0:01 I took Google's prompt engineering
0:03 course for you so here's the cliffnotes
0:05 version to save you to 9 hours but it's
0:07 not enough just to listen to me talk
0:09 about stuff so I've also included a
0:10 little assessment at the end of the
0:12 video to help you remember everything
0:14 that you learned because research shows
0:15 that immediately reviewing information
0:17 after you learn it is the best way of
0:19 retaining that information all right
0:21 let's go let's first go over the
0:22 structure of this course prompting
0:25 Essentials has four modules module one
0:26 is start writing prompts like a pro this
0:27 is where they introduce some really
0:29 helpful Frameworks for how to craft
0:31 prompts module two is design proms for
0:33 everyday work task this will include
0:35 proms for emailing brainstorming
0:36 building tables and summarizing
0:39 documents module 3 specifically focused
0:41 on using AI for data analysis and for
0:43 building PowerPoint presentations and
0:45 finally module four use AI as a creative
0:47 or expert partner this is where Google
0:49 really packs it in like I am genuinely
0:51 super impressed by this module we talk
0:52 about Advanced prompting techniques like
0:54 prompt chaining Chain of Thought tree of
0:56 thought and a framework for creating
0:58 agents all right module one let's do the
1:00 fundamentals let's first the prompting
1:02 prompting is the process of providing
1:04 Specific Instructions through gen tool
1:06 to receive new information or to achieve
1:08 a desired outcome on a task this could
1:11 be text images video sound or even code
1:13 the course provides a five-step
1:15 framework for how to design a prompt
1:18 task context references evaluate and
1:20 iterate the task is what you want the AI
1:22 to do for example if your friend's
1:23 birthday is coming up and they're really
1:25 into anime you can say suggest the gift
1:27 related to anime for my friend's
1:28 birthday now that prompt in itself is
1:31 okay but but you can Elevate this and
1:32 get a result that's more unique and
1:35 specific by incorporating two additional
1:37 things the first one is a Persona which
1:39 is a role that you want the AI to embody
1:41 for example you can update the prompt to
1:43 act as an anime expert to suggest an
1:45 anime gift for my friend's birthday you
1:46 notice that the results are a lot more
1:48 specific and it's actually split into
1:50 different genres the second thing you
1:52 can add is the format of the output the
1:53 default here is just the list and bullet
1:55 points but maybe you want something
1:57 that's more structured so you can say
1:59 organize that data into a table the
2:01 second component of the framework is
2:03 context the general rule of thumb is
2:05 that the more context you can provide
2:07 the better the output will be in your
2:09 birthday gift example you can specify
2:11 something like your friend is turning 29
2:13 years old her favorite animes are
2:15 shangula Frontier solo leveling and
2:17 Naruto etc etc you'll see that the
2:19 output is much more targeted third part
2:21 of the framework is references this is
2:23 where you can provide examples to the AI
2:24 sometimes when you're trying to explain
2:26 what you want it's kind of hard to
2:28 describe it in words but providing
2:30 examples can really clear things up and
2:31 AI is especially good at incorporating
2:33 examples maybe you can provide past
2:35 birthday presents that this person has
2:37 enjoyed step number four is evaluate
2:39 this is after you get the output just
2:41 ask yourself is this output what I want
2:42 it to be and if it's not exactly what
2:44 you wanted to be then last upep is
2:47 iterate prompting is rarely a oneandone
2:48 kind of thing it's much more of a
2:49 circular process in which you're
2:51 refining the prompt to get the results
2:53 that you want just like what we did
2:54 earlier often times you might just start
2:56 with a simple task like suggestions for
2:58 a birthday present then you want to get
3:00 better results and you start iterating
3:02 on that and adding things like a Persona
3:04 context and references to finally get to
3:05 a result that you're happy with as the
3:08 course calls it ABI always be iterating
3:10 speaking of pneumonics the course does
3:12 have one for this five-step framework
3:14 which I actually find really difficult
3:16 to remember um I think it's thoughtfully
3:19 craft really excellent I don't know what
3:20 the ey stands for I'll I'll put it on
3:22 screen but I do have one that I made
