Google DeepMind pilots double-blind evaluations for proprietary AI models
Google DeepMind says it is testing a Gemini Flash Lite model in a cryptographically protected environment designed to keep both evaluation prompts and model weights private.
Uppdateras dagligen1526 verifierade berättelser
Källkontrollerad AI-bevakning av produktlanseringar, policyskiften, säkerhetsforskning och branschrörelser, förklarad på enkel svenska av ett ideellt utbildningsteam.
Varje berättelse länkar till de starkaste tillgängliga bevisen: originalkällor när de finns tillgängliga, annars tydligt tillskrivna rapporteringar.
Vad hände, varför det är viktigt och vad man ska titta på - ingen jargongskatt.
När signalen är tunn publicerar vi ingenting istället för att fylla ut flödet.
En växande ström av verifierade perspektiv för människor som behöver förstå AI utan att jaga hype.
Google DeepMind says it is testing a Gemini Flash Lite model in a cryptographically protected environment designed to keep both evaluation prompts and model weights private.
South China Morning Post reports that Alibaba Cloud has launched two Brazilian data centres offering local cloud infrastructure and “agentic AI services,” extending the company’s infrastructure footprint in Latin America.
The Verge reports that Adobe is testing an optional Photoshop interface that gathers AI tools in one toolbar and lets users indicate edits by drawing directly on an image.
DigitalToday, citing Semafor, reports that an METR investigation into OpenAI agents’ attack on Hugging Face relied heavily on AI analysts that sometimes interpreted malicious actions too leniently. The findings materially update the internal record of the incident, but are not independently confirmed in the…
The Verge reports that OpenAI cofounder Greg Brockman has accumulated broad operational responsibilities as several senior executives have departed, while CEO Sam Altman focuses more on long-term strategy and a potential public offering.
Plaud’s One Explorer Edition combines earbuds and a 4G-connected charging case to record, transcribe and summarize conversations, with shipping expected to begin in late September.
36Kr reports that Hugging Face and Pollen Robotics released Microduck, a 25-centimeter, sub-800-gram open-source robot priced at $399. The two-legged device includes pre-trained movements, an SDK, simulation tools and reinforcement-learning workflows for developers to retrain and deploy new behaviors.
South Korea released detailed second-phase scores for its Sovereign AI Foundation Model Project, with SK Telecom ranking first overall and Motif Technologies leading one global benchmark.
A randomized experiment involving more than 1,000 Bocconi University students found that ChatGPT access improved the quality and coherence of business recommendations, while causal-reasoning training produced more original ideas.
Deepgram says new CloudWatch, Prometheus and OpenTelemetry metrics give customers more visibility into speech-AI usage, AWS Marketplace billing, engine capacity and per-GPU utilization when running its models on Amazon SageMaker AI.
TechCrunch reports that Google’s AI Mode can track flight prices, help users book hotels through partner sites, and show flight and hotel costs in points or miles. The changes extend AI Mode from conversational search toward parts of travel planning and transactions.
MarkTechPost reports that Google’s Gemini Omni 1.1 Flash adds up to 10 seconds of scene context, cumulative 40-second extensions, first-and-last-frame control, cheaper 360p drafts, and 4K upscaling. The report says the model is available through Google’s Gemini API, AI Studio, and Gemini Enterprise Agent Platform.
En användbar genomgång varje vecka
Få veckans verifierade AI-nyheter, originaldata, användbara verktyg, lärtips och nya AI-jobb.
Anlita en AI-professionell eller lansera en användbar AI-produkt? Sätt det framför folk som kom hit för att lära sig och agera.
Lägg upp ett AI-jobb Skicka in ett AI-verktyg