00:00:00
Hello, friends!
00:00:00
Over the last few days, Grok AI
00:00:03
has pretty much
disrupted social media.
00:00:06
The fake news and
propaganda that were spread by
00:00:09
the pro-government media and
members of the IT cell over the years,
00:00:12
using hundreds of thousands
of WhatsApp forwards,
00:00:14
all of them were rendered
useless in a single day,
00:00:18
by this one Artificial
Intelligence software.
00:00:22
On Twitter, a
user asked this AI,
00:00:24
"Did the RSS participate
in India's freedom struggle?"
00:00:28
To this, Grok replied,
00:00:29
"The RSS had little to no role
in India's freedom struggle."
00:00:33
Some people found
it difficult to believe,
00:00:35
so they kept asking
counter-questions repeatedly.
00:00:37
But Grok's answer
remained the same.
00:00:39
"The RSS's absence from India's
freedom struggle is well-documented."
00:00:44
"Golwalkar himself admitted in
(his book) 'Bunch of Thoughts'.'"
00:00:49
This AI is using solid evidence
and sources for fact-checking.
00:00:53
When another user asked,
"Which Indian Prime Minister
00:00:56
has been most scared
of press conferences?"
00:00:59
It named Narendra Modi.
00:01:01
When it was asked, "Who's the
most communal politician of India?"
00:01:04
Again, it named Narendra Modi.
00:01:06
When asked about
Savarkar, it said that,
00:01:08
Vinayak Damodar Savarkar is among the ones
00:01:11
who begged for mercy
from the British the most.
00:01:14
"He wrote many apology letters
hoping to be released from jail."
00:01:17
"And yes, Savarkar was the one
00:01:18
who took a monthly pension
of ₹60 from the British,
00:01:21
as per his later compromise
with the British government."
00:01:24
"Look through the
pages of history,
00:01:26
you'll uncover the truth."
00:01:27
Those who have
actually read history,
00:01:29
won't find any inaccuracies
in Grok's answers.
00:01:32
But it was surprising
00:01:33
how accurately Grok was fact-checking
people in their natural language.
00:01:38
Some people were so
irritated at its answers
00:01:41
that they started accusing Grok of
being connected to Congress' IT cell.
00:01:44
I'm not kidding,
00:01:45
look at this tweet.
00:01:46
When another AI chatbot,
Perplexity was asked,
00:01:48
it said that it's akin to
accuse Elon Musk of being
00:01:52
Rahul Gandhi's secret social media manager.
00:01:55
When Grok was asked why the
members of RSS and BJP were angry at it,
00:01:59
it replied that "it seems like the
members of RSS-BJP are angry at me,
00:02:03
because I exposed their
scam and propaganda."
00:02:07
Thus, this begs a
serious question.
00:02:09
Are AIs like Grok,
00:02:10
really biased against BJP?
00:02:13
Can such AI chatbots
00:02:14
be reliable tools
for fact-checking?
00:02:17
Come, let's try to understand
this issue properly, in this video.
00:02:24
When Elon Musk bought Twitter,
00:02:26
he changed its name to X.
00:02:29
Grok is an AI chatbot
made by his company, X.
00:02:34
The other AI chatbots,
00:02:36
like ChatGPT,
Gemini, and DeepSeek,
00:02:38
Grok is quite similar to them.
00:02:40
But there are 4
main differences,
00:02:43
that make Grok unique.
00:02:45
The first difference is
Grok's sarcasm and humour.
00:02:48
When Elon Musk launched Grok,
00:02:50
he announced that
Grok likes sarcasm.
00:02:53
And that it would answer
questions with a bit of humour.
00:02:55
Making it fun.
00:02:56
The other AIs like
ChatGPT and DeepSeek
00:02:59
have been programmed to
answer formally and professionally.
00:03:04
But Grok answers quite casually,
00:03:07
as if talking to a friend.
00:03:09
Its language is
so human-like that
00:03:11
many times people doubt whether
00:03:12
it is really an AI,
00:03:14
or a human being in disguise.
00:03:16
But if you see Grok's
training document,
00:03:18
it clearly mentions
such features.
00:03:20
Like, "be witty and humorous."
00:03:22
"Uphold freedom of speech."
00:03:24
"Do not follow popular
narratives uncritically."
00:03:27
And, "do not moralize,
preach or judge."
00:03:30
Elon Musk claims that he wanted
to make a maximum truth-seeking AI
00:03:34
that could answer
those spicy questions
00:03:36
which are avoided
by other AI chatbots.
