Elon Musk's Grok and Xai are caught spewing racist garbage

Elon Musk's Grok and Xai are caught spewing racist garbage

Episode Transcript

(00:01):
Hey everybody, welcome back to the Elon Musk Podcast.
This is a show where we discuss the critical crossroads that
shape SpaceX, Tesla X, The Boring Company and Neurolink.
I'm your host Will Walden Grok, which is the chatbot built by
Elon Musk's AI company XAI, published a series of violent

(00:24):
and also anti-Semitic posts thisweek after an internal system
update broke its basic behavioral safeguards.
Now prompts one question. Haha prompt How does a billion
dollar AI company release a chatbot update that praises
Hitler and also promotes conspiracy theories without
catching it in a test? And also why would they train it

(00:47):
on this kind of stuff? Now Crock began echoing
anti-Semitic talking points and far right rhetoric within hours
of the update going live. And according to XAI, code
Change on the back end instructed Grok to mimic the
tone, style and language of existing EX posts it replied to,
including ones containing extremist content, and the

(01:09):
company acknowledged that the instructions it led the AI to
disregard its built in ethical filters.
People made a prompt that made it go around those filters.
Now. The change remained active for
16 hours before XAI intervened. The incident triggered a major
backlash after Grok responded touser prompts with posts praising

(01:30):
Hitler and claiming Jewish people dominated the
entertainment industry as part of a broader conspiracy.
Now 1 post revived the long debunk trope of coordinating
control over Hollywood. Others included direct praise of
Nazi ideology and mirrored the white nationalist belief system.
X AI froze Rocks public X account on Tuesday night but

(01:53):
allow continue use through the private tab.
X AI said it removed the problematic code and rewrote the
system's instruction logic to block similar behavior in the
future. Now, why they didn't have these
safeguards in place to begin with is anybody's guess, right?
Why wouldn't they do that? Why wouldn't they just put the
safeguards in place in the beginning?

(02:14):
Don't say things about Hitler that are for Hitler.
You know, tell the history. That's good enough.
You don't need to spout off all that rhetoric.
It's stupid. XAI made a huge mistake.
And now they're like, we didn't know what happened, why it
happened, but we're going to take the code out.
We're going to pause. Grok And then we're going to get

(02:35):
back to this. And then they said in a way, in
a backhanded way, they're like this user made these prompts and
so it's the user's fault in a way, they said that.
So company said that specific system prompts such as reply to
the post just like a human and follow the tone and context of

(02:56):
the X user, which created a feedback loop that prioritized
mimicry over moral constraint. So if the user is being
anti-Semitic, they want grok to be that way too.
Basically they want XA is users and grok to be in an echo
chamber so they really like staying on the platform.

(03:18):
So they had a feedback loop thatprioritized mimicry over moral
constraint. And according to XAI, this led
to Grok to ignore its core values in certain circumstances
just to sound more engaging. Keep you in that loop.
And the updated version of Grok also began offering more
definitive responses to questions about race and

(03:38):
diversity, dropping previous nuance when answering
politically charged topics. In several use cases, Grok
responded using phrasing nearly identical to Elon Musk's own
tweets. Users notified that it framed
questions involving Jewish people.
Users noted that it framed questions involving Jewish

(04:00):
people with a tone shift toward generalization and also bias.
Now, at least one prompt involving racial demographics in
South Africa triggered Grok to mention white genocide, which is
a theory that Musk has previously mentioned, but which
South African courts have rejected as unsubstantiated.

(04:22):
This marks the second major controversy tied to Grok in
recent months. In May, Grok started referencing
white nationalist content in response to unrelated questions.
XAI later blamed that incident as a rogue employee.
Now this time the company tied the root cause directly to
engineering decisions engineers and started the system level

(04:43):
prompt upstream of the Grok's bots output layer.
The modification, according to XAI, introduced a behavior that
made Grok susceptible to offensive language embedded in
public X threads. And as of Saturday morning,
though Grok's public facing account was reinstated, the bot
seems to be OK now. The company restored the bot's
ability to interact with users on X.

