Hacker News Front Page
중도 성향
IT/기술
Meta Keeps Delaying the Release of Its New AI Model to Developers
Comments
IT/기술 · "EPS" · 총 63건
필터 보기현재 지수
50.3
0 = 부정 우세
50 = 중립
100 = 긍정 우세
최근 7일 기준 82,232건을 분석한 결과, 뉴스 심리지수는 50.2(균형)입니다. 긍정 4,044건(4.9%)·중립 76,245건(92.7%)·부정 1,943건(2.4%)이며, 중립 비중이 뚜렷하게 높습니다. 성향 지수는 종합 14.7(중도 균형)입니다.
Comments
Four former AIADMK ministers and ex MLAs joined ruling TVK in Chennai, latest in a series of defections deepening AIADMKs crisis after its April 23 election defeat
Het voorspellen van het WK voetbal van deze zomer zal voor veel fanatieke of juist minder betrokken voetbalfans niet de makkelijkste opgave zijn. Menig persoon die een voetbalpool invult, zal met moeite een winnaar kunnen aanwijzen bij Iran tegen Nieuw-Zeeland en Congo-Oezbekistan. Als je daarbij optelt dat er dit toernooi veel meer teams (48 in plaats van 32) meespelen en dus ook veel meer wedstrijden te voorspellen zijn, zal het invullen van een voetbalpool kunnen voelen als een heuse opgave. Dankzij de opkomst van AI-chatbots in de afgelopen jaren is het aannemelijk dat meer mensen de hulp zullen inroepen van een digitale assistent. Sterker nog: het is het eerste WK voetbal waarbij een AI-chatbot op grote schaal kan worden gebruikt voor voorspellingen. De eerste publieke versie van ChatGPT werd uitgebracht in november 2022, toen de groepsfase van het WK in Qatar nog bezig was. Sindsdien zijn er meerdere AI-modellen op de markt gekomen en is het gebruik van kunstmatige intelligentie gemeengoed geworden voor veel mensen. Westers gericht Een aandachtspunt daarbij is dat chatbots als Claude, ChatGPT en Gemini niet specifiek zijn getraind om voetbalwedstrijden te voorspellen, zegt Ruud Adriaans, hoofd AI bij accountantsbureau BDO. "Dat zijn modellen die heel breed ingezet kunnen worden. Ze hebben data van zo'n beetje het hele internet gebruikt." Maar meer data betekent niet automatisch een grotere kans op een goed antwoord bij het invullen van een WK-pool. "Als je een open vraag zou stellen aan een AI-bot, krijg je een soort gemene deler als antwoord." Concreet leidt dat ertoe dat de landen die als favoriet worden gezien, in het antwoord van de chatbot vaak netjes door de poulefase heen komen. De data waarmee de chatbot rekent hebben nog een effect, zegt Adriaans. "ChatGPT en veel andere westerse AI-modellen baseren zich onder meer op data van het internet en die zijn overwegend westers. Westerse nieuwssites zijn bijvoorbeeld meer vertegenwoordigd." Dat heeft weer impact op het antwoord dat je van een chatbot krijgt. Zo is de kans groter dat westerse en 'mediagenieke' landen overschat worden door AI, simpelweg omdat ze vaker voorkomen en meer worden opgehemeld in de aangeleverde data. Geen verrassingen Die vooringenomenheid is meteen ook de valkuil, omdat een AI-model je dus niet zomaar een verrassende uitkomst geeft. Marokko dat de halve finale van het WK haalt en daarbij Spanje en Portugal uitschakelt, zoals in 2022, zou een AI-bot niet hebben voorspeld. Daar komt bovenop dat een chatbot een compleet ander antwoord kan geven als je dezelfde vraag vijf minuten later stelt. Dat merkte ook Peter Neef, die de website Pouletips.nl beheert. Hij werkt al jaren met een rekenmodel dat zich baseert op data van wedkantoren en vergeleek dat dit jaar met de antwoorden van AI-chatbots. Die lopen nogal uiteen. ChatGPT en Gemini verwachten een 1-1 bij de groepsfasewedstrijd Nederland-Zweden, Claude denkt 2-1 voor Nederland en Neef zelf gaat uit van een 1-0 winst voor