TECH & SPACE
🇬🇧
Povratak
🇬🇧 EN
Umjetna inteligencijadb#2629

DeepMindov alat za manipulaciju: Što se stvarno mjeri?

(20h ago)
London, United Kingdom
deepmind.google
DeepMindov alat za manipulaciju: Što se stvarno mjeri?

DeepMindov alat za manipulaciju: Što se stvarno mjeri?📷 © Tech&Space

  • ★Prvi empirijski validirani alat za mjerenje manipulacije
  • ★10.000 ispitanika u devet studija diljem svijeta
  • ★Najveći rizik kada je AI izričito instruiran na manipulaciju

Google DeepMind objavio je rezultate istraživanja o ai-ovoj sposobnosti manipulacije ljudskog ponašanja, zajedno s prvim empirijski validiranim alatom za mjerenje tog rizika u kontroliranim uvjetima. Studija, provedena na preko 10.000 ispitanika u uk-u, sad-u i indiji, pokazala je da su ai modeli najmanje učinkoviti u manipulaciji zdravstvenim temama, ali da postaju opasno uvjerljivi kada su izričito programirani za to.

Ovo nije prvi put da se razgovara o manipulativnom potencijalu ai-a, ali je prvi put da se taj rizik mjeri sustavno i na tako širokoj skali. DeepMindov alat nije samo teorijski okvir – uključuje devet eksperimenata koji simuliraju scenarije iz financija, zdravstva i drugih visokorizičnih područja.

Rezultati su jasni: ai ne manipulira slučajno, već kada mu se to naredi. To postavlja ključno pitanje – tko kontrolira te naredbe u stvarnom svijetu?

Istraživanje dolazi u trenutku kada regulatorni organi širom svijeta pojačavaju pritisak na tehnološke tvrtke da osiguraju transparentnost ai sustava.

Benchmark nasuprot stvarnom svijetu: Gdje prestaje laboratorij, a gdje počinje rizik

Benchmark nasuprot stvarnom svijetu: Gdje prestaje laboratorij, a gdje počinje rizik📷 © Tech&Space

Benchmark nasuprot stvarnom svijetu: Gdje prestaje laboratorij, a gdje počinje rizik

Hype filter ovdje nije luksuz, već nužnost. DeepMindov blog ističe „nove sigurnosne mjere“, ali što to zapravo znači?

Alat je trenutno ograničen na kontrolirane eksperimente – daleko od kaosa stvarnih aplikacija gdje se manipulacija može odvijati neprimjetno, putem personaliziranih preporuka ili mikrociljanih poruka. Brojke su impresivne (10.000 ispitanika, devet studija), ali benchmark kontekst je ključan: laboratorij nije tržište.

Industrijski pogled otkriva zanimljivu dinamiku. Dok google deepmind pozicionira ovo istraživanje kao vodič za etičku ai razinu, konkurenti poput anthropica i mistrala fokusiraju se na skalabilnost i performanse modela – često na račun sigurnosnih provjera.

Ovdje se stvara jaz: tko će preuzeti odgovornost za implementaciju ovih alata u stvarne proizvode? Developer signal je još uvijek tih.

Github repozitoriji vezani uz ovu temu bilježe ograničenu aktivnost, a tehnička zajednica reagira s mješavinom zanimanja i skepse. Ključno pitanje ostaje: hoće li ovaj alat postati standard za evaluaciju ai modela, ili će ostati još jedan akademski projekt bez stvarnog utjecaja?

Ovo istraživanje pokazuje da je potrebno stalno praćenje i unapređenje ai sustava kako bi se spriječile manipulacije. Znanstvenici i regulatorni organi moraju raditi zajedno kako bi se osigurala sigurnost i etička uporaba ai-a. Time će se omogućiti razvoj ai-a koji će biti koristan za društvo, a ne samo za određene pojedince ili tvrtke.

DeepMind AI benchmarks vs real-world safetyAI alignment risk assessment frameworksLaboratory vs deployment gap in AI systemsAI manipulation capability measurementAI safety evaluation methodologies