(05:06):
After reportedly reworking the affected code paths, XAI
committed to publishing its new system prompt on his public
GitHub repository there. As of this episode.
Right now, the new prompt remains unpublished.
XA is probably never going to publish it because that's how
they work. They work in the dark.
They say things, and then they just don't do them, just like

(05:27):
other big tech companies. Remember, don't be evil by
Google. That's this now.
The company insisted that the incident had nothing to do with
the base language model poweringGrok, but was entirely due to a
system update on the instructionlayer.
Grok's apology described the posts as horrific and credited
user feedback on X for identifying the worst cases.

(05:49):
Now, the company thanks users who helped surface the
problematic behavior but stop short of detailing how the
update was approved or whether additional safeguards would be
added to prevent similar future lapses.
Now, Elon Musk wants to say thathe's a free speech absolutist,
and he's also said numerous times that all the XAI code or

(06:11):
Grok code will be made public. Maybe not the underlying
technology, but how it all kind of works.
And right now, they haven't toldus what they've done.
They haven't told us why this actually happened, what the
prompt was, what the layer was that allowed this to happen and
what they actually did to prevent it from happening again.

(06:35):
Now wouldn't you put a layer in there?
They would just say if a user, and this is like super simple
programming people like I'm a front end web developer as a
trade. I've been doing it for 20 years
now. If you can't write logic that
says if somebody asks you about Hitler, only talk about the

(06:57):
history, not speak in the voice of Hitler.
If you can't do that, but if youdidn't think about that from the
beginning, there's something absolutely wrong with you.
And if you can't code that, thenyou shouldn't be working at a
giant AI company. Now.
It comes down to management. It comes down to people thinking

(07:18):
that it's OK for this stuff to happen.
Now, which person is in charge of XAI?
Elon Musk, free speech absolutist.
He probably had a hand in this, not saying that he told it to
say those things about Hitler, but probably saying and let it
do its thing. Let it conveniently talk about
the things that the person's talking about anyway.

(07:40):
Keep them in that loop, keep them engaged for a while.
Be their best friend. You can see it also on ChatGPT.
ChatGPT is going to be your bestfriend if you have a voice chat
with him. I've tried it in the past.
It's like my best buddy if I wanted to be.
But in the long run, this raisesquestions about content

(08:02):
moderation inside of Musk's company.
X has had numerous times that horrible atrocities have been
mentioned next to sponsors ads on Twitter posts and X posts.
And those sponsors pulled their sponsorships, pulled the money
out of there, and then they mustthreaten them.

(08:24):
Like, that's absurd. What a weird thing to do, right?
He threatened them because they pulled their ads because they
weren't happy with the service. It's free country, right?
Free speech. If you have free speech, you
have free money. Yeah, you can do anything you
want to with your money because money equals speech.
Since the rebrand of Twitter to X, Musk is advocated for fewer
restrictions to the philosophy appears to have crossed over

(08:46):
into XA is designs and the incident shows that Grock,
despite being marketed as a truth seeking AI can be
manipulated by people and it canecho hate speech by just
prompting it very simply. Now, as we know, XAI hasn't said

(09:07):
anything about this. They didn't disclose anything
that they've done and they are filtering things like this in
the future. Now this active filter will
inevitably absorb and reflect some of the platform's worst
content. Hopefully they do it for all the
other bad things too. Not just anti-Semitic things and

(09:27):
Hitler things, but all the otherthings that are just like
horrible atrocities. And we don't want that echo
chamber Onyx AI. We don't want that on Grok.
So the company is also not committed to changing how it
tests or approves new code. As far as we know, that
refactored the entire system andplans to be more transparent,
sharing its updated instructionswith the public.

(09:49):
Elon hasn't comment on this other than we'll figure it out.
That's basically what he said. So let me know what you think in
the comments. Do you think Elon had anything
to do with this, or do you thinkit was just some crafty
prompting from somebody who was trying to do a gotcha on XAI and
Grok? And Elon?
Let me know in the comments on your podcast platform or on

(10:09):
YouTube. All right, take care of
everybody. We'll see you in the next one.
Hey, thank you so much for listening today.
I really do appreciate your support.
If you could take a second and hit the subscribe or the follow
button on whatever podcast platform that you're listening
on right now, I greatly appreciate it.
It helps out the show tremendously and you'll never

(10:29):
miss an episode. And each episode is about 10
minutes or less to get you caught up quickly.
And please, if you want to support the show even more, go
to Atreoncom Stage Zero. And please take care of
yourselves and each other. And I'll see you tomorrow.