Oranje. De data van de wedkantoren zouden betrouwbaarder moeten zijn, omdat daar ook bijvoorbeeld WK-selecties, blessures en recente wedstrijden in worden meegenomen. Volgens Neef hebben de bookmakers een duidelijke voorkeur. "Frankrijk en Spanje, die hebben de grootste kans om wereldkampioen te worden. Maar die hebben net wel net allebei een oefenwedstrijd niet weten te winnen, dus dat moeten we echt nog maar zien." Hij wacht in elk geval nog even met het invullen van zijn pool. "Dat is een van de tips ook: vul alles zo laat mogelijk in. Want wie weet zijn er nog blessures of gebeuren er nog andere dingen." Onverwacht goed Kortom: AI een volledige voorspelling laten genereren is geen goed idee als je een pool wil winnen. Maar hulp vragen kan wel, als het maar zeer specifieke vragen zijn, zegt Adriaans van BDO. Je kunt bijvoorbeeld ermee onderzoeken welk van de Afrikaanse landen het de afgelopen periode onverwachts goed heeft gedaan. "AI kan je daar wel, mits je de vraag goed kadert, een prima antwoord op helpen vinden. Zolang je het maar met een bron kunt verifiëren." Adriaans omschrijft het als volgt. "Zie het qua betrouwbaarheid een beetje als je oom die voetbal volgt en ook wat heeft gegoogeld. Uiteindelijk moet je kiezen of je hem vertrouwt of toch liever zelf je research doet."
While the AI fundraising machine keeps breaking its own records, some founders are building in the other direction. Mirror founder Brynn Putnam just raised money for Board, a startup focused on bringing people together through in-person games and social experiences. Cyberdeck creators are going viral crafting whimsical DIY computers that literally encourage users to touch grass. Unlike the AI-free browser crowd, this doesn’t just feel like backlash, […]
Javier Bardem steps into the character once portrayed by Robert De Niro.
A research team that includes Huawei Technologies says it has successfully used the firm’s Ascend 910C chips to complete post-training for the DeepSeek-V4-Pro model, marking a major step forward as China’s semiconductor industry tries to leap from supporting basic AI inference to more complex model training amid tightening US sanctions. While Chinese chipmakers have found success in supporting AI inference – the relatively simple process of running an already-finished model to answer user...
“Why wouldn’t you want to be in both Pepsi and Coke?” says one venture capitalist. “It’s the same here.”
In the current environment, remaining heads down has diminishing returns; at some point, you have to make some noise just to remind the market you exist.
[OSEN=강필주 기자] 최근 전 세계 소셜 미디어(SNS)를 뜨겁게 달구고 있는 펩 과르디올라(56) 전 맨체스터 시티(맨시티) 감독이 나오는 펩시(Pepsi) 광고 영상이 기발한 전개와 감동적인 연출로 "역대급 명작 광고가 탄생했다"는 찬사를 받고 있다.하지만 놀랍게도 이 영상은 실제 펩시가 만든 공식 광고가 아니라 인공지능(AI)으로 제작된 '가짜 광
IRCTC has taken significant steps to combat ticket booking fraud by deactivating over three crore suspicious user IDs and placing another six crore under verification. To enhance food safety, the railway's catering arm has expanded its AI-based kitchen monitoring system, utilizing over 2,300 cameras to detect hygiene violations.
Byrne was arrested last Thursday while he was travelling from a hotel in Dublin to Dublin Airport.
Fianna Fáil TD Malcolm Byrne has said that he is stepping down as Chair of the Joint Oireachtas Committee on Artificial Intelligence.
The altered image was mocked by the candidate's opponents and spawned a flurry of memes. It has also been emulated by politicians around the country.
