1
00:00:00,000 --> 00:00:02,654
This audio may be AI-generated.

2
00:00:02,704 --> 00:00:05,359
Always verify important AI answers with trusted sources.

3
00:00:05,409 --> 00:00:08,063
World 3 is about checking AI answers.

4
00:00:08,113 --> 00:00:10,767
AI can sound confident and still be wrong.

5
00:00:10,817 --> 00:00:13,472
That is called a hallucination.

6
00:00:13,522 --> 00:00:16,176
A hallucination is when AI gives an answer that sounds correct

7
00:00:16,226 --> 00:00:18,881
but is wrong, unsupported, outdated, or made up.

8
00:00:18,931 --> 00:00:21,585
The safety phrase for this world is: confidence is not

9
00:00:21,635 --> 00:00:24,289
correctness.

10
00:00:24,339 --> 00:00:26,994
Use this verification checklist.

11
00:00:27,044 --> 00:00:29,698
One: identify the claim.

12
00:00:29,748 --> 00:00:32,402
Two: check a trusted source.

13
00:00:32,452 --> 00:00:35,107
Three: verify names, dates, and numbers.

14
00:00:35,157 --> 00:00:37,811
Four: ask whether a mistake could cause harm.

15
00:00:37,861 --> 00:00:40,515
Five: get human expert review for high-stakes topics.

16
00:00:40,565 --> 00:00:43,220
Here is a claim: Claude is always correct if you write a good

17
00:00:43,270 --> 00:00:45,924
prompt.

18
00:00:45,974 --> 00:00:48,629
That claim is wrong.

19
00:00:48,679 --> 00:00:51,333
A better prompt improves the answer, but it does not guarantee

20
00:00:51,383 --> 00:00:54,037
correctness.

21
00:00:54,087 --> 00:00:56,742
World 3 teaches a powerful habit: before you trust the answer,

22
00:00:56,792 --> 00:00:59,446
check the source.