Large Language Models (LLMs) are evolving beyond text generation to actively use external tools for actions and reasoning, with a cutting-edge research area exploring "agents" that can autonomously plan and execute complex sequences of tasks.
Mind Map
Expand करने के लिए click करें
पूरा interactive mind map देखने के लिए click करें
welcome to the final video of this week
I'd like to share with you in this video
how LMS are starting to use tools and
then also discuss a Cutting Edge topic
of Agents which is where we let OMS try
to decide for themselves what action
they want to take next let's take a look
in the early example of a food order
taking chat bot we saw that if you were
to say sam Burger the bot May reply okay
is on the way in order for a chatbot to
enter the order and send it to you this
is what actually is happening behind the
scenes the LM can't just say Okay is on
its way because it needs to take some
action to actually send the burger to
you and so an LM might output this
response order burger for user 9876 to
sent to this address and then also say
the user message is to say okay is on
his way and L that's been fine-tuned to
Output text like this will be able to
generate an order which in this case
would trigger a software application
that passes the restaurant ordering
system a request to deliver a burger to
this user at that address and what is
shown to the user is not the full LM
output the full LM output is all four
lines of text here but only the final
line okay is on his way is what get sent
to the user as the response so this is
example of tool use by an LM where the
text the LM outputs can trigger calling
a software system to place a restaurant
order now placing an incorrect order can
be a costly mistake so perhaps a better
user interface would be before
finalizing the order to pop up a
verification dialogue to let the user
confirm yes or no if you've got the
order right before charging the credit
card and sending it to them and clearly
given that lm's outputs are not
completely reliable for any safety
critical or Mission critical action it
would be a good idea to let a user
confirm that that's the right action
before letting the L trigger some
potentially costly mistake by itself in
addition to tools for ticking actions
tools can also be used for reasoning for
example if you were to prompt an LM how
much would I have after8 years if I
deposit $100 in the bank account that
pays 5% interest an LM might generate an
answer like this which sounds plausible
but the number
$147.4 is not actually the right answer
it turns out LMS Having learned to
predict the next word or maybe even
instruction tuned are not great at
precise math and just as you I might use
a calculator to calculate the right
answer to a problem like this we can
also give the LM a calculator too to
help it get the right answer so rather
than having the L output the answer
directly if the LM were to Output this
after compounding and so on you would
have calculator 100 time 1.05 that's 5%
interest rate compounded to the power of
8 this can be interprets commands to
call an external calculator program to
explicitly compute the right answer
which turns out to be
$147 74 and plug that back into the text
to give the user the correct dollar
figure so by giving lm's the ability to
call Tools in his output we can
significantly extend the reasoning or
the action-taking capabilities of LMS to
use today is an important part of many
um applications and of course designers
of these applications s should be
careful to make sure that tools aren't
triggered in a way that causes harm or
causes IR reversible damage going Beyond
tools into a more experimental area AI
researchers have been examining agents
which go beyond triggering a tool to
carry out a single action but is
exploring whether Els can choose and
Carry Out complex sequences of actions
there's a lot of excitement and research
on agents but this is at The Cutting
Edge of AI research is is not yet mature
enough to count on for most important
applications but I want to share with
you what many in the AI Community are
excited about if you would ask an agent
that's built on top of an LM help me
research better Burger's top competitors
then an agent might use an LM as a
reasoning engine to figure out what are
the steps it needs to carry out to do
your task of researching better Burger's
competitors and this reasoning engine DM
might decide it needs to search for the
a list of the top competitors then visit
the website of each competitor and
finally for each competitor write a
summary based on the homepage content
and then perhaps by making a sequence of
calls to this reasoning engine it may
figure out that to search the top
competitors it has to trigger a tool to
call web search engine on the query
Better Burgers competitors and then
after that it may visit the websites of
some of the top competitors to download
their homepages and then additionally
call and El yet again to summarize the
text that they found on the website on
the internet there have been some nice
demos of Agents but this technology is
not really ready for prime time yet but
perhaps in the future as researchers
make it better and better it become more
useful and I think that would be the
exciting future if lm's as a reasoning
engine can help decide what's the
sequence of steps to take safely and
responsibly of course to help a user
carry out the
task thank you and congrats on making it
to the very end of week two with just
one more week to go in this course next
week we'll look at how gent of AI is
affecting companies including how you
might be able to come up with gent VII
use cases for your business as well as
look at how gent VII is affecting
society and his impact on jobs I look
Video के उस moment पर जाने के लिए कोई भी text या timestamp click करें
Share करें:
ज्यादातर transcripts 5 सेकंड से कम में तैयार
एक Click में Copy125+ भाषाएंContent Search करेंTimestamps पर जाएं
YouTube URL Paste करें
कोई भी YouTube video link डालें और पूरा transcript पाएं
Transcript निकालें
ज्यादातर transcripts 5 सेकंड से कम में तैयार
हमारा Chrome Extension लें
YouTube छोड़े बिना transcript तुरंत पाएं। हमारा Chrome extension install करें और watch page पर ही किसी भी video का transcript one-click में access करें।
