Reddit discussions and opinions, which, of course,

there are plenty on Reddit, but it does help provide

a streamlined experience that filters out irrelevant content.

So, for example, supposing I'm researching reviews

on the iPhone 15 Pro, I can simply enter reviews

on the iPhone 15 Pro in the Reddit focus mode.

And for this one, I'm actually curious about everything.

So let's select it all.

And Perplexity then scours Reddit to provide the consensus answer

on the camera, battery life, performance, design and price of the iPhone 15 Pro.

Really handy.

Okay, so after that deep dive into the search focus modes,

you are now well equipped to use perplexity for

whatever task you have at hand.

Next, we're going to take a deep dive into the library

and more specifically collections to see how you can easily stay organized

in perplexity AI.

Before we get there, though, I will make a quick detour

on the Discover page.

This page is so much fun as it is updated pretty frequently

and allows you to discover search topics that are really,

really interesting to dig into.

And I would encourage you to use the Discover page

as it does help you get more familiar and comfortable with using Perplexity AI.

Okay, on the library page is where we'll find

all those threads we just created.

You can imagine after using this tool for a bit you may have hundreds

of threads.

So how do you stay organized?

Well, first things first, you can actually easily click

the three little dots and delete all threads.

Not a bad goal when you just need a clean start sometimes.

Clicking the plus icon actually allows you to start a new thread just

as you would on the home page.

Now what things get really powerful is the collections feature.

So let's click get started.

I'm going to make a new collection for meal prep

and I'm going to give it a food icon.

The description is optional but I'm going to enter meal ideas

for the week.

Now this AI prompt section while optional is very very handy

as it allows you to give instructions to perplexity

that affect every thread in that specific collection.

So think of this as custom instructions if you're more familiar with ChatGPT.

So I'm going to let Perplexity know that it's a dietitian

that's going to help me plan quick and easy to make nutritious meals.

Before hitting create, you can also decide the privacy level of

your collection, either by making it shareable

so you can easily share a link with friends and family or

by making it secret.

I'll leave it shareable for now and then click create.

And just like that, my collection is now on the left.

So I'm going to go ahead and click into it.

And coming into the collection, you can already see

that it has some suggested questions for me,

which are actually pretty handy.

So let's take the first question for a spin.

What are some creative and nutritious meal prep ideas

for breakfast, lunch and dinner?

And pretty quick, it gives us some seemingly easy

to make breakfast, lunch and dinner options.

Very, very useful.

And even at the end, it goes on to say these meal prep ideas

are not only nutritious, but also convenient for those

with busy schedules.

Now I'll go back to my meal prep collection

and give it a more complicated question and see what it comes up with.

I'll ask it to create a recipe that has 40 grams of protein

and five grams of fiber using chicken and vegetables and also to make it spicy.

It then goes on to give me a recipe with ingredient list

and the instructions to hit my desired macronutrient levels.

And as you can see at the bottom, it gives us a nutritional

information summary of the recipe.

Now, for this recipe, let's try generate an image for it

and see what comes up. So I'll hit generate image

and then choose the photograph style.

Wow, this is not a bad image at all.

And if you squint your eyes hard enough, in the bottom right corner,

you will see the little AI emblem noting that this image was generated by AI.

Just a heads up, there are a couple of times where I try to generate images

in here and they don't really turn up.

So your mileage may vary with this feature.

Also, just a note, Perplexity's image generator

really isn't meant to be used like tools such as DALLE and Mid Journey

and is more so meant for situations like this where you'd like

to add a bit more context to your answer.

Okay, heading back to our library, supposing I wanted to create

a new collection to store some of the other threads we'd done today.

Well, that's very easy to do.

Once again, we just hit the plus sign next to collections and then enter our details.

You can simply click the three dots on the bottom right of any thread,

click add to collection, and then select the collection you'd

like to add it to.

At the time of recording, threads can only be in one collection

at a time.

Okay, moving on. Perplexity does a great job

of reading files and images.

So let's go through a couple examples to show you what I mean.

We'll start by uploading this picture and then asking perplexity

AI to describe the image.

And this is pretty wild.

It correctly identifies that there are four individuals gathered

around a wooden table in the picture.

It also lets us know that there's a man wearing a beanie

and glasses sitting in the middle.

And goes on to further describe the features

of everyone else in the picture.

It then goes to let us know what's on the table, such as a red cup,

a blue cup, a glass jar with a spoon and what seems to be a sugar container.

Absolutely wild.

Now, just for a bit of fun, I am curious to see what happens when

you click generate image from an image.

So let's try that and choose the photograph

option once again.

Okay, So this is both hilarious and absolutely creepy.

