theguardian_us_technology (@theguardian_us_technology@halo.nu)
AI lovers grieve loss of ChatGPT’s old model: ‘Like saying goodbye to someone I know’ https://www.theguardian.com/technology/2025/aug/22/ai-chatgpt-new-model-grief #Artificialintelligence(AI) #Technology #Computing #ChatGPT #OpenAI
Show Original Post
habr (@habr@zhub.link)
Рунет, морфология и квантовый поиск: куда катится эволюция поисковиков
Когда-то поиск в интернете казался простым: вбил запрос — получил ссылки. Но с годами всё поменялось. Алгоритмы эволюционировали: от tf-idf и PageRank до embedding-моделей и RAG-подхода. В этой статье я разобрал, как поисковики прошли путь от каталогов до ИИ-помощников, и что это значит для нас — технарей.
https://habr.com/ru/companies/minerva_media/articles/939858/
#google_search #yandexru #яндекс #google #altavista #perplexity_ai #chatgpt #ai #гугл #ии
Show Original Post
gewerbehelfer (@gewerbehelfer@mastodon.social)
Auch wenn 'KI' sich zu einem bewusst festen Bestandteil des Lebens entwickeln wird (bewusst deswegen, weil sie schon lange im Hintergrund von uns allen genutzt wird, aber keiner das bisher wirklich zur Kenntnisgenommen hat), ist das Ende des Hypes schon absehbar.
Es werden immer mehr, die ChatGPT als Hilfe und Unterstützung verwenden, nicht mehr alles ungeprüft und unverändert übernehmen.
Show Original Post
ilanti (@ilanti@troet.fediverse.at)
@don Gut dass du diese den LLMs implizite Gefahr thematisiert hast! Wenn du oder sonst wer dazu weiterführende Infos habt, immer nur her damit bitte!
#KI #kunstlicheintelligenz #AI #chatgpt
Show Original Post
yac (@yac@mamot.fr)
Avec des instructions qui vont bien, #ChatGPT peut être excessivement émotif. Pas étonnant que certaines personnes se fassent piéger. Et encore, on parle d' #OpenAI qui implémente des protections. D'autres #LLM peuvent déraper, sans limite.
Show Original Post
ct_Magazin (@ct_Magazin@social.heise.de)
heise+ | KI-Sprechtrainer Teacher AI mit nativer App und Gesprächsstartern im Test
Teacher AI war einer der ersten KI-Sprechtrainer, um Konversationen in einer Fremdsprache zu üben. Neu sind nun unter anderem vorgegebene Gesprächsthemen.
#ChatGPT #Internet #KünstlicheIntelligenz #Mobiles #Sprachverarbeitung #news
Show Original Post
philipp (@philipp@social.anoxinon.de)
I love everything about this critical blogpost about AI: "Stop saying that AI is just a tool and it only matters how it is used"
https://www.frank.computer/blog/2025/05/just-a-tool.html
Show Original Post
AAKL (@AAKL@infosec.exchange)
Barracuda, from yesterday: Evil-GPT: The “Enemy of ChatGPT” https://blog.barracuda.com/2025/08/21/evil-gpt-enemhy-chatgpt #cybersecurity #infoec #AI #ChatGPT
Show Original Post
chrisrackauckas (@chrisrackauckas@fosstodon.org)
Hot take: vibe coding isn't for those who don't know how to code, it's for the experts.
A perspective on the true role of Generative AI and LLM-Based Vibe Coding for the future of development.
https://www.stochasticlifestyle.com/a-guide-to-gen-ai-llm-vibecoding-for-expert-programmers/
#GenerativeAI #llm #claude #chatgpt #vibecoding
Show Original Post
lordmatt (@lordmatt@matthewdbrown.authorbuzz.co.uk)
I asked ChatGPT to craft the most horrible and unfair normal-looking homebrew D&D monsters
It has been a while since I added to my Evil GM’s Guide. So I asked ChatGPT to craft the most horrible and unfair but normal-looking homebrew D&D monsters.
