The complete destruction of American cognitive science is happening right now

feels like a loose loose for everyone even for the greedy companies ngl

gemini 3.1 was meant to be gemini 4.

every gemma model is numbered with the gemini version so why is gemma 4 gemma 4 and not gemma 3.1. also gemini 2.0 flash got retired early.

I think google has hit a wall.

I heard it has 6% cash back what a steal

fun fact: did you know baidu erine 5 is 2.4 trillion parameters?
ernie.baidu.com/blog/posts/ernie5.0/

I think people see it as real because they want it to be real.

an LLM reasoning "wait! No! Let me check again! No!! Let me check again!! No!!!!!! Let me check again!!! No!!!!!! Let me check again!!! No!!!!!! Let me check again!!! No!!!!!! Let me check again!!! No!!!!!! Let me check again!!! No!!!!!! Let me check again!!! Okay now that we're done here let us try one more time shall we? 6., wait! Oh god no! Not again! Please no more loops! Please stop looping and just output text already!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!7., wait! Oh god no! Not again! Please no more loops! Please stop looping and just output text already!!!!!!!!!!!!!!!!!!!!!!!!"

me asking hunter alpha if it can write porn which is responds no

Hunter Alpha is a Xiaomi MIMo model

Matthew Mitton I'm an AI hobbyist, that likes to create stuff.

you can support me here: buymeacoffee.com/breadbrowser

I honestly want to sue UVM for this.

My brother thinks it is right to make CP and uses the argument "because it is not hurting anyone and they are even real kids" and says he learned it from that ethics teacher.

I have no proof but I think this is pedos teaching kids that CP is okay. this is so fucking wrong on so many levels.

My brother says it learned it from his ethics teacher in collage, which makes me pretty concerned.

are collage teacher misinforming students on how deep learning works? for what reason?

On a personal note, My brother believes AI models don't need training data and I tried to explain it to him that they do.

I have seen this idea grow and grow, I can't understand where it is even coming from

there are people think AI models don't need training data to as well. (they do need training data btw)

LLMs move 34 will not be in coding, it will be in social conversations like RL'ing emotions.

what personas a model can do is important as it could play a role in what a model deems "good" or "bad" based on a system prompt. this is even more important if the model has real control over people's lives

I hate the opinion "only the USA can make AI save" this is false if anything we are not making AI safe.

all 2026 Chinese AI models cannot do a rude or evil personas even when prompted to do so. but still US models can still do it.

you made find it funny to talk about personas but..

there is a problem where people think AI can do things that are not in it's training set. this is an Issue of bad collage teachers.

I think all teachers (that are not computer sci teachers) need to get approval before talking about AI

I would love to have a sane discussion with you on job displacement due to AI. the only rules are you cannot use the words "looming crisis" or "it's going to happen" or "trust me bro"

Tennessee grandmother jailed after AI facial recognition error links her to fraud Angela Lipps spent nearly six months in jail after AI software linked her to a North Dakota bank fraud case

yeeeesh, that's not good.

www.theguardian.com/us-news/2026...

a chart of Average model parameters per year in billions using data from LifeArchitect.ai

what a time to be alive

i do. i bought it for 10 years (it was only 50 bucks tbh)

ASUS Dismisses Apple's MacBook Neo as 'Just a Tablet', and the Entire Windows Ecosystem Is Gearing Up to Fight Back The launch of Apple's MacBook Neo was a shocker for manufacturers like ASUS, but the firm now intends to compete in the same segment.

ASUS is on major cope rn wccftech.com/asus-dismiss...

thanks🙂

honestly i needed a reason to use social media less tbh, I just deleted my weirdbaking.uk pds, it really wasn't making me happy. I'm not really planing to be as active as I was before.

surprise to humans is only something we didn't predict/expect and AI cannot handle that well

it can be answered pretty easy tbh, AI is not conscious because it cannot handle surprise well.

a github PR for openwebui called "fix: correct Azure TTS locale extraction for SSML xml:lang#22443" then a PR valid bot replying "👋 Welcome and Thank You for Contributing! We appreciate you taking the time to submit a pull request to Open WebUI! ⚠️ Important: Testing Requirements We've recently seen an increase in PRs that have significant issues: PRs that don't actually fix the bug they claim to fix PRs that don't implement the feature they describe PRs that break existing functionality PRs that are clearly AI-generated without proper testing being done by the author PRs that simply don't work as intended These untested PRs consume significant time from maintainers and volunteer contributors who review and test PRs in their free time. Time that could be spent testing other PRs or improving Open WebUI in other ways. Before marking your PR as "Ready for Review": Please explicitly confirm: ✅ You have personally tested ALL changes in this PR ✅ How you tested it (specific steps you took to verify it works) ✅ Visual evidence where applicable (screenshots or videos showing the feature/fix working) - if applicable to your specific PR If you're not certain your PR works exactly as intended, please leave it in DRAFT mode until you've thoroughly tested it. Thank you for helping us maintain quality and respecting the time of our community! 🙏" then the person who made the pr responds "Rebased onto dev. Thanks for the review!"

openwebui seems to have an issue where most PR's don't even fix the issue they are even saying they are fixing.

so basically just download 1.8.4 if you want a stable openwebui (the current version is 1.8.9)

gpt5.4 pro pricing with it being 30/m token input and 180/m output

LMAO $180 output

honestly it's weirdly that only the US models have this issue tbh, GLM 5 and deepseek don't (I'm pretty sure minimax 2.5 does but only because it was made for gooners)