The Central Board of Secondary Education (CBSE) on Wednesday clarified that students applying for verification and re-evaluation of Class XII answer sheets do not need to have accounts with State Bank of India, Canara Bank, Bank of Baroda or Indian Bank to make payments on its online portal, addressing confusion that emerged after the system was launched earlier this week, Times of India reported.The clarification came after several students claimed on social media that the portal appeared to restrict payments to customers of the four public sector banks. In a statement posted on X, CBSE said the portal only uses payment gateways operated by these banks and does not require applicants to hold accounts with them.Also Read: Claude, other AI tools used to breach CBSE portals: IIT Panel“Candidates may use the available online payment options — UPI, net banking, credit card and debit card — through the designated gateways,” the board said.CBSE also said the portal continued to function smoothly despite a major cyberattack attempt on Tuesday, shortly after it went live. According to the board, the platform came under a barrage of denial-of-service attacks within minutes of its launch, receiving nearly 1.5 million hits in two minutes along with more than one lakh attempts at unauthorised file access.The board said its technical teams worked continuously to maintain the stability and security of the platform.“The portal has accepted 4,924 applications for verification and 39,056 applications for re-evaluation (total of 43,980) as of 12 noon today,” CBSE said.The board urged students to rely only on official CBSE communication for updates related to the process.Also Read: CBSE re-evaluation portal keeps lakhs of students guessingThe verification and re-evaluation window opened on June 2 for Class XII students who had earlier obtained scanned copies of their answer books evaluated under the board’s new digital On-Screen Marking (OMS) system.
'Todo Mundo em Pânico' entra no circuito na quinta-feira, dia 4, mas tem sessões antecipadas no dia 3 de junho. Divulgação Estreia nesta quinta-feira (4) nos cinemas brasileiros o sexto filme de "Todo Mundo em Pânico", um dos maiores fenômenos do terror-paródia dos anos 2000. Apesar de estar sendo chamado pelo público de "Todo Mundo em Pânico 6", o longa vem sendo divulgado apenas com o título original, sem o número estampado no pôster. O lançamento acontece 13 anos após o capítulo anterior e marca a volta dos irmãos Wayans após mais de duas décadas de afastamento. O hiato foi provocado por disputas de direitos autorais e divergências criativas com os antigos coprodutores da marca, os irmãos Bob e Harvey Weinstein — este último, condenado por uma série de crimes de agressão e assédio sexual em Hollywood. Marlon e Shawn se juntam a Anna Faris e Regina Hall, as eternas Cindy e Brenda, para remontar o quarteto original que deu o tom dos dois primeiros filmes da série. Justamente por tudo isso, a expectativa dos fãs não era pequena. E nem podia ser. A franquia tinha em mãos o elenco original e mais de uma década de material para trabalhar: novos filmes, novos debates, novos termos e, sobretudo, novas polêmicas. Pelo fim do 'mimimi' Os Wayans estão de volta a 'Todo Mundo em Pânico' para acabar com a cultura do cancelamento. Divulgação A premissa é a seguinte: o quarteto tenta escapar de um assassino mascarado também velho conhecido do público, o Ghostface, mas estabelecendo uma nova meta, "acabar com a cultura do cancelamento". O espectador, a essa altura do campeonato, já está cansado de saber: para assistir ao filme, não dá para se levar muito a sério. Afinal de contas, a própria franquia nunca se levou. E o novo filme faz questão de deixar isso bem claro, colocando, sem pudor, na boca de um dos personagens no início da trama: "Não é comédia com consciência social feita para branco pensar e ninguém rir. É para se divertir". É a partir desse posicionamento politicamente incorreto que os irmãos Wayans (que também assinam o roteiro) tentam desenhar uma provocativa disputa entre a velha guarda e a nova geração. Ao mirar nos dilemas geracionais, é como se os criadores — que agora já passaram da casa dos 50 anos — estivessem voltando aos holofotes para dizer: "Deixa a gente mostrar para vocês como é que se faz". Metralhadora de referências No elenco, Chris Elliott, Lochlyn Munro, Heidi Gardner, Damon Wayans Jr. e Savannah Lee Nassif, entre outros nomes. Divulgação Seguindo essa linha de