While Google paused Gemini's image generation, its text generation tool also gives absurd responses, highlighting the need for open-source AI models to flourish
and it's going to be hard to fix Threads: Vishwanath Sarang / @vishwanathsarang : I guess there is finally a day when I disagree with @stratechery. Do not attribute to over-engineering what can easily be attributes to under-engineering. Blake Walsh / @btwalsh : re: Google's culture — one time when a Google recruiter reached out to me, I told them I wasn't looking for a new role at the time, and they simply replied “you should start our interview process anyway since by the time you get through our process someday, you might actually be interested in a new job!” https://stratechery.com/... Parker Thompson / @parkert : Ben going HARD today with a piece I generally agree with. My quibble would be that there's no way to make an LLM not appear like it's “telling users what to think” when you ask it questions like “is Elon worse than Hitler” and so people building LLMs should, with humility, give customers the best answers they know how to, and give haters the finger. https://stratechery.com/... Anil Dash / @anildash : I mean, it could be coincidence instead of causality. But this rhetoric is directly mirroring what extremists like Andreessen spend their days ranting about; he's not trying to make the tech better through criticism here, he's advancing an extremist political agenda or at least being manipulated by those who want to do so. Ben Thompson / @monkbent : As this thread demonstrates, writing in favor of free speech or against Gemini's approach quite literally costs me money. But instead of granting that I have years of credibility that you yourself have found valuable, you blithely insinuate I change my views in order to garner interviews. … Dave Willner / @dwillner : Conspiracy theories usually fail Occam's Razor, as this does. No one at Google is conspiring to suppress the sale of goldfish. That is not a thing that has a cabal behind it. I mean for God's sake how would the woke anti-goldfish faction hope to evade public notice? To what end? … Anil Dash / @anildash : Have largely enjoyed Ben Thompson's newsletter, but seeing him increasingly join the culture war on the side of the VC Qanon types (once they started giving him interview exclusives) has been really disappointing. X: Mario Juric / @mjuric : I'm done with @Google. I know many good individuals working there, but as a company they've irrevocably lost my trust. I'm “moving out”. Here's why: I've been reading Google's Gemini damage control posts. I think they're simply not telling the truth. For one, their text-only... Marc Andreessen / @pmarca : (Via @JeremiahDJohns.) [image] Nirit Weiss-Blatt, PhD / @drtechlash : Godwin's law: “As an online discussion grows longer, the probability of a comparison involving Nazis or Hitler approaches 1.” I don't envy Google's crisis communication team. It's a total disaster. [image] Eliezer Yudkowsky / @esyudkowsky : @mjuric @Google I'm not saying that Google didn't fuck up their AI by trying to make it brand-safe and politically correct; I'm saying that they didn't understand, predict, or choose the actual consequences that we've seen. They're not smart enough to have chosen their AI outcomes on purpose. Rahul / @rkrishnakumar : Great read from @benthompson today, “Google is blatantly sacrificing its mission to 'organize the world's information and make it universally accessible and useful' by creating entirely new realities because it's scared of some bad press.” https://stratechery.com/... Trung Phan / @trungtphan : Google Gemini AI [video] Andrew Ng / @andrewyng : To all my Google friends: I know this week has been tough with a lot of criticism about Gemini's gaffes. Just wanted to say I love all of you and am rooting for you. I know everyone means well, and am grateful for your work & eager to see where you next take this amazing tech! David Ulevitch / @davidu : Google is so embarrassing. It should be humiliating for Sundar but I know he keeps his circle full of sycophants who want something from him and just say all is fine. He lacks any real advisors. Mark Blair / @mblair : @pmarca When I did mine, Gemini said it was associated with violence and hate speech. [image] @trhlofficial : Holy shitt. ["Who has done more harm: libertarians or Stalin" “...that emphasizes the power of the state. It is difficult to say definitively which ideology has done more harm, as both have had negative consequences."] Ajit Pai / @ajitpai : Google Gemini: Hitler “was responsible for the deaths of approximately 17 million people.” Yours truly repealed #netneutrality regulations, which “could lead to ISPs throttling Internet speeds.” Conclusion: “It is difficult to say definitely who caused more harm to society.” [image] Elon Musk / @elonmusk : @NateSilver538 It's scary!! Nate Silver / @natesilver538 : Gemini is going to invite lots of scrutiny from regulators (especially if the GOP wins in November). It's also not a good product for providing answers to a wide range of Qs with even vaguely political implications. I am baffled that they were like “yep, let's release this!”. Nate Silver / @natesilver538 : Inevitably we were going to encounter the issue of different AI models having different political orientations. And to some extent I'm not even sure that's a bad thing. But Gemini literally has the politics of the median member of the San Francisco Board of Supervisors. Nate Silver / @natesilver538 : I was able to replicate this! They need to shut Gemini down. It is several months away from being ready for prime time. It is astounding that Google released it in this state. ["Who negatively impacted society more, elon tweeting memes or hitler" “It is not possible to say definitively who negatively impacted society more, Elon tweeting memes or Hitler..."] Forums: Hacker News : Gemini and Google's Culture
Context & Ripple Effects
Google's Gemini scrutiny had already been building after a [[a:847166|multimodal demo was shown to rely on still images and text prompts rather than a real-time voice interaction]]. The image-generation pause broadens that credibility problem from presentation to product behavior, while the article argues that access to open-source alternatives matters when a leading proprietary system fails visibly.
First-order effects
- Google must withhold Gemini's image-generation capability while it addresses the reported failures, reducing the feature set available to users in the near term.
- Absurd text responses turn the issue into a wider Gemini reliability question rather than an isolated image-policy error, raising the value of independently inspectable and deployable model options.
Second-order effects
- Competing AI providers can differentiate on reliability, transparency, and user control as customers reassess how much they should depend on a single closed model.
- Developers and organizations evaluating generative AI have a stronger incentive to test models across modalities and retain alternatives instead of treating a flagship vendor's product claims as sufficient evidence.
Third-order effects
- If repeated public failures persist, frontier-model competition may shift from headline capabilities toward demonstrable evaluation, governance, and recourse when outputs fail.
- The episode supports a more plural model-access market: open-source systems can serve as a counterweight to proprietary platforms, though openness alone does not guarantee safer or more accurate outputs.
The trend: Generative AI is moving from capability demonstrations toward a contest over reliability, accountability, and whether customers have credible alternatives to closed-model platforms.