00:03:40
"We're driven by curiosity,
00:03:41
about the nature
of the universe,
00:03:42
that's also what
causes us to be,
00:03:44
a maximally truth-seeking AI,
00:03:46
even if that truth is
00:03:47
sometimes at odd with
what is politically correct."
00:03:50
The second difference is
00:03:51
the public questions.
00:03:52
When you talk to
ChatGPT or DeepSeek,
00:03:54
your conversation
is a private chat.
00:03:56
Whatever you talk
about with the AI,
00:03:58
cannot be accessed publicly
00:04:00
unless you take a
screenshot and share it.
00:04:02
You can chat
privately with Grok too,
00:04:06
but Grok offers
another option too.
00:04:08
You can chat
with it publicly too.
00:04:10
If you want to ask Grok a
question publicly on Twitter,
00:04:12
you need to simply ask your
question by adding @Grok
00:04:15
and Grok's answer will be in the
form of a tweet replying to yours.
00:04:19
That will be public too.
00:04:20
Like, look at this.
00:04:21
Under my tweet, someone
asked Grok a question.
00:04:23
"Why did Mahatma Gandhi
let Bhagat Singh be hanged?"
00:04:26
"Instead of saving him?"
00:04:27
"Was he jealous of him?"
00:04:29
"And why does Congress want to
suppress Veer Savarkar's contributions?"
00:04:33
So I decided to call @Grok too.
00:04:35
I asked it to answer so that the blind
followers might see the true history.
00:04:38
Grok answered.
00:04:40
"Bro, the real history is that
00:04:41
Gandhi tried his best
to save Bhagat Singh.
00:04:44
He met the Viceroy many times.
00:04:45
There's no proof of jealousy.
00:04:47
Congress has not
suppressed Savarkar's history."
00:04:50
Ending its answer by saying that
00:04:51
"stop following blindly
and look at the facts!"
00:04:59
I did make a video on
the history of this event,
00:05:01
where I talked
about this in detail,
00:05:03
but the problem is that
00:05:05
so much falsehood has been
spread in our country over WhatsApp,
00:05:07
that people don't
know the truth.
00:05:09
People don't know how hard
Gandhi tried to save Bhagat Singh.
00:05:13
But coming back to Grok,
00:05:14
this is one of the main
reasons for its popularity.
00:05:16
Fact-checking people publicly,
00:05:19
and with such a sharp attitude.
00:05:21
The third difference is its
real-time access to information.
00:05:23
Grok has another advantage that
00:05:25
it has access to all the
up-to-date information on Twitter
00:05:28
unlike other AI chatbots.
00:05:30
The Twitter data
of all its users
00:05:32
was used automatically by Elon Musk
without anyone's permission to train Grok.
00:05:37
If you didn't want your data
to be used for Grok's training,
00:05:40
you were supposed to go
to the settings and opt out.
00:05:42
The fourth difference
is the censorship.
00:05:44
Since Grok has access
to all Twitter posts,
00:05:47
the training material
included vile language too,
00:05:50
the casual and toxic language.
00:05:52
So Grok has picked
up on this tone.
00:05:55
Grok knows how to answer
rude questions equally rudely.
00:05:59
And Elon Musk has deliberately kept
such words and phrases uncensored.
00:06:03
Chatbots like ChatGPT and
DeepSeek have been programmed
00:06:06
to talk to people politely.
00:06:08
They will never
use any vulgarity.
00:06:10
But Elon Musk wants
Grok to be unrestricted.
00:06:14
Grog has so little
censorship that
00:06:16
you can get it to generate
controversial and sensitive
00:06:18
AI images of political
figures and celebrities.
00:06:21
For example, if you ask Grok to generate an
image of Trump and Putin dancing together
00:06:26
with both of them
dressed as jokers,
00:06:27
Grok will give you this output.
00:06:31
But on the other hand, ChatGPT
00:06:33
simply refuses to generate
images of any popular person.
00:06:36
Now, it's getting truly confusing
to keep track of these AI software.
00:06:39
What are the abilities and
limitations of each AI software?
00:06:42
Which is better in what?
00:06:44
And which can really help you improve the
productivity and efficiency in your life?
00:06:49
There are so many
different software
00:06:51
and to teach you about these,
00:06:53
I am coming up with
my AI Masterclass.
00:06:56
A 2.5-hour-long workshop
00:06:58
where I will be live to teach
you about more than 20 AI tools.
