00:00so with this lesson plan you can tell
00:02your students that you're going to talk
00:04about the birds and bees of AI
00:07where do answers come from or in other
00:11how does AI work so this lesson is going
00:14to give your students an overview of how
00:16generative AI works and why and how it's
00:20different than services like Google
00:23all right so you tell your students
00:25today we're going to talk about AI
00:28and the first thing I want to know from
00:29you like what do you think is AI good or
00:35so I want you to take a minute or two
00:38and I want you to think on your own like
00:39what are the positives and what are the
00:41negatives of AI in terms of your own
00:46so you give them some time to think
00:47about that then you say okay I want you
00:49to pair up and I want you to talk with a
00:51partner talk with someone near you with
00:53the pros and cons you came up with
00:55give them another couple minutes to talk
00:58then you say okay let's let's discuss
01:00this amongst all of us then you
01:03facilitate a discussion so you still
01:05notice we're doing the think pair share
01:07the reason why it's going to create a
01:10lot more safety and a lot more Dynamic
01:12engaging conversation if you just ask
01:15hey in general is AI good or bad you're
01:18going to have the same students
01:19responding that you always have
01:21responded to your prompts so use the
01:23think pair share method to get more
01:25students involved and in the
01:26conversation and we generally genuinely
01:29want everyone to be thinking about the
01:30positives and the negatives of AI
01:33okay so then you say you know what it's
01:35it's normal for for people to feel
01:38anxious about Innovations like AI
01:42in fact I'm going to give you three
01:43different quotes from very prominent
01:47who are largely skeptical of these sort
01:50of Innovations so it's dangerous and
01:53will never be practical
01:54the novelty in a fad
01:57has too many shortcomings to be
01:59seriously considered
02:01so these are all quotes from people who
02:03it turns out all of these quotes can
02:06apply to AI but they're not actually
02:10these are quotes from innovations that
02:12we now all take for granted
02:15like I said it's common to have these
02:16sort of concerns whenever we are
02:19approached or meet something new and
02:21Innovative so for example that very
02:22first quote about being dangerous and
02:26um that's actually a quote that the full
02:28context is if God had intended us to fly
02:32he would have given us Wings it's
02:33dangerous it will never be practical
02:35that quote is actually from Orville and
02:38Wilbur Wright's father so the first
02:41people to have manned powered flight
02:44their father said that this is dangerous
02:46and will never be practical
02:48that quote about it being a novelty and
02:50bad that's actually full of context is I
02:52predict that the horse will be with us
02:55the automobile is a novelty in a fad and
02:58that comes from the United States
03:00President William McKinley
03:02and then finally the telephone has too
03:05many shortcomings to be seriously
03:06considered as a means of communication
03:09this is of course from William Orton the
03:12president of Western Union the most
03:15powerful telegram company
03:17in the United States so
03:20what I want you to know you tell your
03:22students what I want you to know is that
03:24at the end of the day there are
03:26positives and negatives to every
03:27Innovation but I don't want you to be
03:30stuck on the wrong side of history for
03:33this and other Innovations you run into
03:35so I want to explore
03:37the positives and negatives of AI and I
03:40want you to really understand how they
03:41work so you can understand for
03:43yourselves like why these are beneficial
03:45or why these are dangerous
03:48okay so to do that I want you to learn
03:51something new the same way AI learns so
03:55you can see how it collects and presents
03:57information so I'm going to teach you a
03:59brand new language this language doesn't
04:01exist and so you're going to learn it
04:03from scratch the same way that
04:06generative AI models learn things
04:08so I'm going to present you a number of
04:10words from this new language
04:13then you tell your students what I need
04:15you to do as my generative AI is I need
04:18you to look at this list of words I need
04:20you to tell me which of these words
04:23which were words refer to cats
04:26and which words refer to neither
04:31so if I ask you okay which of these are
04:36as a generative AI you would have no
04:40because you haven't done the first step
04:42of AI building which is the training
04:45first someone has to give you what we
04:47call labeled data I have to tell you a
04:50few of the answers and then you can
04:51start to find patterns and then discover
04:53your own your own subsequent uh words
04:58that match that pattern
05:01so what I'm going to do is I'm going to
05:02give you some training data I'm going to
05:04show you a couple bird words and when I
05:06do I want you to notice what happens in
05:09your own mind like what do You observe
05:13all right so the first bird word is
05:17okay I'm going to show you the second
05:18word again pay attention what happens in
05:20your brain when you see this second bird
05:29so you may have a sense now of what bird
05:33words look like but I'm going to show
05:36third bird word is vast
05:39okay so we have one bird word left
05:42I want to know does anyone have any
05:44ideas for what that could be so you ask
05:47everyone yeah you can take some guesses
05:49from people for the most part they'll
05:51come up with it it's veal so bio is the
05:56so that was an example of labeled
05:59training data I told you some of these
06:02words are bird words and then you were
06:05able by and large to find the fourth
06:07bird word on your own that's because
06:09you've been doing something called
06:10pattern recognition and you discovered
06:13the pattern the bird words in this