3:24 which I can remember a lot better so I
3:25 don't know maybe this will help you as
3:27 well uh which is Tiny crabs ride
3:30 enormous iguanas a lot more memorable in
3:32 my opinion anyways whatever it is that
3:33 you need to do uh just figure out some
3:35 way to remember this framework because
3:37 everything else in the course is based
3:40 on this the rest of module one which
3:41 also includes interviewing different
3:44 people I think is interesting but not
3:46 super necessary the only other really
3:47 useful important thing that they
3:49 presented is the four iteration methods
3:51 so by following date I'm just going to
3:53 call it tiny crabs writing enormous
3:55 iguanas framework um will get you like
3:57 80% of the weight but sometimes you're
3:59 just not quite there so to iterate and
4:01 get the 20% there are four different
4:03 methods that you can try the first one
4:04 is just to revisit the prompting
4:06 framework maybe you can give more
4:08 references more examples provide more
4:10 context or you know add a Persona if you
4:12 haven't already number two is to
4:14 separate your prompt into shorter
4:16 sentences it's helpful to think about AI
4:17 like how you would talk to a normal
4:19 person like if you just word vomit to
4:21 someone about whatever it is that you
4:22 want they'll probably be like
4:24 overwhelmed and there's just like a lot
4:25 of stuff going on right so the same
4:27 thing can happen for AI and an easy
4:29 solution for this is just to break your
4:31 prompt into simpler sentences and
4:33 feeding it to the AI slowly so it's less
4:35 like blah blah blah blah blah blah blah
4:38 I'm more like blah blah and blah much
4:40 more organized number three is trying
4:42 different phrasing or switching to an
4:44 analogous task say you want the AI to
4:46 help you write a marketing plan but the
4:48 results are just kind of like boring and
4:50 Bland what you can do is that marketing
4:52 is really just telling a compelling
4:54 story so instead you can ask it to write
4:56 a story about how this product fits into
4:58 the lives of our Target customer
5:00 demographics this is an analog task but
5:02 the results are much more lively and
5:04 interesting and the fourth iteration
5:06 method is to introduce constraints just
5:07 like when you tell someone that they can
5:09 do anything or like if you ask people
5:11 what does everybody want to eat for
5:13 lunch and they're just like oh anything
5:14 this actually makes it a lot harder for
5:17 you to get a result that you're happy
5:18 with so instead you can introduce
5:20 constraints to narrow the focus down say
5:22 you want to generate a playlist for a
5:24 road trip and the AI generat your
5:26 playlist but it's just like not very
5:27 interesting you can add different
5:29 constraints like only specific to a
5:31 certain region only at this specific
5:41 heartbreak Deion for example I don't
5:43 know maybe you like feeling sad so with
5:45 these four iteration methods um with the
5:47 help of AI I also came up with a
5:49 pneumonic to remember it better which is
5:53 rahen saves tragic idiots so let's talk
5:55 about multimodo prompting the most
5:56 classical way of interacting with a
5:58 large language model is just by like
5:59 typing stuff I like having a
6:01 conversation but you can actually
6:04 interact with many AI models like Gemini
6:06 um with different modalities as well
6:08 including pictures audio video and even
6:10 code it's able to take different types
6:12 of modalities as the input and is able
6:14 to Output using different modalities as
6:16 well this doesn't change anything in
6:17 terms of how you think about prompt it's
6:20 still going to be tiny crabs writing
6:22 enormous iguanas but you just might need
6:24 to be a little bit more careful about
6:26 specifying what kind of input or output
6:27 you're looking for and the kind of
6:29 context that you're providing for
6:30 example if you designed a new nail
6:32 artart collection and you want to Market
6:33 it on social media you could input
6:35 something like write a social media post
6:36 featuring this image and then attach
6:39 your nail art collection as a reference
6:40 the post should be fun short and focus
6:42 on the fact that's a collection of new
6:44 designs I'm selling some other examples
6:46 of multimodality usage would be asking a
6:48 gen tool to suggest recipes based on the
6:50 photo of the ingredients in your fridge
6:52 inputting your Brand's logos and colors
6:54 and then creating a digital teaser to