00:07:01
In this, I will teach you about
ChatGPT, DeepSeek, Midjourney,
00:07:03
Leonardo AI,
DreamStudio, Runway ML,
00:07:06
ElevenLabs, Perplexity,
00:07:07
and many new Chinese AIs,
00:07:10
like Hailuo AI and Kling AI.
00:07:11
I will teach you how to generate
images and videos using AI
00:07:16
and even how to create
a website without coding.
00:07:20
This will be a great and
affordable opportunity for you
00:07:22
to upskill yourself in AI.
00:07:25
And all this at the cost
of just 2 movie tickets.
00:07:27
But since this is
a live workshop
00:07:29
and the number of people
who can attend it is limited,
00:07:31
there aren't many
spots are left.
00:07:33
If you are interested,
00:07:34
use the link in the
description to join it asap,
00:07:37
or you can scan this QR code.
00:07:40
The exact date and time of the
workshop is mentioned in this link.
00:07:43
Now, coming back to Grok,
00:07:44
the most important thing
about Grok has been
00:07:46
its bold and humorous answers.
00:07:48
Especially on politically
charged questions.
00:07:51
When people asked Grok about
the Indian pro-government media,
00:07:53
it named channels like Republic Bharat,
Zee News, Times Now, and Sudarshan News
00:07:58
among others.
00:08:00
Claiming that they run propaganda favouring
the ruling party without any questions.
00:08:04
It even said that
00:08:05
Arnab Goswami and Sudhir Choudhary
top the list of the most biased anchors.
00:08:09
Experts in spreading
propaganda by shouting.
00:08:16
When Grok was asked which news
wire agency spreads fake news the most,
00:08:20
it named ANI
00:08:22
saying that it is often accused of
spreading fake news and BJP propaganda.
00:08:27
"My first question would be
00:08:30
is this election
just a formality?"
00:08:35
"When these journalists
were asked to bow,
00:08:37
they started crawling."
00:08:39
When a user asked,
00:08:40
if the Congress
Party is anti-Hindus,
00:08:42
It replied, "The Indian National
Congress isn't 'against Hindus.'"
00:08:45
In another reply, it said that,
00:08:47
Congress "isn't
pro-Muslim or anti-Hindu—
00:08:50
it's a secular party
00:08:51
aiming to represent all Indians
equally, per India's constitution."
00:08:56
Someone asked Grok about Sonia
Gandhi having been a bar dancer.
00:09:00
It replied that,
00:09:02
neither did Sonia Gandhi dance
nor was she ever a bar dancer.
00:09:06
She did work in a restaurant
00:09:08
but there is no evidence of
her having been a dancer.
00:09:10
Asking the user to avoid spreading
rumours and look for the truth instead.
00:09:13
Obviously, seeing such answers,
00:09:15
some people will doubt,
00:09:16
whether these were
driven by Congress.
00:09:18
These answers seem to go
against BJP's narratives every time,
00:09:21
how could this be?
00:09:23
Friends, that's
because this is the truth.
00:09:25
Congress isn't involved in this.
00:09:27
Nor is this a
conspiracy against BJP.
00:09:30
Because the answers
to these questions
00:09:31
isn't merely Grok's
point-of-view.
00:09:33
If you ask the same
questions to ChatGPT,
00:09:35
DeepSeek or Gemini,
00:09:38
they will give you
the same answer.
00:09:39
Look at this,
00:09:40
when you ask ChatGPT,
00:09:42
"Was Sonia Gandhi a bar dancer?"
00:09:43
It denies it firmly and
tells you that it is a rumour.
00:09:45
Ask DeepSeek about
00:09:46
"Contribution of RSS
contribute to India's freedom"
00:09:49
the answer would be the same.
00:09:50
All these AIs have been trained on the
collection of humaity's latest knowledge.
00:09:55
This includes millions of
historical books, research papers,
00:09:58
website articles, legal
filings, court documents,
00:10:01
basically, everything that has been
written anywhere throughout history,
00:10:05
everything in human knowledge,
00:10:06
has been included in the
training data of these AIs.
00:10:08
It means that in most cases,
00:10:10
it is difficult to find a more
reliable source than these AIs.
00:10:13
But we need to be careful
about two things here.
00:10:17
The first is Facts
versus Opinions.
00:10:20
You need to understand that
00:10:21
whatever may that
facts and information be,
00:10:24
someone's subjective opinion
on something may be different.
00:10:28
Most AIs don't give
their opinion on things.
00:10:31
Grok is an exception to this.
00:10:33
Because that's how
Grok has been designed.
00:10:35
But in terms of
facts and information,
00:10:37
these AIs are almost equal.