06:15language start with the letter b
06:18okay so let's do the same thing this
06:20time we're going to switch over to cat
06:22so I'll show you a couple of the cat
06:25and then you tell me when you have a
06:28sense of what the next cat word is so a
06:38all right so now you can ask okay any
06:40ideas we've got one cat word left is
06:43what the last cat word is
06:46so you can pull the class and see you
06:48know who thinks what everyone thinks
06:50most students will think it's lorat
06:54so you say okay great so most of us
06:55think it's loret it turns out though in
06:58this language lorat doesn't refer to a
07:04but this is normal this happens all the
07:06time with AI and with training during
07:09the training process it makes mistakes
07:11and this is how it learns
07:13but it turns out that these sort of
07:15mistakes are really important to know
07:17about that the quality of your AI output
07:21has everything to do with the quality of
07:23the data that you put into it
07:25for example there was a study done where
07:28doctors were trying to use AI to
07:31determine whether or not moles were
07:35so they took a bunch of pictures of
07:37moles they fed it through an AI system
07:39they trained it on you know told the AI
07:41okay these ones are malignant these ones
07:43are not and then it they let it go with
07:47a large sample set of unlabeled data
07:50and what the aai discovered was pretty
07:54profound and pretty amazing it
07:57discovered that rulers cause cancer
08:01or rather what it discovered was that
08:04if a picture had a ruler in it it was
08:08far more likely to be a malignant mole
08:14that's because the data it was trained
08:18the day the pictures that the AI was
08:21given of malignant moles largely had
08:24rulers in them because they were taken
08:26in a doctor's office where the ruler was
08:28used to give a scale of the size of the
08:31and the benign moles they were taken
08:34outside of a doctor's office or without
08:36a ruler just because they were you know
08:41and so the the AI found a pattern but
08:44it's not the kind of pattern that we
08:45would find particularly valuable found a
08:47pattern that hey if there's a ruler it's
08:49more likely to be malignant
08:51so this is really important to think
08:53about that whenever you're using a
08:54generative AI that you understand where
08:57the data has come from because if the
08:59data is skewed if the input is skewed
09:01the output will be skewed
09:03so we want to think about the data
09:04particularly in terms of like political
09:08gender bias race bias any sort of biases
09:11that go into our input of our aim
09:14training will likely become likely be
09:17shown in the output as well
09:20okay so we want to be careful of our
09:23training and just know that the quality
09:25of our input will determine the quality
09:27of our output we also know that the
09:29route is not our last cat word so let's
09:31take our last guess for what the right
09:35you ask your students again most of them
09:37will say rat and that's correct so rat
09:40is the final cat word
09:44is that cat words have to have double
09:46letters and end with the at sound
09:50so we now have two different patterns
09:52for this new language we know that cat
09:55words must contain a double letter and
09:58end with an at sound
10:00and we know that bird words start with a
10:05so now you have all the information you
10:07need to be a generative AI for this new
10:10language in particular if I want you to
10:13do one specific task for me if I want
10:18and come up with a word that means
10:23so you tell your students I'm going to
10:24give you 30 seconds to see if you can
10:27combine the patterns that we know about
10:29this new language to generate a brand
10:32new word that doesn't exist
10:37so see if you can take those two
10:38patterns and come up with a definition
10:42give them 30 seconds give them a minute
10:45and you ask people to share okay what
10:46did you come up with or a flying cat and
10:48what you're looking for is any example
10:50of a word that starts with a B contains
10:52two letters in it then ends with an at
10:55sound so you can ask them okay how did
10:56you spell it make sure that they got all
10:59the right words all the right characters
11:00and you say great that's it you
11:02understand how generative AI works it
11:05collects it understands it learns a
11:07number of patterns and then combines
11:08those patterns to generate new and novel
11:13and that's distinct from something like
11:17so we think about where AI generates its
11:19answers from it's tempting to think of
11:21it as like oh it's the next generation
11:23of Google but it is fundamentally
11:25different this is not a search engine
11:28services like Google or search engines
11:29they are a collection of data
11:33they've surfed the web to go aggregate a
11:36bunch of discrete pieces of information
11:37so grabbed a bunch of pieces of
11:39information and stored it in a database
11:41that says okay this piece of information
11:42is here this piece of information is
11:44here and so when someone searches for
11:46something it just says okay well here's
11:48everything I know about that type of
11:49information and here's where to go find
11:52it's like a giant dictionary just you
11:55you look up a word it tells you what
11:56that word means that's what Google can
11:58do you want to know about something it
11:59says oh here's all the information I
12:00found about that topic
12:03it's very different from our generative
12:06generative AI is not a collection of
12:08data it's a collection of patterns
12:10there isn't so much as like a big
12:13database full of information so much as
12:15there is what we call a neural network
12:16of understanding of patterns
12:19in these patterns each of these
12:21individual patterns you can think of it
12:22kind of like a Lego block
12:24where if you understand a set of
12:28patterns you can combine them to create
12:30new and novel information
12:33so just like we took a uh the