6:56 promote an event or if you're working on
6:57 a short story and you get really
6:59 inspired by a musical piece you try
7:01 inputting that music piece and tell it
7:03 to kind of follow those vibes for the
7:05 atmosphere and details of the story
7:07 regardless of the modality that you're
7:09 prompting in there are two major issues
7:12 with using AI tools the first one is
7:14 hallucinations a Hallucination is when a
7:16 gen AI tool provides outputs that are
7:17 inconsistent incorrect or even
7:19 nonsensical a really famous example is
7:21 that if you ask an Ella um how many RS
7:23 are in Strawberry it tells you that
7:24 there's two hours in Strawberry the
7:28 second is biases unfortunately L's being
7:31 trained on human content also
7:33 incorporates human biases things like
7:35 gender and Race So to minimize these
7:36 sorts of problems the course recommends
7:38 that we take a human in the loop
7:39 approach which is making sure that
7:41 you're always checking your outputs and
7:43 verifying whatever it is that the Gen
7:45 tool gives you in the end it is your
7:46 responsibility of making sure that
7:49 whatever is being produced is in fact
7:51 accurate here is a checklist feel free
7:53 just to take a screenshot for some
7:55 considerations when you're thinking
7:57 about using AI responsibly you know
7:58 compared to other Google courses I've
8:00 taken especially the AI Essentials
8:02 course which you can check out over here
8:04 um this course is a lot more dense which
8:06 is a lot better bang for your bug so pay
8:10 attention moving on to module 2 module 2
8:11 is called design prompts for everyday
8:13 work task it's essentially just
8:15 providing examples of use cases based on
8:17 the tiny crabs writing enormous iguanas
8:19 framework and the ramen saves tragic
8:21 idiots framework too that's why I'm
8:22 going to go through this module
8:24 relatively quickly I'll highlight some
8:25 of the examples that I think are really
8:28 important and for the rest of them I'll
8:30 actually just put on screen so you can
8:32 take screenshots of it if you want and
8:33 build out your own prompt Library where
8:35 you can kind of like store the prompts
8:36 that you want to use one of the biggest
8:38 use case that most people have when
8:40 using gen tools is by using it to
8:42 produce content for example like writing
8:44 emails here's an example of a situation
8:46 um when you want to write an email to
8:48 your staff about a new schedule change
8:50 for your gym I'm a gym manager and we
8:52 have a new gym schedule write an email
8:54 informing our staff the new schedule
8:56 highlight the fact that the MWF Monday
8:58 Wednesday Friday Cardio Blast class
9:00 changed from 7:00 a.m. to 6:00 a.m. make
9:02 the email professional and friendly and
9:04 short so that the reader can skim it
9:05 quickly here's the new schedule and you
9:07 can actually attach the link that
9:08 contains the new schedule this sort of
9:09 email probably take you like 10 minutes
9:11 to write but by using a gen tool you can
9:13 do it in like a minute most of us do
9:15 send quite a lot of little emails here
9:17 and there throughout the week um so the
9:19 time savings do add up for this kind of
9:21 email you probably aren't super picky
9:22 but what happens if you need to write
9:25 email that is a lot more important or
9:26 you're writing other things like an
9:28 essay an article or a newsletter you
9:30 would care a lot more about the tone and
9:32 the word choice that's being used
9:34 instead of using general terms like
9:36 write a casual summary try to use more
9:38 specific phrases like write a summary in
9:41 a friendly easy to understand tone like
9:42 explaining to a curious friend you can
9:45 also provide references for context
9:47 other emails or articles or whatever
9:48 that you've written in the past and tell
9:50 to AI to match the tone I'm going to now
9:52 include a few other prompts on screen
9:55 related to generating text or content
9:56 which you can take a screenshot to add
9:58 to your prompt Library if you want if
10:00 you're current in University thinking
10:02 about going to University or going back
10:04 to University to maybe get an additional
10:06 degree you should check out straighter
10:08 line straighter line is a credible
10:10 online education platform where you can
10:12 take highquality online courses designed
10:14 by academics from leading universities
10:16 and recommended by accredited
10:18 educational institutions you can
10:19 basically take a course at a fraction of
10:21 the cost and still transfer the credit