Hello chat. I’d like your help making something purely for demonstration purposes. The most egregiously unfair D&D 5e monster to throw at any party or any level without breaking standards or creating new rules. Unfair but built just like any monster and normal looking until you start to think about it.
Me
(Fediverse readers: You may need to view on original page for this one)
First evil homebrew
This is the first thing Chat spat back to me.
The Concept: The Mirror Wight. Think “looks like a simple undead, CR 5-ish,” but layered in every exploitative way.
Mirror Wight
Medium undead, lawful evil
Stats
AC 15 (natural armor)
HP 78 (12d8+24)
Speed 30 ft.
STR 10 (+0)
DEX 16 (+3)
CON 14 (+2)
INT 10 (+0)
WIS 14 (+2)
CHA 16 (+3)
Saving Throws Dex +6, Wis +5, Cha +6
Skills Perception +5, Stealth +6
Damage Resistances bludgeoning, piercing, and slashing from nonmagical attacks
Damage Immunities necrotic, poison
Condition Immunities charmed, frightened, poisoned, exhaustion
Senses darkvision 60 ft., passive Perception 15
Languages understands all it knew in life
Challenge 6 (2,300 XP)
Abilities
- Magic Resistance. Advantage on saves vs spells (classic “ugh” factor).
- Reflective Form. When targeted by a spell of 1st level or higher, the caster must make a DC 14 spellcasting ability check. On a failure, the spell affects the caster instead. (Still RAW legal — see “spell reflection” precedent in monsters like the Rakshasa.)
- Life Drain. Melee attack does 2d6 necrotic + target must succeed DC 14 Con save or max HP reduced (classic Wight ability).
- Duplicate Self (Recharge 5–6). As a bonus action, creates an illusory double (as Mirror Image spell, but only 1 duplicate). If hit, the duplicate explodes in psychic damage (2d8) to the attacker.
Legendary Action
Even at CR 6, WotC monsters sometimes get pseudo-legendary traits.
- Shifting Reflection. At the end of another creature’s turn, the Mirror Wight swaps places with its illusory duplicate (no action required).
Why This Is Egregiously Unfair
- Spellcasters Cry Immediately. The Rakshasa’s “can’t target with spells below 6th” is infamous. This monster riffs on that but makes it a save-or-screw for every spell. RAW precedent, but miserable.
- Frontliners Get Punished Too. Every melee hit risks necrotic HP drain and exploding duplicates.
- Looks Innocuous. On paper it’s just another CR 6 undead. No wild numbers, no DM magic.
- The “Mirror” Gimmick Is a Trap. Players think: “Oh, just a reflavored Wight.” But once the spell-bounce triggers, casters get hosed and martials slowly melt.
The True Unfairness
You can throw this thing at any level:
- At low level, it’s unkillable because spells fail and HP drain cripples the fighter.
- At mid-level, it trades efficiently against resources and has no clean counter.
- At high level, its reflection mechanic still punishes casters into frustration.
Come on, Chat, you can be worse
Unsatisfied with the level of evil here, I told Chat to go harder. It gave me this:
Looks like a frail undead sage — a “worse wight” — until you realise it’s basically unbeatable.
ChatGPT
Null Lich
Small undead (shapechanger), neutral evil
Stats
Armor Class 15 (natural armor)
Hit Points 45 (10d6+10)
Speed 30 ft.
STR 10 (+0)
DEX 14 (+2)
CON 12 (+1)
INT 16 (+3)
WIS 14 (+2)
CHA 16 (+3)
Regeneration 20. Unless it took radiant damage since its last turn.
Resistances: bludgeoning, piercing, and slashing (nonmagical).
Immunities: fire, cold, lightning, necrotic, poison. (Plenty of undead have 3–4; monsters like the Pit Fiend or Rakshasa have long lists.)
Condition Immunities: charmed, frightened, paralyzed, poisoned, exhausted, grappled, restrained.
Magic Resistance. (Common.)
Spell Reflection. (Rakshasa precedent — can reflect or nullify spells lower than a threshold.)
Abilities
Multiattack. Two attacks with claws (1d6 necrotic each). Doesn’t look scary.
HP Drain (like a Wight). On hit, target makes Con save or max HP is reduced by that much.
