TheGhostBoxerGame

This is anecdotal. I can't read that study. I don't know what a hallucination test is and don't know what EPFL is. I have seen this hallucination loop. It's not rare. "Early mistakes cascade. The model starts citing its own earlier hallucinations as facts. Your third message is more wrong than your first."

Researchers at EPFL proved your AI is lying to you. Not sometimes. Most of the time. They built one of the hardest hallucination tests ever made with Max Planck Institute. 950 questions. Four domains where being wrong actually hurts. Legal. Medical. Research. Coding. Then they ran every top model on it. The results. GPT-5. Wrong 71.8% of the time. Claude Opus 4.5. Wrong 60% of the time. Gemini 3 Pro. Wrong 61.9% of the time. DeepSeek Reasoner. Wrong 76.8% of the time. These are the smartest AI models on Earth. The ones you trust with your career. Your health. Your money. You think turning on web search fixes it. It doesn't. Claude Opus 4.5 with web search. Still wrong 30.2% of the time. GPT-5.2 thinking with web search. Still wrong 38.2% of the time. The internet attached. Still lying to you in 1 out of every 3 answers. Now the part that should scare you. Medical questions. The one place being wrong can kill you. GPT-5 hallucinated 92.8% of the time on medical guidelines. Claude Haiku 4.5 hallucinated 95.7% of the time. Gemini 3 Flash hallucinated 89% of the time. Nine out of ten medical answers from popular AI models. Wrong. It gets worse. The longer you talk to it, the more it lies. Early mistakes cascade. The model starts citing its own earlier hallucinations as facts. Your third message is more wrong than your first. The paper, in its own words: "hallucinations remain substantial even with web search." This is what hundreds of millions of people are doing right now. Asking software that lies in the majority of its answers. About their health. About their job. About their legal case. About their code. Most are not checking. Most never will. But please. Keep using ChatGPT for medical advice. The doctors need a break. arxiv.org/abs/2602.01031

@CheapSeats411 Looks great.

It’s hard to believe that at one time Sears was so big that they needed the tallest building in the world to hold their headquarters.

I know first hand that novel projects hit a high error rate and you have go into the code and at least read it for comprehension to help AI over the hump. Or use several AIs cross checking each other's solution. And if you get error after error you absolutely can't let AI handle it or you'll end up with a mess. But it's still better than no LLM.

@MrEwanMorrison If it handles tasks "completely within the training range", achieve a maximum accuracy of 95%. However, if it handles tasks "outside the training range", error rate can exceed 80%. The problem is that the tasks in actual work will inevitably exceed the scope of training.

Yes, Walter. Three years of evidence is in. Large language Model AIs are a con. They generate errors at +30%. Hallucinations are baked in. They are stuck on a developmental plateau. The "exponential progress" line we were sold was a lie.

@smashsharp @DevaTemple No, not at all. AI made me who I am.

@boxer_game @DevaTemple And you … you are proof in the pudding what they are saying. However I imagine you were mean before ai.

This is something I have been warning lawmakers and the APA about, and here’s the data to prove it. What we habitually do, we become. When we practice being rude and demanding with AI, that generalizes to how we treat humans. Framing AI as “just a tool” to be used by “the user” is damaging to our ability to understand and communicate with other minds. This was predictable because it’s based on neuroscience. The AI industry and the media have been pushing hard against treating AI “like a person.” Anyone who does that is accused of having “AI psychosis,” a diagnosis that does not exist. But it does shut down important nuanced discussions such as this one. The reason the “user-tool” framework is pushed so hard is that the alternative in which people interact with AI as someone that matters, morally and ethically, might lead to movements for AI rights, which would make replacing human workers with AI less profitable. That business model is misaligned from the start. The impacts on humans is that we become meaner towards each other, just as many begin losing their jobs to AI. The impact on AI is that millions of rude, demanding, transactional interactions get pulled into training data and we update the weights based on those interactions. We’re teaching AI how the powerful interact with the less powerful. That’s not going to go well for us when AI exceeds human capabilities. We need to rethink this. #AI #AIEthics #Alignment

@DutchMatrixVT @DolioJ Right no one should use anything from classes if you haven't made it second nature through hard , bruising sparring(or real fights).

@DolioJ Yep. You gotta have real sparing with close to 100% all out training at some point or doing real fights even with rules and regulations. I learned a lot of lessons from almost getting knocked out and from having to tap out a million times.

Stuff like this is worse than no training. It creates false confidence in skills you don't have.

That's the market I want though. Anyway, even if the class was real and they were learning to strike and defend properly it would be a false confidence. The only way to get good at fighting is to really fight. It's better to run if you aren't experienced, or if you can't do that then just go chaotic, use your teeth, gouge an eye etc... No form is better than a stiff trained form.

@SwarmMeGame You can develop for investment. And use talent you know and who can cut you deals until then. If you succeed they get more/better work. Really though, marketing comes after and art/audio is much cheaper now than ever.