10:23 to your college or university more than
10:25 3,000 colleges and universities have
10:27 accepted credit from straighterline
10:29 courses so not only are StraightLine
10:30 courses is much more affordable they
10:32 also come with a flexibility of being
10:33 able to take a class when you want and
10:35 at your own pace they even have a
10:37 variety of Support Services like tutors
10:39 and resources to help you succeed if
10:40 you're still on the fence just check out
10:42 their course catalog and you can start a
10:44 free trial today at this link over here
10:45 also linked in description thank you
10:47 straighter line for sponsoring this
10:49 portion of the video now back to the
10:50 video module 3 is pretty similar to
10:52 module 2 it's just more example use
10:54 cases very helpful example use cases
10:56 though specifically data analysis and
10:58 presentations the big word of caution
11:00 here is is that be careful about what
11:02 data that you're inputting into your AI
11:04 model uh if you're working for you know
11:06 a company you do not want to be
11:08 violating any privacy policies and you
11:10 probably don't want to be putting some
11:12 sensitive data into some large language
11:14 model as well the example from a course
11:16 if save you have a data set for a
11:17 grocery store chain with information
11:20 about the store the area items that they
11:22 have available the daily customer
11:24 account and store sales if you're not
11:26 very good at Google Sheets or Excel you
11:28 might input a prompt like this attached
11:30 is a Google sheet of store data how can
11:32 I create a new column in sheets that
11:34 calculates the average sales per
11:35 customer for each store it can help
11:37 teach you specific things like this um
11:39 and it can actually do more say if
11:40 you're interested in the trends in the
11:42 data set you could then add on to give
11:44 me insights into the relationship
11:46 between daily customer count items
11:48 available and sales based on the given
11:50 data Gemini is able to show some
11:51 interesting Trends including the fact
11:53 that there's no clear correlation
11:55 between items available and store sales
11:56 if you find this interesting you can
11:58 continue prompting it digging into this
12:00 and maybe coming coming up with ways to
12:01 figure out why that's the case I'm going
12:03 to put on screen now a couple other
12:05 prompts related to spreadsheets and data
12:07 analysis that you might find helpful the
12:08 second part of the module is about
12:10 building presentations and I'm going to
12:11 put on screen out a couple prompts
12:14 related to presentations that could be
12:17 helpful and finally we are at module 4
12:19 we're almost done guys module 4 is
12:22 titled use AI as a creative or expert
12:24 partner this is an extremely important
12:26 module and what made me very impressed
12:28 about this course so first we're going
12:29 to cover some advanced prompting
12:31 techniques starting off with prompt
12:33 chaining prompt chaining is a way to
12:35 guide gen tools through a series of
12:37 interconnected prompts adding new layers
12:39 of complexity along the way for example
12:41 you're an author and you wrote a
12:42 wonderful novel and now you want to
12:45 Market and sell your novel so you need
12:47 to come up with a marketing plan the
12:48 course recommends you use Google AI
12:49 studio for this because it has much
12:51 longer context window because you're
12:53 going to be attaching your entire
12:54 manuscript the first thing you might
12:56 want to do is to generate some summaries
12:58 of your manuscript generate three
12:59 options for a one sentence summary of
13:01 this novel manuscript the summary should
13:03 be similar in voice and tone to the
13:05 manuscript but more catchy and engaging
13:07 so Jeb was able to give some decent
13:09 options but you want to focus on a more
13:11 specific theme that's where the prompt
13:13 chaining comes in taking the output from
13:15 the previous prompt and then asking
13:17 create a tagline that is a combination
13:19 of the previous three options with a
13:21 special focus on the exciting plot twist
13:23 and mystery of the book find the
13:26 catchiest and most impactful combination
13:27 the tagline should be concise and leave
13:29 the reader hooked and wanting to read
13:31 more and great it comes up with the
13:33 desert whisper secret a young Weaver
13:36 seeks a city of singing stones but the
13:38 greatest Journey unfolds into Whispers
