20 ms·
OpenAI reportedly nears breakthrough with "reasoning" AI
- kordlessagain 2y agoOpenAI has revealed a new five-tier classification system to measure its advancement towards developing artificial general intelligence (AGI). The levels range from current conversational AI to hypothetical future systems that could manage entire organizations. OpenAI believes its technology like GPT-4 is currently at Level 1 but on the verge of reaching Level 2 "Reasoners" with human-like problem solving abilities. While the framework provides a roadmap of OpenAI's ambitious goals, many questions remain on how to meaningfully quantify progress towards the nebulous target of AGI. The system is likely best interpreted as a communications tool to entice investors rather than a scientific way to track the development of human-level AI.
- TacticalCoder 2y agoWhy don't they start by allowing my paid subscription to use a model that'd be mostly beating Claude 3.5 Sonnet on benchmarks, instead of a model trounced on nearly every single benchmark by Claude? I'm probably cancelling my ChatGPT subscription soon: show me progress or I switch and give my munnies to Anthropic.
- JSDevOps 2y agoWhere have I heard this “level” thing before. Oh yeah that’s right. Musks Tesla and self driving cars.