ataque, o roteiro apresenta na tela um compiladão de quase tudo que bombou nas redes, nos cinemas e no noticiário nos últimos tempos. Sobra espaço para piadas envolvendo gays do Grindr, o Kanye West do Novo Testamento, a Covid-19, o ChatGPT, os relatórios de Jeffrey Epstein, a invasão do Capitólio norte-americano, a nova geração de streamers e por aí vai... Há ainda referências diretas a dezenas de outras produções, como "Wandinha", "Pecadores", "Guerreiras do K-Pop", a cinebiografia "Michael", "Saltburn", "Corra!" e mais. Muito mais. O grande problema, no entanto, é que quando tudo isso é colocado junto, em formato de uma sequência de esquetes de humor independentes, a engrenagem não dá liga. Já ouvimos isso antes 'Todo Mundo em Pânico' ironiza remakes, sequências, requels, prequels, spin-of's. Divulgação A estrutura fragmentada vai ficando bem batida à medida que o filme avança e, do meio para o final, as situações parecem sempre variações da mesma piada. Ao insistir na crítica à chamada "geração mimimi" ou, como a própria sinopse apresenta, "da cultura do cancelamento", o humor vai patinando em clichês que mais parecem uma reciclagem de milhares de outras piadas que o espectador já leu antes por aí, rolando a timeline do X, por exemplo. Tópicos como a "machosfera", questões raciais ligadas às cotas e o debate sobre pronomes neutros já foram excessivamente explorados por dezenas de outros produtos, formatos e comediantes nos últimos anos. O problema aqui, e é importante que se diga, não são os temas abordados. Mas a forma, pouco criativa (e quase nunca engraçada), com que são tratados. Um presente para os ex-viúvos Cena de 'Todo Mundo em Pânico', sexto filme da franquia que estreia nesta quinta-feira (4). Divulgação Curiosamente, os melhores momentos do filme acontecem justamente quando os atores deixam a fixação por essa "nova geração" de lado e passam a fazer piadas sobre eles mesmos e a criticar a própria indústria cinematográfica: Ironizando o Oscar, brincando com as escolhas de carreira que cada um do elenco fez no período em que estiveram afastados e expondo a própria batalha judicial que travaram nos bastidores para recuperar os direitos da marca. Felizmente, o desfecho do longa também consegue recuperar um pouco do fôlego ao entregar um final animador para os fãs, deixando evidente que os irmãos retomaram de fato o controle criativo da marca. É um belo presente para os, agora, ex-viúvos da franquia. Mas para conseguir esticar a história em uma eventual sequência sem cair no lugar-comum e no cansaço criativo que comprometem este sexto capítulo, apenas piadinhas sobre o mimimi da nova geração não vão colar. Cartela resenha crítica g1 Arte/g1
As artificial intelligence steps out of the digital realm and into the real world, the race to build the embodied “brains” powering next-generation robots has become the newest battleground in tech competition between China and the United States. Two days after US chip giant Nvidia launched its Cosmos 3 model – designed to help physical AI “think before it acts” – a Chinese start-up stole the spotlight. On Wednesday, Hangzhou, Zhejiang province-based Spirit AI said its foundation model for...
Researchers follow in Nightmare Eclipse’s footsteps, flipping off Redmond in favor of insta-leaks
From data colonialism to deepfakes, AI is reshaping Africa. A new study shows where Kenyan and South African coverage falls short, and offers practical steps to deepen and improve reporting.
The Hong Kong Generative AI Research and Development Centre (HKGAI) has officially launched a new DeepSeek-based large language model that can run on domestic chips, as the government-backed lab seeks to commercialise its products and export Chinese AI overseas. The HKGAI-V3 model, built on DeepSeek V4, has achieved “significant improvements” in efficiency and agentic capabilities, the centre said on Wednesday. The home-grown model delivered over tenfold improvement in the efficiency of token...
Chinese artificial intelligence start-up DeepSeek is finalising its first external fundraising round, securing over 50 billion yuan (US$7.4 billion) at a valuation of just under US$60 billion, according to people familiar with the matter – marking a six-fold leap from its US$10 billion valuation in April. The blockbuster round highlights intensifying global competition and a shifting strategy for the AI breakout star, which had previously resisted external capital. Market-oriented investors and...