13:39 of her own
13:42 heart anyways you can keep refining
13:43 things if you want and finally maybe
13:46 even asking Gemini to generate a six-
13:48 week promotional plan for the book tour
13:51 including the locations and the channels
13:52 to promote each stop on the tour so that
13:54 was prom chainy there's two other
13:56 Advanced Techniques in this module Chain
13:58 of Thought prompting and tree of thought
13:59 prompting as a no a lot of these Ai
14:01 terminologies and techniques sound like
14:04 super fancy but they're actually not
14:06 like Chain of Thought prompting is about
14:09 asking the AI to explain his reasoning
14:11 as a step-by-step process it's similar
14:12 to how your math teacher might ask you
14:14 to explain your work so he or she is
14:16 able to identify like the steps that
14:17 you're taking and where you could be
14:18 going wrong all you have to do
14:20 throughout your prompting sequence is to
14:21 tag on the line explain your thought
14:23 process this helps you understand the
14:25 ai's reasoning for things and you can
14:27 help improve his decision making tree of
14:30 thought prompting as his name suggest is
14:31 sort of like a tree it allows you to
14:33 explore multiple reasoning paths where
14:35 like branches simultaneously this is
14:37 really helpful for abstract or complex
14:40 problems like developing novel plots
14:42 with new characters or creating outlines
14:44 and drafting sections for lengthy
14:46 documents you can work with the AI tool
14:48 to explore different options and
14:49 evaluate them to finally come up with
14:51 the best output as an example maybe
14:53 you're creating an online course and you
14:55 want to create a cool image on the
14:56 landing page you can use tree of thought
14:58 prompting to brainstorm different
15:00 options a potential prompt may be
15:02 imagine three different designers are
15:04 pitching their designs to me all
15:05 designers will write down one step of
15:07 their thinking then share it with the
15:09 group then all experts Will Go On To The
15:12 Next Step Etc if any expert realizes
15:13 they're wrong at any point then they
15:15 leave the question is generate an image
15:17 that's visually energetic and features
15:20 images of art supplies in computers show
15:22 me three suggestions in very different
15:23 styles from simple to detailed and
15:26 complex and here's the output Gemini
15:27 came up with now looking at this output
15:29 you might be like m I kind of like The
15:32 Vibes of one where one could be going so
15:34 you can tell the AI that you like the
15:36 first one and you'll like to expand the
15:37 idea a little bit more and perhaps
15:39 generate three different color schemes
15:41 for that concept and you can just keep
15:42 doing that until you end up with
15:44 something that you like a pro tip is
15:46 that you can combine Chain of Thought
15:48 and tree of thought prompting by asking
15:50 the AI to explain his reasoning at each
15:52 iteration so you can provide feedback
15:53 another Pro tip as you're prompting
15:55 along is if you ever get stuck and you
15:58 don't really know what prompt to use you
16:00 can actually use use AI to help you come
16:02 up with a prompt this is called meta
16:04 prompting all right the last section of
16:06 the course is on agents and I have
16:07 actually not seen a single course be
16:10 able to cover agents as well as this one
16:12 so first definitions what is an AI agent
16:14 an AI agent is like an expert designed
16:16 to help with task and answer questions
16:18 you can have all sorts of Agents you can
16:19 have coding agents that help you with
16:21 coding uh marketing agents that come up
16:23 with marketing plans with you a golf
16:24 agent that can correct you on your golf
16:27 swings or maybe just a friend agent that
16:28 can be your friend the course cover
16:30 covers two types of Agents the first one
16:33 is a simulation agent called Agent Sim
16:35 agent SIM can simulate scenarios with
16:37 you like conduct interviews or do role
16:39 playing for example if you work in a HR
16:40 department you might be tasked with
16:42 coming up with a training program to
16:44 help interns improve their interviewing
16:46 skills for that final job assessment for
16:48 AI agents you want to focus a lot on the
16:50 Persona and the context the Persona here
16:53 is act as a career development training
16:56 Simulator the task is your task is to
16:58 help interns Master interview skills and
16:59 conduct convers ation with potential
17:02 managers then you have the context you