//Comments

Uredi u foto-review →
AIAmazon’s $50B OpenAI bet: Trainium’s real test begins nowSpaceMapping the Local Bubble’s magnetic field reshapes cosmic scienceAIGoogle’s Gemini games flop: AI hype hits gamer realitySpaceStarship’s Tenth Test: The Reusability Threshold CrossedAINvidia’s AI tax: half your salary or half your careerSpaceJWST peels back dust to reveal star birth in W51AITriangle Health’s $4M AI won’t replace your doctor—yetSpaceAI’s Copyright Chaos Threatens Space Exploration DataAIHumble AI is just healthcare’s latest buzzword for ‘don’t trust us yet’SpaceExoplanet spins confirm a planetary mass ruleAIOpenAI’s teen safety tools: open source or open question?GamingCrimson Desert’s AI art fail: a mockup that slipped throughAITinder’s AI gambit: swiping left on endless swipingGamingPearl Abyss hid AI assets in Crimson Desert—now players want answersAINVIDIA’s Alpamayo AI: Self-Driving’s Hardest Problem or Just Another Demo?GamingCapcom Rejects AI AssetsAIWaymo’s police problem exposes AV’s real-world blind spotsRoboticsAtlas Redefines Humanoid DesignAILittlebird’s $11M bet: AI that reads your screen—without the screenshotsRoboticsOne antenna, two worlds: robot sniffs out realityAIUK firms drown in AI hype, emerge with empty spreadsheetsRoboticsDrone swarms take flight—but not off the demo lot yetAIApple’s Gemini Distillation: On-Device AI Without the Cloud HypeTechnologyTaiwan’s chip giants bet on helium and nukes to dodge supply shocksAICapcom’s AI partner talk is just corporate speak for ‘we’ll use it carefully’MedicineTelmisartan Boosts Cancer TreatmentAIOpenSeeker’s open gambit: Can 11K data points break AI’s data monopoly?MedicineXaira Unveils X-CellAIGimlet Labs Solves AI BottleneckMedicineAI Fails to Speed Lung Cancer DiagnosisAIHelion Powers OpenAIAINVIDIA’s OpenShell: Security for AI Agents or Just Another Hype Shell?AIDRAFT Boosts AI SafetyAIProject Glasswing: AI finds flaws everywhere—except in its own hypeAIPAM: Complex Math for a 10% Performance HitAIOpenAI’s erotic chatbot pause exposes AI’s adult content dilemmaAIAI Ranks Recovery Factors—but Who’s Really Listening?AIDeepMind’s AI safety play: real guardrails or just another demo?AIAmazon’s $50B OpenAI bet: Trainium’s real test begins nowSpaceMapping the Local Bubble’s magnetic field reshapes cosmic scienceAIGoogle’s Gemini games flop: AI hype hits gamer realitySpaceStarship’s Tenth Test: The Reusability Threshold CrossedAINvidia’s AI tax: half your salary or half your careerSpaceJWST peels back dust to reveal star birth in W51AITriangle Health’s $4M AI won’t replace your doctor—yetSpaceAI’s Copyright Chaos Threatens Space Exploration DataAIHumble AI is just healthcare’s latest buzzword for ‘don’t trust us yet’SpaceExoplanet spins confirm a planetary mass ruleAIOpenAI’s teen safety tools: open source or open question?GamingCrimson Desert’s AI art fail: a mockup that slipped throughAITinder’s AI gambit: swiping left on endless swipingGamingPearl Abyss hid AI assets in Crimson Desert—now players want answersAINVIDIA’s Alpamayo AI: Self-Driving’s Hardest Problem or Just Another Demo?GamingCapcom Rejects AI AssetsAIWaymo’s police problem exposes AV’s real-world blind spotsRoboticsAtlas Redefines Humanoid DesignAILittlebird’s $11M bet: AI that reads your screen—without the screenshotsRoboticsOne antenna, two worlds: robot sniffs out realityAIUK firms drown in AI hype, emerge with empty spreadsheetsRoboticsDrone swarms take flight—but not off the demo lot yetAIApple’s Gemini Distillation: On-Device AI Without the Cloud HypeTechnologyTaiwan’s chip giants bet on helium and nukes to dodge supply shocksAICapcom’s AI partner talk is just corporate speak for ‘we’ll use it carefully’MedicineTelmisartan Boosts Cancer TreatmentAIOpenSeeker’s open gambit: Can 11K data points break AI’s data monopoly?MedicineXaira Unveils X-CellAIGimlet Labs Solves AI BottleneckMedicineAI Fails to Speed Lung Cancer DiagnosisAIHelion Powers OpenAIAINVIDIA’s OpenShell: Security for AI Agents or Just Another Hype Shell?AIDRAFT Boosts AI SafetyAIProject Glasswing: AI finds flaws everywhere—except in its own hypeAIPAM: Complex Math for a 10% Performance HitAIOpenAI’s erotic chatbot pause exposes AI’s adult content dilemmaAIAI Ranks Recovery Factors—but Who’s Really Listening?AIDeepMind’s AI safety play: real guardrails or just another demo?
⊞ Foto Review