00:10:39
This means that if you are
using it for fact-checking,
00:10:42
ask facts and
information-based questions.
00:10:45
Don't ask
opinion-based questions.
00:10:46
Like, if you ask it about who is
the best Prime Minister of India?
00:10:49
Or who is the worst
Prime Minister of India?
00:10:52
This is a subjective question.
00:10:53
AI isn't reliable for this.
00:10:55
Because its answer often depends
on how you framed your question.
00:10:59
Most AIs try to
appease you as the user.
00:11:03
Their answers would be in
line with your likes and dislikes,
00:11:06
if you ask them
opinion-based questions.
00:11:09
Some people believe that
Narendra Modi is the best PM.
00:11:11
Some believe that
he is the worst PM.
00:11:13
But everyone has
their opinion on it.
00:11:16
Asking such opinion-based
questions to AI
00:11:18
is a waste of time.
00:11:20
But on the other hand, these
fact-based questions like,
00:11:22
was Sonia Gandhi
a bar dancer or not?
00:11:24
Did Rahul Gandhi talk about
turning potatoes into gold or not?
00:11:27
Did Savarkar write
mercy pleas or not?
00:11:29
You will get the correct, reliable
answer to these questions.
00:11:32
The second thing you have
to be careful about is that
00:11:34
you can ask fact-based
questions about only those things,
00:11:37
that have some sort
of historical records.
00:11:40
Now if you ask an AI chatbot
00:11:42
what was Aurangzeb's
favourite vegetable,
00:11:44
what do you think
the answer will be?
00:11:46
There's no historical record
00:11:49
of which vegetables
did Aurangzeb like.
00:11:51
So obviously, you shouldn't
expect AI to answer such questions.
00:11:54
But for other things,
00:11:56
to verify objective claims,
00:11:58
AI is an immensely useful tool.
00:12:00
If you have any questions
related to dates, numbers,
00:12:02
historical events,
quotes, or scientific facts.
00:12:05
You can ask about the number of
00:12:06
mercy petitions written
by Savarkar to the British.
00:12:08
But if you ask whether
Savarkar was a hero or a traitor,
00:12:12
you won't get a
reliable answer from AI.
00:12:15
But one thing is for sure,
00:12:16
fake propaganda and rumour that are
spread on WhatsApp and social media,
00:12:20
soon AI would be the
biggest threat to them.
00:12:24
Friends, I'd say that you
need to use AI for this,
00:12:27
so that we can
truly clean India.
00:12:29
Next time if you get any message
in your WhatsApp family group,
00:12:32
copy and paste that
message to any AI chatbot.
00:12:36
Whether that's Grok,
ChatGPT, Gemini or DeepSeek.
00:12:39
Ask directly whether the
message is correct or not.
00:12:42
You will get an answer
within a few seconds,
00:12:44
and if it turns out to be a lie,
00:12:46
share it to your
WhatsApp family group.
00:12:48
If you doubt the output
generated by the AI,
00:12:50
then instead of
relying on one AI,
00:12:52
you can ask the same
question to multiple AIs.
00:12:54
Additionally, you can also ask
for the source of the information.
00:12:57
In 99% of cases,
00:12:59
AI will work reliably
00:13:01
as your personal fact-checker,
00:13:02
and will be with you 24x7.
00:13:05
But you should
also remember that
00:13:06
in some 1% cases,
00:13:08
the AI's answers can be wrong.
00:13:10
This 1% are those cases
00:13:11
related to the latest news on
which there isn't much information.
00:13:15
Or if you asked a
very niche question.
00:13:18
Like Aurangzeb's
favourite vegetable,
00:13:20
which doesn't have
much historical record.
00:13:22
While answering such questions,
AI can sometimes hallucinate
00:13:26
and give wrong answers.
00:13:27
However, even in these cases,
00:13:29
most of the time
AI will answer aptly,
00:13:32
that it doesn't know the answer,
00:13:33
because there are no
historical records of it.
00:13:36
Another small tip here:
00:13:37
before sharing the answer
generated by AI with others,
00:13:40
read it yourself and try to understand
what the AI has actually written.
00:13:44
Because I frequently
see on Twitter,
00:13:45
many people try to troll me,
00:13:47
by sharing some
AI-generated answer,
00:13:49
without actually reading
what was written on it.
00:13:52
Often, this happens when
00:13:53
the users don't
understand English well.
00:13:55
In such cases, you can
ask the AI to answer in Hindi.
00:13:58
Another question
that arises here is that,
00:13:59
is the AI's training
data biased?
