Generative AI significantly simplifies and accelerates the development of sophisticated software applications, particularly those dealing with unstructured data, by moving beyond traditional machine learning approaches to more accessible prompt-based development.
Mind Map
Clique para expandir
Clique para explorar o mapa mental interativo completo
welcome back last week we discussed how
generative AI can be used either via web
user interface or be built into a
software application in this week we'll
take a look at how many amazing software
applications are being built using G of
AI and we'll also take a look at some
technology options that go beyond just
prompting and that allow you to do much
more with gent of AI for example having
it operate on your own propri documents
rather than just on what is learned from
public sources on the internet let's
take a look we saw last week a few
examples of generative AI applications
such as writing answers to questions
that may require access to information
about your company's parking policy in
this example or reading restaurant
reviews on the internet to help with
reputation monitoring or building a
chatbot to help take food orders it
turns out that that while some
applications like this did exist and
were built before the rise of gen AI gen
AI has made building these applications
much easier and in many cases it's made
them work much better as well let me
illustrate with the example of reading
restaurant reviews for reputation
monitoring a few years ago if you wanted
to build a system for reading restaurant
reviews it would have taken writing a
lot of software code that looks like
this pages and pages of software that
you need machine learning Engineers to
write and specifically the process of
building a restaurant reputation
monitoring review system would have
looked like this you would use
supervised learning that's that
technology that maps from inputs a to
outputs B and if I were building this
system I would start by collecting maybe
a few hundred or a few thousand data
points with examples like this I would
have a review that suit Dum things ever
eaten that sounds delicious and label
that as a positive review the colorful
table cost made me smile that's positive
or not worth the three Monon wait that
be a negative review and the process of
building the system would involve first
getting label data then finding AI team
to help train an AI model on the data to
learn how to Output positive or negative
depending on different inputs a and then
finally you might have to find a Cloud Server
Server
like AWS or Google cloud or a zuro to
deploy and run the model so that when
you then input best bub I've ever had
that would hopefully recognize this as
having a positive sentiment and this
process would often take months in
contrast if you were to use prompt based
development this is the code you would
need to develop a sentiment classifier
first here's how I we specify a prompt
in code my prompt which have set equal
to 2 positive text there's the
instruction text classify the following
reviewers having the positive negative
sentiment and then here is the review
text and after specifying The Prompt in
codes I just need one line of code to
call the large language model to get a
response back and then I'm going to have
it display or print the response so this
is pretty much all the code it takes to
build such a system and in fact in the
next video I'll share you an optional
exercise where you can try out this code
yourself whereas with the traditional
approach to building a sentiment
classifier using supervised learning the
timeline for the project might have been
a month to get say a th labeled examples
with a th000 reviews and positive
negative labels after collecting the
data it might have taken a team say
three months to train the AI model on
data and then another three months to
deploy it and make sure it's running
well and it's regular and robust I don't
know if this seems like a long time to
you but for many really good machine
learning teams I've worked with this 6
to 12 month timeline was pretty
realistic for what it took to build and
deploy a valuable AI model and this
worked and this was very valuable for a
lot of applications but this just took a
long time in contrast for prompt based
AI this is what it feels like you can
specify a prompt in minutes or maybe
hours and then deploy the model in hours
or maybe days so there are now many
applications that have previously taken
me and very good machine learning teams
maybe six to 12 months to build that
today I think there are millions of
people around the world that can now
build in maybe days or a week and this
is fantastic because this lowering of
the barrier to entry to building such
applications is leading to flourishing
of a lot more AI applications with one
important caveat which is that as we
discussed last week gen of AI tends to
work much better for unstructured data
like text and images and audio um but
with that admittedly important caveat
the number of AI applications built on
top of gender AI is just letting the
community do much more than ever before
in the next optional video I'd like to
invite you to try out some codes with me
for reading restaurant reviews and
classifying sentiments it's fine if
you've never seen or written a line of
code before in your life but I'm hoping
to convey to you how code is needed to
do this now and let you try it out
yourself so I hope you take a look
though also feel free to skip it if you
wish and after that we'll come back and
talk about what building a genitive AI
software project feels like when we talk
Clique em qualquer texto ou marcação de tempo para ir diretamente àquele momento do vídeo
Compartilhar:
A maioria das transcrições fica pronta em menos de 5 segundos
Cópia com um cliqueMais de 125 idiomasPesquisar no conteúdoIr para marcações de tempo
Cole a URL do YouTube
Insira o link de qualquer vídeo do YouTube para obter a transcrição completa
Formulário de extração de transcrição
A maioria das transcrições fica pronta em menos de 5 segundos
Instale nossa extensão para Chrome
Obtenha transcrições na hora sem sair do YouTube. Instale nossa extensão para Chrome e acesse com um clique a transcrição de qualquer vídeo direto na página de reprodução.
