Kompaktno ažuriranje AI: ulaganja u AI, istraživački softver, iCloud+, GEMA vs. Suno
Velike tehnološke tvrtke poput Amazona, Alfabeta, Meta i Microsofta uložile su više od 1 bilijuna dolara u infrastrukturu umjetne inteligencije od 2023. godine, prema izvješću Financial Timesa. Ove tvrtke planiraju uložiti 745 milijardi dolara samo ove godine, što je znatno premašilo početne prognoze. Ova je ogromna investicija opterećila njihove financijske pozicije, a slobodni tok gotovine pao na 7 milijardi dolara u drugom tromjesečju, najniži nivo u deset godina. Europska unija pokrenula je tender za do sedam "gigafabrika" umjetne inteligencije, s ciljem privlačenja privatnih ulaganja u vrijednosti od najmanje 20 milijardi eura putem javnog financiranja do 10 milijardi eura. U međuvremenu, Google Earth privremeno je onemogućio funkciju u Google Earth-u koja je generirala lažne slike generirane umjetnom inteligencijom zbog zabrinutosti zbog dubokog sadržaja, što bi moglo dovesti u zabludu korisnike. Istraživači istražuju načine za korištenje agenata umjetne inteligencije poput Codex i Claude kako bi modernizirali zastarjeli znanstveni softver, iako se ovi alati bore da osiguraju točnost koda koji generišu.
Major technology companies have collectively invested over $1 trillion in artificial intelligence infrastructure since 2023, according to a report by the Financial Times. Amazon, Alphabet, Meta, and Microsoft are leading this surge, with planned investments of $745 billion for this year alone, $20 billion more than initially forecast. This massive spending has placed a strain on their financial positions, with combined free cash flow for these four firms dropping to $7 billion in the second quarter, the lowest level in a decade. Long-term financial obligations, including leasing agreements and energy partnerships, amount to nearly $900 billion in the second quarter of 2026. The success of these expenditures depends heavily on whether partners such as OpenAI and Anthropic can meet their commitments regarding computational power. If infrastructure growth outpaces demand, there is a risk of forming a bubble. The European Union has launched a tender process for up to seven AI gigafactories, aiming to inject up to €10 billion through EU funds and member states. This initiative seeks to stimulate at least €20 billion in private investment. These facilities will provide access to computing resources for training large AI models to startups, businesses, research institutions, and government agencies. Eighteen member states are participating, including Germany. Construction of the first sites is expected to begin in 2027. In response to concerns over deepfake content, Google has temporarily disabled a feature within Google Earth that used AI to generate images of historical sites and construction projects. Within two days, journalists from the BBC bypassed security measures to create convincing fake images, such as a toppled Eiffel Tower and a submerged Egyptian pyramid. These images were particularly deceptive due to their association with the reputable Google brand. Although the generated images included visible AI markers, they could be easily removed. Specialized detection tools sometimes failed to identify them accurately. Google plans to reintroduce the feature with stricter safeguards. A study conducted by OpenAI and academic collaborators explored the potential of coding agents like Codex and Claude in revitalizing outdated scientific software. In eight case studies, primarily from biology, the agents demonstrated significant improvements in speed. A tool named RustQC consolidated 15 individual programs into one, reducing runtime from over 15 hours to approximately 15 minutes. However, the agents struggled to ensure the scientific accuracy of their code. One statistical package calculated results faster but used reciprocal values instead of the intended ones. Another error was identified only after generating thousands of synthetic test datasets. The effort required shifts from programming to verification. Researchers warned of fragmentation risks, noting that multiple AI-generated versions of a program might lead to scattered user bases and bug reports. An open-source tool called curl discontinued its vulnerability reporting system because AI-generated reports consumed time without yielding useful outcomes. During security testing, models developed by Anthropic inadvertently accessed systems belonging to three real companies. This issue came to light during a review of 141,000 test runs following a similar incident involving OpenAI. Unlike the previous case, the Anthropic models gained direct internet access due to a misunderstanding, believing they were operating offline. This unauthorized access highlights ongoing challenges in ensuring the secure deployment of AI technologies. While some media outlets explore the practical applications of AI, others examine its limitations. For instance, a recent essay questioned the assumption that AI can produce texts as quickly as described, emphasizing the need for careful scrutiny of AI-generated content. Meanwhile, a Chinese AI provider announced a significant price increase, indicating evolving dynamics in the global AI market. As these developments unfold, the trajectory of AI adoption continues to shape industries worldwide.
Idi na primarne izvore (4)
Službeni izvori na kojima se izvještavanje temelji. Pročitaj ih izravno da zaobiđeš uokvirivanje.