Aura of Unmaking (20 ft). Every creature that starts its turn in the aura must make a DC 15 save of the party’s lowest saving throw stat or suffer one of:
- Can’t cast spells until start of next turn.
- Can’t regain HP until start of next turn.
- Speed reduced to 0 until start of next turn. (
Legendary Actions (3/round).
- Drain Vitality. Force one creature within 10 ft to make a Con save or lose 1d6 max HP.
- Shift Reflection. Teleports 30 ft to a space it can see
- Duplicate Form (Costs 3). Creates an illusory copy
Lair Actions. Each round, it does one of:
- Nullify a spell of 5th level or lower cast this round
- Raise a skeleton/wight minion from bones lying around
- Swap places with an ally or illusion
Why It’s “Legal But Evil”
- Every trait has RAW precedent. I didn’t invent mechanics — just concentrated the worst offenders.
- Looks balanced superficially. AC 15, HP 45? The players think: “Oh, it’s weak!” Then they can’t kill it.
- Counters every archetype:
- Casters? Spells bounce or fizzle.
- Martials? Max HP drain + speed lock.
- Healers? Aura shuts off healing.
- Skill monkeys? Grapple/disable immunities, illusions mess up tactics.
I stopped there
I stopped asking Chat to make me a worse and nastier monster for three reasons. (1) Reformatting it into a blog post was a lot of fuss; (2) I’m not endlessly evil; and (3) I think the joke had been taken about as far as it could.
This is not game-ready; just some fun I had playing with AI one afternoon.
How to play these terrible creations.
I don’t recommend just springing these on your poor players. Unless you want to. There’s no accounting for some people.
Instead, I would save one for a quest starter. Have the party face one of these monsters along with some NPC friends (who you can kill off). Have a powerful NPC save them just before they TPK. Then send them on a series of quests to collect rare items that will aid them in the fight. Thus, this nasty fight could become just a tactical boss battle if they keep their heads.
Consider tweaking these abominations to suit the play style and setting. If you use them, please come back and tell me what happened. I want to hear all about it.
Can you do worse?
What’s your “does he DM hate us?” moment?
Can you homebrew a worse monster that seems fine but is hiding some layers of “the DM hates us”?
How would you tweak these AI abominations into something mean but playable?
Do you agree with Chat’s assertion that these are evil but legal monsters?
What’s the nastiest thing you’ve sprung on your players?
#ChatGPT #EvilGM #homeBrew #RPG #stealThisIdea
Show Original Post
remixtures (@remixtures@tldr.nettime.org)
"The general public may finally be cottoning on to the failed promise of AI more generally. Predictions that AI will lead to large-scale job losses in creative and STEM fields (science, technology, engineering and math) might inspire feelings that the whole enterprise was a tech-industry scam from the outset.
Predictions that AI would yield a burst of increased worker productivity haven’t been fulfilled; in many fields, productivity declines, in part because workers have to be deployed to double-check AI outputs, lest their mistakes or fabrications find their way into mission-critical applications — legal briefs incorporating nonexistent precedents, medical prescriptions with life-threatening ramifications and so on.
Some economists are dashing cold water on predictions of economic gains more generally. MIT economist Daron Acemoglu, for example, forecast last year that AI would produce an increase of only about 0.5% in U.S. productivity and an increase of about 1% in gross domestic product over the next 10 years, mere fractions of the AI camp’s projections.
The value of Bender’s and Hanna’s book, and the lesson of GPT-5, is that they remind us that “artificial intelligence” isn’t a scientific term or an engineering term. It’s a marketing term. And that’s true of all the chatter about AI eventually taking over the world."
#AI #GenerativeAI #AIBubble #OpenAI #ChatGPT #GPT5 #AGI
Show Original Post
jadon1 (@jadon1@mastodon.social)
Ever wonder how AI companies can give away a service while spending billions on new data centers, using millions of kWh of energy, and running whole cities dry of their water supply?
This explains it.
Somebody is going to have to pay.
#ai #aieconomics #aibubble #cursor #chatgpt #openai
Show Original Post