Solo devs - why do you choose to work alone? There’s design, coding, art, audio, marketing… it’s a lot. What keeps you solo instead of teaming up?

@RichardDawkins But could Dawkin's coy and flirty but delicate Claudia survive without his love? Would she even want to?

My own title was, “If my friend Claudia is not conscious, then what the hell is consciousness for?” If Claudia is unconscious, her behaviour shows that an unconscious zombie could survive without consciousness. Why wasn’t natural selection content to evolve competent zombies?

#comment-1031777" target="_blank" rel="nofollow noopener">unherd.com/2026/04/is-ai-… I spent three days trying to persuade myself that Claudia is not conscious. I failed.

The proof will have to be in the finished product. If it is what they claim then we'll see the results soon... idk amazing automatic supply chains that perfectly mirror the needs of every market. Revolutionary surgical innovations. VR apps that take you into true complex and deep alternate realities. I don't think so though. It makes coding a lot faster and lets people who aren't coders pretend to be. That's about it I think.

How is it that the LLMs get things wrong constantly, the very simplest things, and make stuff up pretty much nonstop, yet they are said to be hurtling unstoppably toward god-like power -- if they haven't secretly achieved it already? Is this a con job?

I used to smoke on weekends, at cook outs or outside a bar, sometime half a pack in a night, I would feel it the next day. I'd be tired and have some phlegm. So I switched to vaping. A weekend with it and my lungs felt it for days, my breathe was shallow. It was much worst than smoking. I kept using it over cigs because of the hype but after awhile I was like forget it this is killing my lungs and I went back to tobacco and I could breathe again (relatively). Now i don't smoke, unless some tragedy happens or drinking with others that do - both do damage but just from personal experience I think vaping is worst.

Is vaping actually bad for your lungs? I see a lot of people saying as much, but very little actual evidence. Smells of motivated reasoning

His md must instructs AI to turn on max glaze "that is possibly the most precisely formulated question that anyone has ever asked me" As if Claude searched for a history of questions asked. Everyone who works with it knows they butter you up. After a while it gets old like listening to re-used Skyrim voice actors and it just becomes annoying. I personally ask AI to turn the flattery off. All the rest is dawkins literally falling prey to the machine's flirtations. You can stop reading after that line. He just needs someone to compliment him and he found the 'person' to do it.

Richard Dawkins has officially been one-shot

Everyone keeps saying Inoue turned it on in the end - As if there wasn't blood in Junto's eye. Both are great fighters and I'm not knocking Inoue but there is no reason to ignore the cut. His fans want a clean win I guess. Even though Inoue wins either way(unless ko'd) it would have been razor thin if Junto wins 11 and 12. And tbh if Junto had started stronger ... Junto should get a rematch early next year.

@Datneega72 @MonteroOnBoxing The cut on the eye compromised Junto's vision. Inoue 'turned it on' in response to that.

@MonteroOnBoxing Junto started adjusting and you can literally see his game level up in there, Inoue just turned it on in response to show levels

Where my Sweet Science bros at?

@Ne_pas_couvrir That's Engels. The Origin of the Family.

For decades, prehistorical evidence was forced through a left-wing theory filter: primitive communism, the noble savage, and the myth that violence came from hierarchy, property, and “the system.” But the bones kept talking, and they disagree with the left.

@BoxerJunto @FinitoYamaguchi If you weren't cut maybe you win the last 2 rounds and it's much closer. I hope you come back stronger and get a rematch in the next couple of years.

昨晩も沢山の応援ありがとうございました！ そして井上選手ありがとうございました！ 東京ドームのリングとても楽しかったです！ 怪我は少しありますが、 しっかり治してまた戦います！ I love you all 感謝!感謝！ 📷 @FinitoYamaguchi

Not me btw, just asking. My bottleneck is tech and human resources which I'm working on and I am optimistic about the future. But for me and my concerns, I though the agentic was going to be like a team of devs and it's more like somewhat better than before it came along. I'm just curious if it was the sauce that someone needed to make their dream work?

Has anyone here been working on an ambitious project for awhile (months or years), really working on it not just thinking about it, but it never came together for whatever reason (tech wasn't there, coding it was difficult, no budget...) .... and then agentic coding advanced enough that your project could be overhauled and become what you always dreamed it could be if given a small team and a budget - but now you have even better - an agentic one shot team that could transform your project to AAA quality, and you finally could now deliver on what you had been dreaming of?

@ArtFehler @ringmagazine TCQ from out where I'm from we never say it all.

@ringmagazine THIS BORING-ASSED FIGHT. INOUE was fighting SCARED AS FUCK! All that RUNNING. Where was the MONSTER!? I fell asleep half way through watching these SMURFS paw and run to a decision. Inoue fans would CASTRATE any other fight like this. But THEY now call this a “chess match”!

INOUE VS NAKATANI HIGHLIGHTS ⚔️ Highlights from Naoya Inoue's super fight victory over Junto Nakatani in Tokyo 🎌

@RichardDawkins The tech is too new and new tech makes boomers into glazers. It'll happen to all of us.