It made a picture of four people sitting on a table,

but they do seem to be a bit possessed.

Anyway, moving on.

Let's upload a CSV file to Perplexity and ask it to summarize it

for any insights.

Now, the CSV file contains a ton of hotel data so the copilot steps in

and asks us what it is exactly that we'd like summarized.

I'll tell it to summarize booking patterns and general trends.

It then proceeds to give us a pretty comprehensive overview of

our uploaded file.

For example, column six lets us know that most bookings are for one

or two adults with children and babies being less common.

Now, of course, this is just the tip of the iceberg.

And the purpose of this really is to show you the capabilities

and not run a comprehensive data analysis.

So now at this point, you should be pretty well equipped

to do some pretty incredible stuff in Perplexity AI and more specifically,

Perplexity Pro if you have it.

With that said, let's head over to our profile settings to go

through some advanced but pretty handy features.

On your account page, you can actually select the

AI model you would like Perplexity to use as a default.

So the default model is of course the one provided by Perplexity,

which is pretty fast.

However, if you have a preference such as GPT-4

or CLAR 2.1 or even Gemini, you can easily select it

from this screen here.

And all your future searches will be using that model by default.

The next handy feature here is if you switch to your profile tab,

you can also give it custom instructions that the AI should know.

And once again, if you use chat GPT, this is akin to

the custom instructions there.

As an example, I'll let it know that I'm someone

that prefers brief and well-structured answers

in a bullet point format.

And you know what? I will enter my location

as it suggests a location, Boulder, Colorado, and then hit save.

When you do that, follow up questions are then brought

into the mix.

For example, it asks you about your professional skills,

your favorite outdoor activities and stuff like what kind of music you like.

These questions are designed so that Perplexity

can get a better understanding of the type of answers it thinks you'd

like to see the most.

However, you don't have to fill out any of this information

and you can easily skip it.

Just a Note, if you do enter anything here,

it does affect all your new threads.

So sometimes I'd say it's better to use collections

versus this global setting, but your mileage may vary.

Let's quickly take our new settings for a test drive And I'll ask it

to describe how satellites work.

And as you can see, for such a pretty broad topic,

such as how satellites work, it gave me a pretty brief

and well-structured answer in bullet point format,

exactly what I asked for.

And to turn off the custom instructions, you can simply go back to your profile

and then hit clear.

Or you can also hit pause at the top of the page.

Before we touch on some prompting tips, I'm going to take just

a quick minute to head over to the Playground.

This is pretty much just a fun place where you can have a conversation

with different AI models, both from Perplexity

and other providers.

So, for example, using Perplexity's default 7B

online model, I'm going to ask it to tell me

about the matrix.

And just like that, it gives us an answer.

But supposing I prefer to use another model.

Let's pick LLAMA2 and ask the same question

and see what sort of answer we get there.

As you can see, LLAMA provides a more essay type response,

whereas perplexity provided a more straight to the point and business

like answer.

Really, really cool.

Okay, let's look at some prompting tips. For this section, I'm actually not looking

to reinvent the wheel. And perplexity actually recently

released a very handy blog on some prompting tips.

So we'll just take a moment to walk you through that.

First of all, the anatomy of a prompt.

The prompt should seamlessly blend your goal and background

into an easy to read sentence.

It should contain keywords which better guide perplexity

to give you the right answer. And the formatting is an important aspect

to give you good results.

So, for example, supposing you're using perplexity

for market research, you can let it know that you're a product manager

and need a summary of market trends for electric bikes in the US.

And in this case, the keywords Perplexity picks up most

on are market trends, electric bikes, the US and product manager.

They then have a few other examples on the page.

So, for example, supposing you're a student writing an essay,

you can then let it know that you're a student aiming to excel in

your psychology paper and need an essay outline on the effects

of social media on mental health with your keywords

in that ask being essay outline, social media, mental health and student.

Very, very handy resource.

If you'd like to read this more in depth, I will link it down below as

it's a great resource to help you structure

your prompts depending on your goal.

Awesome job making it this far in.

At this point, you should now be able to save hours of

your life from trudging through horrible search results on Google

and being able to use Perplexity to get the answers you need in a fraction

of the time.

In terms of some key takeaways, Perplexity is an answer engine

and not a search engine or even chat bot. So you kind of do have to approach this

a bit differently than you'd approach Google

or even chat GPT.

Perplexity seems to be more accurate with academic and factual information

and is a great resource for those type of settings.

When using this AI tool, make sure you choose

the right focus mode to really increase your efficiencies.

Also, use collections to stay efficient and organized.

I'd love to hear how you've been using Perplexity AI

or Perplexity Pro down in the comments below.

Thanks for watching and catch you in the next one