12:37combination of the idea of a bird the
12:39idea of a cat and we combined them into
12:40something new to make a flying bird
12:43excuse me flying cat
12:46I can ask a generative AI to generate an
12:50image for me of something that's it's
12:53as far as I know no one has ever made a
12:55flying cat out of Legos but I can ask a
12:58generative AI image model to generate
13:01one for me because it knows okay
13:03things that fly typically have wings so
13:05I should make something as wings
13:07I know what a cat looks like you know
13:09it's got four legs it's got a cute face
13:11it's got a tail uh so I'll put some
13:14wings on on something that looks like a
13:16cat and then I know that things that are
13:18made out of Lego are blocky and have
13:20these little bumps on top and boom it
13:22combines those patterns into create
13:25something new and novel that has never
13:29and so that's the difference Google is
13:31pulling out is references to things it's
13:33already seen before generative AI is
13:36coming up with something brand new that
13:38may exist you know who knows maybe this
13:41is exactly the same image as someone
13:42else has done but that would just be a
13:44fluke that would just be random chance
13:46it wouldn't be that that the AI is
13:49necessarily pulling that information out
13:50it's just combining the patterns it
13:52knows and the best most effective way
13:56okay so let's talk about benefits and
13:58challenges of this type of Technology
14:01so you can ask your students okay so do
14:04we think AI is going to take your jobs
14:06so is your degree worth nothing because
14:09AI is going to come in and do all your
14:12um no AI is not going to take all of our
14:15jobs but it will change them it will
14:17fundamentally change them same way that
14:19calculators fundamentally change the way
14:21we do math and the internet
14:23fundamentally changed the way we do
14:24everything AI will change the way we
14:28work we don't know exactly how yet but
14:31it's already changing the way we work
14:33and it's very likely to continue to do
14:37okay but in the short term can AI help
14:39you complete your assignments
14:41the answer here is kind of
14:44so it can absolutely be helpful if you
14:47need to write something if you're
14:49brainstorming ideas if you want to learn
14:50a new topic that you don't well know
14:52well AI can help do all of those things
14:55but there's a caveat there's a condition
14:59and that is you have to know whether or
15:01not AI will lie to you
15:04and if you don't already know the answer
15:06is resounding yes AI will lie to you it
15:09won't know that it's lying to you
15:11but it will very confidently combine the
15:13patterns it knows to present something
15:15that looks very believable
15:19but that can be very wrong and that's
15:22what we call a hallucination
15:24so you can ask your students have any of
15:26you experienced a hallucination before
15:31you can get a couple examples from
15:32students and then you can say well it
15:34turns out actually all of you have
15:36experience with AI hallucinations
15:39that's because those quotes I showed you
15:41I asked chat gbt for those quotes
15:44and it turns out they're all lies
15:48completely fabricated
15:51now I believed they were real quotes I
15:53thought they were real genuine quotes
15:55until I did some fact checking on them
15:58it turns out that no that Chad gbt had
16:01just combined the patterns it knows
16:02particularly well it knows what a
16:04compelling quote looks like for someone
16:07who is fearful of innovation
16:10in fact it knows likely it's probably
16:12seen in its in its pattern and
16:14recognition that Orville and Wilbur
16:16Wright's father was actually a pastor
16:19so him you know quoting you know
16:23invoking God to say hey if God had never
16:25God had intended us to lie would have
16:28um I mean that's a quote that many of us
16:29have heard about already so it makes
16:31sense that that would be a real thing
16:33he's a pastor it would make sense that
16:35he would say that but he never did
16:38so this is something you really want to
16:40watch out for when you're using AI in
16:42any context whether it's to do your
16:43homework or to do a job but if you're
16:46asking it for factual information
16:48it may not have access to that and is
16:51very likely to lie to you
16:53if you're asking it to create new and
16:55novel things that's where its strengths
16:59so yes it can help you it can help you
17:01with creativity and brainstorming but
17:03don't assume that all the dancers are
17:08all right so there you can tell your
17:09students so that my friends that's the
17:11birds and Beads of AI That's where the
17:13answers come from and that's the
17:14difference between Google and a tool
17:18so you can use this as a jumping off
17:20point for any number of lessons about AI
17:23if you would like to use a curriculum an
17:26entrepreneurship curriculum that takes
17:27full advantage of AI including
17:30integrating AI into six different
17:32modules including customer storming
17:34solution ideation MVP development
17:37customer interviewing Financial modeling
17:39and defining a student's purpose
17:42check out the experiential
17:44entrepreneurship curriculum you can do
17:48so there we've got a full experiential
17:51entrepreneurship curriculum that teaches
17:5230 plus modern entrepreneurship skills
17:55with a flexible schedules whether you're
17:57in a quarter system and an MBA program
18:01seven weeks eight weeks all the way up
18:02to 15 weeks every entire year-long
18:04programs it's got LMS integration
18:07Advanced attendance tracking it's got a
18:10virtual assistant that can help you with
18:12assessment and all for less money than a
18:14textbook so take a look at the
18:17experiential entrepreneurship curriculum
18:19we call it exec at teachinge.org
![Welp, Here's what the Fed *Just* Said [Fed FOMC Minutes]](/_next/image?url=https%3A%2F%2Fi.ytimg.com%2Fvi%2Fl8u23oq1E9E%2Fhqdefault.jpg&w=3840&q=75)