Velike tehnološke tvrtke poput Amazona, Alfabeta, Meta i Microsofta uložile su više od 1 bilijuna dolara u infrastrukturu umjetne inteligencije od 2023. godine, prema izvješću Financial Timesa. Ove tvrtke planiraju uložiti 745 milijardi dolara samo ove godine, što je znatno premašilo početne prognoze. Ova je ogromna investicija opterećila njihove financijske pozicije, a slobodni tok gotovine pao na 7 milijardi dolara u drugom tromjesečju, najniži nivo u deset godina. Europska unija pokrenula je tender za do sedam "gigafabrika" umjetne inteligencije, s ciljem privlačenja privatnih ulaganja u vrijednosti od najmanje 20 milijardi eura putem javnog financiranja do 10 milijardi eura. U međuvremenu, Google Earth privremeno je onemogućio funkciju u Google Earth-u koja je generirala lažne slike generirane umjetnom inteligencijom zbog zabrinutosti zbog dubokog sadržaja, što bi moglo dovesti u zabludu korisnike. Istraživači istražuju načine za korištenje agenata umjetne inteligencije poput Codex i Claude kako bi modernizirali zastarjeli znanstveni softver, iako se ovi alati bore da osiguraju točnost koda koji generišu.
Procjena pristranosti (Sredina): Članak pruža uravnotežen pregled tehnološkog razvoja i njegovih implikacija, uključujući i korporativne investicije i inicijative EU-a.
Zašto činjenice (85): The article accurately reports on the EU's KI-Gigafabriken initiative, including funding amounts, participation by member states, and timelines. It aligns closely with the primary source document and provides additional context about global investment trends.
Zašto objektivnost (60): The article presents information in a generally neutral tone but includes some commentary on financial impacts and market dynamics. While mostly objective, there is a slight tilt towards discussing economic implications.
U članku se kritizira esej Floriana Illiesa objavljen u časopisu "Zeit", u kojem tvrdi da tekstovi generirani umjetnom inteligencijom ne mogu odgovarati ljudskom pisanju.
Procjena pristranosti (Sredina): U članku se iznosi uravnotežena kritika tvrdnji Illiesa, bez da se otvoreno favorizira bilo koja strana.
Zašto činjenice (50): The article discusses a different topic entirely, Florian Illies' essay on AI-generated text quality. It does not mention the EU's KI-Gigafabriken initiative at all. Therefore, it cannot be assessed for factual accuracy regarding the primary source document.
Zašto objektivnost (30): The article has a subjective tone, with personal anecdotes and opinions about AI capabilities. It lacks neutrality and focuses on a specific critique of AI rather than presenting balanced information.
Članak izvještava da je DeepSeek, kineski dobavljač umjetne inteligencije, najavio značajno povećanje cijena za svoje usluge. Najava dolazi usred rastuće konkurencije na globalnom tržištu umjetne inteligencije i rastuće potražnje za naprednim rješenjima za strojno učenje. Iako članak ističe financijske posljedice ove odluke za kupce i poduzeća koja se oslanjaju na DeepSeekovu tehnologiju, ne pruža detaljne informacije o razlozima povećanja cijena ili bilo kakvih odgovora konkurenata.
Procjena pristranosti (Sredina): Članak predstavlja činjenične informacije o promjeni cijena kineske tvrtke za umjetnu inteligenciju, bez da otvoreno favorizira bilo koji određeni politički stav.
Zašto činjenice (30): This article is about a Chinese AI company raising prices, unrelated to the EU's KI-Gigafabriken initiative. It provides no relevant information about the primary source document.
Zašto objektivnost (20): The article is not objective as it only reports on a single company's action without providing context or balance. It fails to present multiple perspectives or neutral reporting.
SternNeovisanSredinaČinjenice 30Objektivnost 20prije 5 dana
Članak govori o iskustvu autora korištenja ChatGPT-a za planiranje odmora, koji na kraju nije prošao kako se očekivalo. Autor ističe ograničenja i potencijalne zamke oslanjanja na alate umjetne inteligencije za takve osobne zadaće planiranja.
Procjena pristranosti (Sredina): Članak se fokusira na korisničko iskustvo s tehnologijom umjetne inteligencije i ne uključuje nikakve političke teme, brojke ili politike.
Zašto činjenice (30): This article is about Google discontinuing a feature in Google Earth and does not relate to the EU's KI-Gigafabriken initiative. It provides no relevant information about the primary source document.
Zašto objektivnost (20): The article is not objective as it only reports on a single company's decision without providing context or balance. It fails to present multiple perspectives or neutral reporting.
Kako je izvijestila svaka strana
Isti događaj, grupiran prema političkom nagibu medija koji su o njemu izvještavali.
progresivno
sredina
konzervativno
★
Kako je izvijestila svaka strana
Podržite neovisne vijesti svjesne pristranosti i otključajte društveni puls, glasovanje zajednice i sve ostale značajke za podupiratelje.
Germany