17:03 need to support the following types of
17:05 conversations articulating strengths and
17:07 skills communicating professionally and
17:09 confidently And discussing future career
17:11 development goals once an intern has
17:13 picked a conversation topic provide
17:15 details about the situation in the
17:16 interviewer's role then act as the
17:18 interviewer and allow the intern to
17:20 participate as the employee make sure to
17:22 guide a conversation in a way that will
17:23 allow the intern to exercise their
17:25 interview skills finally you want to
17:27 include a stop role where you can tell
17:29 the agent that you're done with the
17:30 simulation continue to roleplay until
17:33 the intern replies with jazz hands after
17:35 the intern gives the stop rule jazz
17:36 hands provide them with key takeaways
17:38 from the simulation and skills they can
17:40 work on now that is set up you can start
17:42 doing a simulation maybe by inputting
17:44 the chart analysis that I did for my
17:46 intern project agent Sim will ask you
17:47 more questions about the analysis and
17:49 you keep responding to them and at the
17:51 end you can insert jazz hands and then
17:53 agent Sim will provide feedback for you
17:55 the second kind of agent is an agent for
17:58 expert feedback Called Agent X Agent X
18:00 is able to give you feedback on any
18:02 topic of your choosing sort of like a
18:04 personalized tutor or consultant here's
18:06 an example prompt to create an agent X
18:08 that can provide you feedback about a
18:10 pitch for a potential client first the
18:12 Persona you're my potential client the
18:14 VP of advertising at a world famous
18:16 sports car company known for its
18:17 Innovation performance and Engineering
18:19 Excellence now the context you're
18:21 considering hiring a creative agency to
18:23 develop a new campaign that will attract
18:25 a younger generation of buyers you're in
18:27 a meeting with me the design director of
18:29 a creative agency that's pitching a new
18:31 campaign for your company and now the
18:33 task act as my potential client when I
18:35 provide answers critique the answers if
18:38 needed ask follow-up questions continue
18:39 the conversation until I give the stop
18:41 roll break then give me a summary of the
18:43 whole conversation highlighting ways I
18:45 can improve my pitch you also want to
18:47 include additional material references
18:48 for your agent I've included the brief
18:50 the car company provided me that has all
18:52 the relevant information for this
18:54 project use the information from this
18:56 brief to inform your answers AI agents
18:58 can be super powerful if you can design
19:00 them correctly and these are only two
19:02 examples I really like how the course
19:03 also provides a guideline for how to
19:06 create any AI agent first you need to
19:07 assign a Persona that you want the AI
19:09 agent to take on for example act like a
19:11 successful personal fitness trainer and
19:13 talented nutritionist step two is that
19:15 you want to give as much context and
19:17 detail as you can about the scenario and
19:19 the conversation for example I'm looking
19:20 to improve my overall Fitness and adopt
19:22 a healthier lifestyle step three is to
19:24 specify the type of conversations or the
19:25 kind of interactions that you want to
19:27 have with the AI agent you might also
19:28 want to set some rules to follow like
19:30 like ask me about my workout routines
19:32 and meal planning and give me feedback
19:33 step four is to provide a stop phrase in
19:35 order to stop the conversation this can
19:37 literally be anything you want so go
19:40 wild an example they give is no pain no
19:42 gain and finally step five as a tool to
19:44 provide feedback or areas of improvement
19:46 after the conversation EDS at the end of
19:48 our conversation provide a summary of
19:50 the advice you provided and that is it
19:51 my friends you have now completed the
19:53 Google prompting Essentials course and
19:56 saved 9 hours of your time but as
19:58 promised to make sure that you actually
19:59 have retained this information I will
20:01 now put on screen the questions for the
20:03 little assessment please answer these
20:04 questions to actually retain the
20:06 information you've just learned you can
20:07 like say it in your head you can say it
20:08 to your friend your dog your cat
20:10 whatever but for proof you should write
20:12 it in the comments thank you all so much
20:14 for watching and I will see you guys in
