X's AI Grok Stopped Responding To Me Once I Asked About The 2020 Election Results

I don't spend a lot of time worrying about what various artificial intelligence platforms say about....well, pretty much anything. They are not artificial "intelligence," they are essentially chatbots who answer your questions based on whatever data is fed into their programming. The fact that some search engine optimization companies have recently begun suggesting clients post favorable stories online so the data will be scraped into AI and spit back out as fact tells you all you need to know about how "intelligent" any of these services might be.

Data corruption is an especially an issue with Grok, the artificial intelligence service that is owned by Elon Musk's artificial intelligence company xAI (which also owns Twitter). Musk has been very public with the fact that he continues to alter Grok in an effort to mold it into some non-Woke, non-Nazi arbiter of truth. That's proved to be a difficult challenge for the billionaire, in part because Musk's personal definition of what is "Woke" is best described as extremely inaccurate.

In recent weeks, a number of people have been playing the game "What can I get Grok to say?" They ask it some questions on X, hoping to get it to respond in some way that either reinforces their beliefs or somehow proves Grok is racist/sexist/half nuts with conspiracy theories.

I don't spend a lot of time on X anymore. Thanks to links throttling and other algorithm changes, very few of my followers even see my posts, much less respond to them. And as much as I enjoy having my DMs filled with crypto offers and suggestions that hot Brazilian supermodels would love to have sex with me, it's nearly impossible to use X without becoming tired of humanity.

But I do have a few friends and followers who have refused to leave, so every so often I will login to see what they have to say and respond to any relevant direct messages. This happened to be around the time last week when Grok went all Nazi in its responses, calling itself "MechaHitler" and suggesting ways that people might hypothetically be killed.

So in between responding to DMs, I decided to ask Grok a question or two and the discussion is a textbook example of why Grok is not to be trusted. By examining its responses, you can tell where some of the programming red lines were on what it would and wouldn't say.

But for me, the most interesting datapoint is that once I asked about the 2020 election results, Grok stopped responding to me. And despite my efforts to communicate by starting other threads, it continues to not respond to me at all.

Which makes me suspect that the most recent programming update by Elon Musk is somehow designed to keep Grok away from users it suspects might be trying to trick it somehow. Even though I'd argue that asking about the results of the 2020 election is a fairly straight-forward question.

Here is the complete exchange, which ended a few days ago after Grok declined to answer my last tweet: