OpenAI claims ChatGPT’s new default model hallucinates way less

Schriftgröße

OpenAI's newest default model for ChatGPT might not make stuff up as much. Hallucinations have been an ongoing problem for AI models, but OpenAI says its new GPT-5.5 Instant model has "significant improvements in factuality across the board." The company claims that, based on "internal evaluations," GPT-5.5 Instant produced "52.5% fewer hallucinated claims" than its […]

Continue reading on The Verge

This article was originally published by The Verge.

Ähnliche Meldungen

In Entwicklung·10 sa önce

Meta CEO Zuckerberg Admits AI Development Pace Not Meeting Expectations

Meta CEO Mark Zuckerberg stated that AI agent development has not accelerated as expected. The company previously laid off 8,000 employees and reassigned 7,000 to AI groups, a restructuring Zuckerberg admitted was not "clean." He anticipates improvements from AI investments in 3-6 months.

TechCrunch

In Entwicklung·12 sa önce

Meta Launches New Pocket App Focused on AI-Generated 'Gizmos'

Meta has launched a new app called Pocket, distinct from the read-it-later service, which allows users to create and share interactive AI-generated 'gizmos' based on prompts. The app is not yet available globally.

The Verge

In Entwicklung·13 sa önce

PamStealer: New macOS Malware Uses Clever Tactics for Stealthy Credential Theft

A new macOS malware, PamStealer, employs sophisticated techniques including a disguised disk image, AppleScript, and Rust-based infostealer to stealthily steal user credentials. It bypasses security features like com.apple.quarantine and uses native macOS APIs for a quieter execution chain.

Ars Technica

In Entwicklung·13 sa önce

Meta Launches Pocket, a New Social App for Creating AI-Powered Gizmos

Meta has launched Pocket, a new social app allowing users to create and share interactive "gizmos" using AI prompts. While not yet available everywhere, the app enables users to scroll a feed of these AI-generated experiences, which can respond to touch, phone tilt, and media from the user's device.

The Verge

In Entwicklung·14 sa önce

Anthropic Reportedly Exploring Custom AI Chip Collaboration with Samsung

AI company Anthropic is reportedly in talks with Samsung to explore a collaboration on custom AI chips, potentially reducing reliance on Nvidia and other suppliers. The company stated its compute strategy relies on a diversified hardware stack.

TechCrunch

In Entwicklung·14 sa önce

US Proposes Overturning Ban on Commercial Supersonic Flights

The FAA proposed a new rule to overturn the 53-year ban on commercial supersonic flights over the US, replacing it with a noise-based standard. This could allow for the return of supersonic airliners if sonic booms are kept below 0.11 psf.

Ars Technica