0:04 what is generative AI good for one of
0:07 the reasons that question is a bit hard
0:09 to answer is because AI is a general
0:11 purpose technology unlike a lot of
0:14 Technologies like a car which you know
0:16 good for transportation or microwave
0:18 oven good for heating up food AI isn't
0:21 useful just for one thing it's useful
0:22 for a lot of things and that almost
0:25 makes it harder to talk about but let's
0:27 take a look at what a general purpose
0:29 technology really means similar to
0:33 Electric AI is useful for many tasks if
0:35 would ask you what is electricity good
0:38 for or what is the internet good for
0:39 these are other general purpose
0:41 Technologies and it's almost difficult
0:43 to think what is electricity good for
0:46 because it's so pervasive and it's used
0:48 around us for so many different things
0:51 in fact as you saw earlier supervised
0:53 learning is useful for many tasks like
0:55 span filtering advertising speech
0:57 recognition and many others and
1:00 generative AI is like this too in the
1:02 last video you saw a few of the tasks
1:05 that an LM can carry out answering
1:06 certain questions and helping with
1:09 writing for example let's discuss more
1:11 broadly a framework for what kinds of
1:15 tasks LS can do first off gbi generates
1:18 text so not surprisingly perhaps is
1:22 useful for writing I routinely use gent
1:25 AI tools as a brainstorming companion so
1:27 if you're trying to name a product you
1:30 can ask it to brainstorm some names and
1:32 it comes up with some creative
1:34 suggestions L can also be good at
1:36 answering questions and if you give them
1:38 access to information specific to your
1:40 company they can help members of your
1:42 team find information that they need in
1:44 this case about the availability of
1:47 parking at the office in addition to
1:50 writing gent of AI is also good for what
1:54 I'm going to call reading toss where
1:56 you're going to give it a relatively
1:58 long piece of information and have it
2:01 generate a short output for example if
2:04 you run an online shopping e-commerce
2:06 company and you get a lot of different
2:09 customer emails jent of AI can read the
2:11 customer emails and help you very
2:14 quickly figure out is this email a
2:16 complaint or not which can be used for
2:20 helping to Route complaints to the
2:21 appropriate Department to be handled
2:24 quickly so given I love my new llama
2:28 t-shirt fabric is so soft that's not a
2:31 complaint but if someone emails I wore
2:33 my llama t-shirt to friend's wedding now
2:35 they're mad at me for stealing the show
2:37 well maybe that is a complaint but J of
2:40 AI can help you route emails to the
2:42 right department and I call this a
2:45 reading toss because it's looking at the
2:48 relatively long piece of text that is a
2:50 customer email and then generating a
2:52 relatively short output just yes or no
2:54 is this a complaint or not and while
2:56 supervised learning can also be used for
2:59 this particular task we'll see later
3:01 that General of AI is allowing these
3:03 sorts of reading tasks this and other
3:05 examples that we'll see later this week
3:08 to be built much more quickly and
3:11 inexpensively lastly genive AI is also
3:15 used for many chatbot types of tasks
3:18 whereas chat gbd and B and B chat are
3:21 general purpose chat Bots gen of AI
3:23 technology large language models is also
3:26 enabling many special purpose chat Bots
3:28 to be built in this example here's what
3:31 a chatbot might be like for taking
3:34 online orders where a user can say a
3:36 like a cheeseburger for delivery and the
3:39 chatbot acknowledges and puts the order
3:42 through for the user now in talking
3:45 through these tasks I find that there
3:47 sometimes useful to distinguish between
3:50 two different types of om based
3:55 applications one is examples like this
3:57 brainstorming one where it be quite
3:59 natural for you to type a prompt like
4:01 this into to CH or B or B chat or one of
4:04 the other free or paid large language
4:07 models on the internet and get a result
4:09 back so I'm going to call an application
4:12 like this a web interface based
4:15 application in contrast in the example
4:18 of recognizing of an email is a customer
4:21 complaint this fits more into a
4:24 company's email routing workflow and it
4:27 doesn't really make sense for anyone to
4:30 cut and paste customer emails one at a
4:33 time into a web interface to get back
4:35 answers as to which ones are actually
4:38 complaint emails so this is an example
4:40 of an LM that would make sense when it's
4:45 built into a larger software automation
4:47 that in this case helps with a company's
4:50 automated email routing so I'm going to
4:53 call this a LM based software
4:56 application the second writing example
5:00 of answering HR questions it turns out
5:03 this also will make more sense as a
5:05 software based LM application because
5:08 it'll need access to information about
5:11 your specific company's parking policy
5:13 for employees whereas a general lar
5:15 langage model on the internet probibly
5:17 doesn't have that information we'll talk
5:19 more later in this course about how this
5:21 technology is built and most of the
5:23 specialized chat Bots will also be
5:27 software based L applications so in the
5:29 rest of this course I'm going to use
5:32 these two symbols to distinguish between
5:34 web interface use cases and software
5:38 based um applications and for many
5:40 people it may be easier to get started
5:43 with some of the web interface use cases
5:45 because you just go to a website like
5:49 trbd or bar or Bing and type in a prompt
5:52 and get result back but I think both the
5:55 web interface based applications and the
5:57 software based applications are
6:00 important and will be very useful for
6:02 for individuals and for companies I
6:05 founded the framework of writing reading
6:07 and chatting as a useful way to think
6:09 about the many different tasks that LM
6:12 large language model can do in the next
6:15 three videos we'll dive more deeply into
6:18 many different examples of writing
6:21 reading and chatting TSS and I hope that
6:23 you find some of them potentially useful
6:25 for your own work so I look forward to
6:27 seeing you in the next video where we'll
6:30 talk more about writing and then I look
6:32 forward to enjoy my burger
