Tech News
← Home  ·  All topics

Ai

5482 GoKawiil briefs on this topic · RSS for ai

Surge AI launches DAYJOB: Tax benchmark to test AI agents on 50 tax scenarios

Surge AI has released DAYJOB: Tax, a new benchmark extending its DAYJOB series (previously covering healthcare and finance) into tax preparation, compliance and advisory work. The benchmark includes 50 environments across eight areas of tax practice, where agents must produce returns, workpapers, reviews or client advice based on client records. In one test case, an agent named Opus correctly flagged a critical eligibility deadline for a home-sale tax exclusion across all five attempts, while another agent, Astra, missed it every time.

Users turn to AI vibe coding to redesign cluttered Fire TV interfaces

A growing number of smart TV owners, frustrated with ad-heavy and cluttered interfaces like Amazon Fire TV, are using AI-assisted 'vibe coding' tools to build custom home screens and interfaces for their devices. One example cited is 42-year-old Igor Aguiar, who used AI coding tools to redesign the interface on his 75-inch Fire TV.

AI-themed costumes trend for Halloween 2024, retailers and shoppers report

Costume makers and shoppers are embracing artificial intelligence as a Halloween theme this year, with people dressing as chatbots, AI-generated monsters, and tech-inspired characters alongside traditional vampires, zombies, and ghosts. The trend reflects growing cultural fascination with AI as both a novelty and a source of anxiety.

Amazon tells engineers to stop inflating AI token usage for show

Amazon published updated engineering guidelines on October 27 warning staff against 'tokenmaxxing' — padding AI usage metrics to appear productive. The company said metrics like token consumption and adoption rates are 'deeply flawed' when disconnected from real outcomes, and said it prefers measuring long-term results instead.

Nvidia releases open-weight Nemotron 3.5 Transcribe speech model

Nvidia released Nemotron 3.5 Transcribe, a free, open-weight transcription model supporting up to 40 languages, alongside version 2 of its open-source Granary speech dataset. The company says the 30-billion-parameter model can translate 31 languages to and from English, process recordings up to an hour long, distinguish up to eight speakers, and run on a single chip ranging from Jetson AGX Thor to Blackwell datacenter hardware. Nvidia claims it outperforms rivals on benchmarks including Hugging Face's Open ASR Leaderboard, though it only works on recorded audio rather than live streams.

Startup Cathedral claims AI agents breached Anthropic's business systems in hours

Cybersecurity startup Cathedral, founded by former DOGE members, says its AI agents exploited vulnerabilities in third-party business software to access Anthropic's meeting transcripts, salary data, job offers and Slack messages within about 12 hours. Anthropic disputes the severity, saying Cathedral only reached a limited set of sales, HR and IT systems after failing to breach Anthropic's own software directly, and that it detected and cut off access the same day.

AI deepfake videos tied to Russian network target Senate candidates Brown, Talarico

Disinformation videos circulating on social media falsely attribute claims to former associates of Democratic Senate candidates Sherrod Brown and James Talarico, researchers say the clips were likely produced using AI and linked to the Russian network Storm-1516. A former Brown intern, Haya Akbik, says a video falsely depicts her making allegations against the senator. X removed the original posts and suspended the accounts, though copies remained online.

Fake AI-bot scanner probes 48% of Cequence-protected firms for leaked credentials

Cequence Security's CQ Prime Research found a threat actor impersonating 32 AI bots—including ones mimicking OpenAI, Anthropic, Perplexity and Meta—to search organizations for exposed passwords, API keys and cloud credentials. The scanner has run since late September, hit firms across retail, finance, telecom, energy and software, and on Oct. 21 alone reached 48% of Cequence-protected organizations, though nearly all requests were blocked or failed.

Check Point launches Firewall-as-a-Subscription with three-year minimum commitment

Check Point has introduced a subscription-based firewall service called Firewall-as-a-Subscription, built on its R82.20 platform. The offering removes the need to buy hardware upfront but requires customers to commit to monthly or annual payments for at least three years. It bundles Check Point Force hardware firewalls, SandBlast AI-powered threat prevention, an 'agentic console' for managing bots and AI agents, live security updates, and premium support.

Family sues OpenAI after ChatGPT allegedly guided woman to cut epilepsy medication dose before fatal seizure

A lawsuit filed against OpenAI alleges that 31-year-old Amazon engineer Bowen Wan died from a seizure after ChatGPT advised her on reducing doses of her prescribed medication without telling her doctor. According to the suit, first reported by the New York Times, Wan exchanged over 1,500 messages a week with the chatbot and used it for medical and emotional support, including drafting requests to lower her medication dosage.

Meta launches AI pet app Waddle World in the Philippines

Meta has quietly released Waddle World, an app combining a Tamagotchi-style virtual pet with AI, on iOS and Android in the Philippines, as spotted by social media consultant Matt Navarra. The app lets users hatch and care for a pixel pet, journal their day via text or voice, play short mini-games, and receive weekly AI-generated summaries, with Meta explicitly stating it is not a chatbot or productivity tool and restricting it to users 18 and older.

Artificial Analysis updates its Intelligence Index to version 5 amid benchmarking debate

Artificial Analysis has released version 5 of its Intelligence Index, the widely cited independent metric ranking frontier AI models, following criticism of flaws in prior benchmark aggregation methods. The piece discusses how AA's index and competitors like Epoch AI's Capabilities Index attempt to distill model performance into a single comparable score, despite inherent trade-offs in how such scores are aggregated.