Google's 9 Hour AI Prompt Engineering Course In 20 Minutes

00:20:17
https://www.youtube.com/watch?v=p09yRj47kNM

摘要

TLDRHierdie video bied 'n samevatting van Google se kursus oor "prompt engineering" wat ontwerp is om jou te help om jou vaardighede in prompt skryf te verbeter. Dit dek vier modules insluitende belangrike raamwerke vir die skryf van prompts, hoe om prompts te ontwerp vir alledaagse take soos e-pos en data-analise, en hoe om die kreatiewe potensiaal van AI te benut. Module een fokus op die basiese beginsels van prompt skryf. Module twee bespreek hoe om prompts te ontwerp vir daaglikse werkstake deur gebruik te maak van 'n vyf-stap raamwerk. Module drie bied voorbeelde van data-analise en skep van aanbiedings. Die laaste module gaan oor gevorderde tegnieke soos ketting van denke, boom van denke, en die skepping van AI agents vir spesifieke take. Daar is ook 'n klem op die belang van menselike evaluasie om die akkuraatheid van AI gereedskap te verbeter, wat hallusinasies en vooroordele kan bevat.

心得

  • 📝 Ontwerp van prompts behels 'n vyf-stap proses.
  • 💡 Meer konteks lei tot beter AI uitvoere.
  • 🔄 Iterasie is noodsaaklik vir verbetering.
  • 🤹 Multimodale prompting gebruik verskeie tipe invoer.
  • 📊 Voorbeelde sluit data-analise en aanbiedings in.
  • 👨‍🏫 AI agents kan simulasies en terugvoer bied.
  • 🧠 Kognitiewe tegnieke soos ketting en boom van denke.
  • 🔍 Mens moet AI-uitvoer evalueer vir akkuraatheid.
  • 🎓 Kursus bevat praktiese assessments vir leerretensie.
  • 🤔 Metaprompting kan help as jy vashaak.

时间轴

  • 00:00:00 - 00:05:00

    Daar is vier modules in die cursus oor die ontwerp van effektiewe aanmanings vir AI. Die eerste module fokus op hoe om 'n aanmaning te skryf en sluit belangrike raamwerke in soos die vyfstap proses: taak, konteks, verwysings, evalueer en iterasie. Om beter resultate te kry, kan 'n persoonlikheid aan die AI toegeken word en moet spesifieke uitsette, soos 'n tabel, aangevra word. Kontekste moet deeglik verskaf word om gerigte uitsette te genereer.

  • 00:05:00 - 00:10:00

    Die video bespreek voortgesette iterasie tegnieke soos om komplekse aanmanings op te deel, verskillende frasering te probeer, en beperkings bekend te stel om die AI se fokus te vern ou. Dit sluit ook multimodale aanmanings in waar verskillende tipes media gebruik word as insette en uitsette. Twee groot uitdagings met AI-instrumente, hallusinasies en vooroordele, word bespreek, asook die belangrikheid van 'n mens-in-die-lus-benadering om uitsette na te kyk.

  • 00:10:00 - 00:15:00

    Module 2 en 3 demonstreer hoe om aanmanings vir alledaagse werkstake, soos e-posse en data-analise, te ontwerp. Belangrik is die gebruik van raamwerke om die toon en konteks aan te pas. Dit bied insigte in die praktyke om geskrewe inhoud te verbeter. Module 3 waarsku oor die sensitiewe data wat by AI ingevoer word en beklemtoon die belangrikheid daarvan om privaatheidsbeleid te eerbiedig.

  • 00:15:00 - 00:20:17

    Module 4 fokus op gevorderde tegnieke soos Prompt Chaining, Chain of Thought, en Tree of Thought om kreatiewe en komplekse take aan te pak. Dit gaan ook oor die skepping van AI-agente, soos Agent Sim en Agent X, om gebruikers te help met simulasies en kundige terugvoer. Die skepping van êrig-ontwikkelde agente vereis die definiëring van 'n persoonlikheid, konteks en gesprek reels.

显示更多

思维导图

视频问答

  • Wat is die basiese raamwerk vir prompt ontwerp in Google se kursus?

    Die raamwerk sluit 5 stappe in: taak, konteks, verwysings, evalueer, en iterasie.

  • Hoe kan jy die AI se uitvoer verbeter?

    Deur meer konteks te verskaf, die prompt in korter sinne te verdeel, verskillende formulering te probeer, of beperkings in te stel.

  • Wat is 'n voorbeeld van gevorderde prompt tegnieke?

    Voorbeelde sluit in kettingdenke en boom van denke prompts.

  • Wat is multimodale prompting?

    Dit behels die gebruik van verskeie tipe invoer soos teks, beelde, en klank vir die prompt.

  • Wat is die doel van die kursus se praktyk-assessering aan die einde van die video?

    Dit help om die kennis wat tydens die kursus opgedoen is, vas te lê.

  • Hoe gebruik AI agents in die kursus?

    AI agents kan gebruik word vir simulasie en om kundige terugvoer te gee.

查看更多视频摘要

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