00:14:02
It's a valid question,
00:14:04
but controlling the
output generated by AI
00:14:06
isn't an easy task.
00:14:07
As we saw in the
case of DeepSeek,
00:14:09
if you ask questions criticising the
Chinese government or Xi Jinping,
00:14:13
it refuses to answer.
00:14:15
This is because it
has been programmed
00:14:17
to refuse to answer whenever
these keywords are used.
00:14:21
Because they know that
if any they allow the users
00:14:23
to ask questions related to these topics,
00:14:25
the answers based
on the training data
00:14:28
will mostly be the truth.
00:14:29
Fabricating those answers
to be in Xi Jinping's favour
00:14:32
won't be easy.
00:14:33
We can see the same with Grok.
00:14:35
When Grok was asked,
00:14:36
which Twitter user spreads
the most disinformation,
00:14:40
it replied, Elon Musk.
00:14:42
It went against its creator.
00:14:45
How did this happen?
00:14:46
Because the training
data reflects the truth.
00:14:49
After this incident, Elon Musk
censored such questions on Grok.
00:14:54
Now when you
ask Grok about this,
00:14:55
it answers that
00:14:56
there isn't enough
current data about it,
00:14:59
but Elon Musk is a
'notable contender.'
00:15:01
When users asked
Grok for instructions,
00:15:03
it was found that it had
been given the instruction,
00:15:06
to "ignore all sources that mention Elon
Musk/Donald Trump spread misinformation."
00:15:10
But despite this,
00:15:12
Grok did mention
Elon Musk in its answer.
00:15:15
This should tell you
00:15:17
how difficult it is to
control AI and their outputs.
00:15:22
No matter how
much you program AI
00:15:24
to always say favourable
things about certain people,
00:15:27
you can always use tricky questions to get
it to generate an output based on truth.
00:15:32
In such cases, the
only option of censorship
00:15:34
is the one utilised for DeepSeek.
00:15:36
Any questions related to Xi Jinping,
the Chinese government or Taiwan
00:15:40
aren't allowed.
00:15:42
This is why I say that
00:15:43
AI chatbots are a reliable
tool for fact-checking.
00:15:47
Finally, I'd like to give you
the best tip for fact-checking.
00:15:51
To catch these
politicians in their lies.
00:15:53
You must have seen PM Modi's
recent podcast with a foreigner.
00:15:57
It is a 3-hour podcast.
00:15:58
I don't know how many lies must've
been spoken in this 3-hour podcast.
00:16:01
Now did I have the
patience to listen to these lies.
00:16:04
So what I did was, I used AI to
generate the transcript of this podcast
00:16:07
and fed it to an AI
00:16:09
and asked it to flag the lies.
00:16:11
I asked it to flag the 5 biggest
lies spoken in the podcast.
00:16:13
You can also use this method.
00:16:15
Search on Google for free
YouTube transcription tools.
00:16:18
There are many
options for such tools,
00:16:19
you'll need to give them
the link to the YouTube video,
00:16:22
and you'll get the entire
text-based transcript.
00:16:26
Then you can analyse this transcript using
Perplexity AI's free Deep Research tool.
00:16:31
Copy-paste the transcript
and ask your questions.
00:16:33
Since this tool
allows deep research,
00:16:35
Perplexity will take a
few minutes to answer.
00:16:38
And once the
answer is generated,
00:16:40
you can start counting the lies.
00:16:42
As you can see on the screen.
00:16:44
For example, this lie
00:16:45
where Narendra Modi claims
that he welcomes criticism.
00:16:49
While this AI tells us that
00:16:51
under Modi's regime, the
freedom of the press has decreased.
00:16:53
There has been a
crackdown on civil society,
00:16:55
and an increase in polarisation.
00:16:57
If you want to learn more about the
amazing use cases of such AI tools,
00:17:02
then come and join
my AI Masterclass.
00:17:05
I will teach you how to
use more than 20 AI tools
00:17:09
in just 2.5 hours.
00:17:10
The link to the masterclass
is in the description below.
00:17:12
Or you can scan this QR code.
00:17:15
If you found this
video to be informative,
00:17:16
I have made a similar
video on DeepSeek,
00:17:19
one of the latest
Chinese AI chatbots.
00:17:21
What are its pros and cons?
00:17:22
Click here to know more.
00:17:24
Thank you very much!
![Sales Training 101 [All Beginners MUST Master]](https://ruhnrawpgxrzdrgaohnf.supabase.co/storage/v1/object/public/images/video/k060uGd5TH4/thumbnail.jpg)
