1
00:00:00,000 --> 00:00:08,480
Now, let's talk a bit about GenAI.

2
00:00:08,480 --> 00:00:10,640
Chances are that at some day, you

3
00:00:10,640 --> 00:00:14,680
will be responsible for hosting a GenAI system.

4
00:00:14,680 --> 00:00:17,240
So what are the GenAI components?

5
00:00:17,240 --> 00:00:19,240
Start with the large language model.

6
00:00:19,240 --> 00:00:22,440
That is considered the brain of the GenAI solution.

7
00:00:22,440 --> 00:00:23,920
And it typically has been trained

8
00:00:23,920 --> 00:00:26,120
on large amounts of data, and it will

9
00:00:26,120 --> 00:00:28,559
be multiple gigabytes in size.

10
00:00:28,559 --> 00:00:32,880
It requires a lot of GPU RAM in order to be used.

11
00:00:32,880 --> 00:00:34,959
Training LLMs from scratch requires

12
00:00:34,959 --> 00:00:37,240
huge amounts of resources.

13
00:00:37,240 --> 00:00:40,200
And for that reason, most companies that run an LLM

14
00:00:40,200 --> 00:00:42,360
don't train the LLMs themselves.

15
00:00:42,360 --> 00:00:45,160
They use a standard LLM, which next they

16
00:00:45,160 --> 00:00:47,520
will be fine tuning only.

17
00:00:47,520 --> 00:00:49,520
The AI client provides a prompt that

18
00:00:49,520 --> 00:00:52,599
is used to get information out of the AI system.

19
00:00:52,599 --> 00:00:54,520
That is where the user is interfacing

20
00:00:54,520 --> 00:00:56,599
and is asking the question.

21
00:00:56,599 --> 00:01:00,000
In a process that is referred to as inference.

22
00:01:00,000 --> 00:01:04,279
The AI agent is a goal-driven system on top of the LLM

23
00:01:04,279 --> 00:01:06,000
that can perform actions for you.

24
00:01:06,000 --> 00:01:08,839
It can perceive, it can reason, and it can act.

25
00:01:08,839 --> 00:01:11,519
So it can make bookings for you, for instance.

26
00:01:11,519 --> 00:01:14,000
Or it can make changes to configuration

27
00:01:14,000 --> 00:01:16,959
or whatever you have configured your agent to do.

28
00:01:16,959 --> 00:01:18,800
Think about questions like, analyze

29
00:01:18,800 --> 00:01:20,720
which servers are running out of disk space

30
00:01:20,720 --> 00:01:22,720
and send me a report.

31
00:01:22,720 --> 00:01:27,959
And agents turn Gen-AI systems into real problem solvers.