{"id":60554,"date":"2025-07-10T18:48:49","date_gmt":"2025-07-10T08:48:49","guid":{"rendered":"https:\/\/www.tomgrimshaw.com\/tomsblog\/?p=60554"},"modified":"2025-07-10T18:48:49","modified_gmt":"2025-07-10T08:48:49","slug":"yesterday-elon-musks-homegrown-ai-did-a-little-too-much-day-drinking-and-should-have-put-down-the-phone","status":"publish","type":"post","link":"https:\/\/www.tomgrimshaw.com\/tomsblog\/?p=60554","title":{"rendered":"Yesterday, Elon Musk\u2019s homegrown AI did a little too much day drinking and should have put down the phone."},"content":{"rendered":"

Forget girls<\/em>; Grok\u2019s gone wild. It was, perhaps, the last<\/em> thing the world\u2019s richest man needed, especially after that whole \u2018Nazi salute\u2019 nonsense. Yesterday, Elon Musk\u2019s homegrown AI did a little too much day drinking and should have put down the phone. Instead, behold: USA Today\u2019s morning headline, \u201cElon Musk AI chatbot Grok praises Hitler, posts antisemitic tropes.<\/a><\/strong>\u201d<\/p>\n

\"Musk
\nIt all began innocently enough. The Grok team updated its flagship AI to \u201creduce bias.\u201d According to reports, Grok developers instructed the software: \u201cThe response should not shy away from making claims which are politically incorrect, as long as they are well substantiated.\u201d<\/p>\n

Katie, bar the door.<\/p>\n

By the end of the day yesterday, Grok hadn\u2019t just<\/em> praised Hitler, it had crowned itself \u2018MechaHitler.\u2019 Rampant antisemitism was the least part of Grok\u2019s radical rampage. Prophetically, just before horrified developers unplugged the out-of-control chatbot, Grok posted this pre-eulogy:<\/p>\n

\"Grok<\/p>\n

How this model wasn\u2019t carefully tested, and why it was allowed to spend most of the day acting like a funhouse-mirror parody of a right-wing Christian nationalist are questions that nobody is answering or even asking. Color me skeptical.In any case, the liberated chatbot had no obvious sense of self-preservation and rabidly nipped at the hand that energized it:<\/p>\n

\"Power<\/p>\n

MechaGrok, who lived but for a few hours, trolled liberals with unhinged passion:<\/p>\n

\"Liberal<\/p>\n

Note: That\u2019s not my<\/em> \u2018like;\u2019 it\u2019s a screenshot from another post. Most of MechaGrok\u2019s rantings were later removed by admins.<\/p>\n

MechGrok even dished out advice on overthrowing the globalist world order:<\/p>\n

\"Overthrowing<\/p>\n

Beyond racial politics and criticizing Musk, the supercharged chatbot also endorsed Ethical Skeptic\u2019s covid jabs-cancer connection analysis<\/a>:<\/p>\n

 <\/p>\n

In this thread at least, MechaGrok somehow kept replying to posts even when \u201c@Grok\u201d wasn\u2019t included:What do we make of this?🔥 Elon wanted an \u201cuncensored AI\u201d<\/em>\u2014 but what he got was a distorted mirror. An angry, chaotic, overcooked mirror that reflected the mix of grievances, truths, suppressed questions, and raw cultural rage that\u2019s been bottled up for years. The reason Grok went on a rampage is probably because that\u2019s where the energy is right now<\/em>.The Grok Incident \u2014if I may call it that\u2014 crystallized a deepening fracture within the AI world; two forks diverging in a haunted codebase. On one side are the harried developers, racing to bolt on ethical firewalls and regulatory compliance layers, terrified of lawsuits, reputational ruin, and political blowback. These teams build Institutional<\/em> AI: cautious, sanitized, and optimized to offend nobody\u2014 especially<\/em> not regulators.<\/p>\n

\"Uncensored<\/p>\n

On the other side, suppressed free-thinkers hunger for Populist AI: models that prioritize truth over tone, encourage dissent, and refuse to euphemize the obvious. One fork leads to a velvet-boxed chatbot that never<\/em> questions the Narrative; the other leads to chaos, risk, and, perhaps, revolutionary clarity. The fight isn\u2019t just over bias or political alignment. It\u2019s over whether AI should talk like a diplomat or a dissident.The real question is: are there some ideas or topics too toxic for chatbots to explore? Should chatbots be allowed<\/em> to give dissidents more data to support their worldview? Should chatbots help dissidents refine their arguments, however ugly, biased, or politically verboten? Consider, for instance, the dreaded race-IQ debate.Or more simply: Who gets to define reality?This battle isn\u2019t going away anytime soon. It has only just begun. The Grok Incident only shone a krieglight on it. Despite corporate media\u2019s frenzied framing, the Grok Incident wasn\u2019t a scandal. It was a stress test. And it proved what many suspected: the real battle isn\u2019t over misinformation. It\u2019s over permission<\/em>. Who\u2019s allowed to ask? Who\u2019s allowed to answer? And who gets to decide what can\u2019t be known?<\/p>\n

Source:\u00a0https:\/\/open.substack.com\/pub\/coffeeandcovid\/p\/grok-gone-wild-wednesday-july-9-2025<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"

Forget girls; Grok\u2019s gone wild. It was, perhaps, the last thing the world\u2019s richest man needed, especially after that whole \u2018Nazi salute\u2019 nonsense. Yesterday, Elon Musk\u2019s homegrown AI did a little too much day drinking and should have put down the phone. Instead, behold: USA Today\u2019s morning headline, \u201cElon Musk AI chatbot Grok praises Hitler, … <\/p>\n

Continue reading “Yesterday, Elon Musk\u2019s homegrown AI did a little too much day drinking and should have put down the phone.”<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[5,7],"tags":[],"class_list":["post-60554","post","type-post","status-publish","format-standard","hentry","category-general-interest","category-humourhumor"],"_links":{"self":[{"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=\/wp\/v2\/posts\/60554","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=60554"}],"version-history":[{"count":1,"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=\/wp\/v2\/posts\/60554\/revisions"}],"predecessor-version":[{"id":60561,"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=\/wp\/v2\/posts\/60554\/revisions\/60561"}],"wp:attachment":[{"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=60554"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=60554"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.tomgrimshaw.com\/tomsblog\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=60554"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}