YBX AI Radar
Channels

AI Research

AI 赛道深度、公司拆解、概念解读和周报月报。

6.0 AI Research Tech Xplore AI
AI Research

The Indian workers training AI robots to take their jobs

AI Summary: With a smartphone strapped to her head, Indian housewife Nagireddy Sriramyachandra films herself slicing mangoes to train AI-powered robots to take on household jobs in the future.

The Indian workers training AI robots to take their jobs
Source: Tech Xplore AI 6.0
8.0 AI Research Tech Xplore AI
AI Research

MIT researchers channel AI to turn hand gestures into robot training data

MIT researchers have developed an AI-powered robot training solution combined with an ultrasound wristband, which captures movements of subcutaneous muscles, tendons and ligaments via the wristband worn on the wrist, and converts hand gestures into robot training data. It can help humanoid robots tackle complex tasks like grasping cups, and is expected to lower the threshold and cost of robot training, according to a report from Tech Xplore AI.

MIT researchers channel AI to turn hand gestures into robot training data
Source: Tech Xplore AI 8.0
AI Radar Summary

MIT研究团队开发了一套结合AI与超声波腕带的机器人训练新方案，通过佩戴腕带捕捉人体皮下肌肉、肌腱与韧带的运动，将手部手势转化为机器人训练数据，可帮助类人机器人解决抓取杯子等复杂操作任务，有望降低训练门槛与成本，相关信息来自Tech Xplore AI的报道。

6.0 AI Research Tech Xplore AI
AI Research

Why old EV batteries still go to waste—and how AI could change that

AI Summary: Even though retired lithium-ion batteries often retain up to 80% of their capacity, significant economic value and climate benefits are currently being lost. A new article in Natur

Why old EV batteries still go to waste—and how AI could change that
Source: Tech Xplore AI 6.0
8.0 AI Research EleutherAI Blog
AI Research

RLHF and RLAIF in GPT-NeoX

This article is sourced from EleutherAI's official blog, which introduces that after the cooperation between EleutherAI and SynthLabs, the open-source large model GPT-NeoX now supports post-training alignment based on Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning from AI Feedback (RLAIF). This update helps developers conveniently fine-tune GPT-NeoX to optimize the alignment between model outputs and human preferences and AI feedback standards, representing a research progress in the field of AI large model training and alignment.

RLHF and RLAIF in GPT-NeoX
Source: EleutherAI Blog 8.0
AI Radar Summary

本文来自EleutherAI官方博客，介绍该机构与SynthLabs合作后，开源大模型GPT-NeoX现已支持基于人类反馈强化学习（RLHF）与人工智能反馈强化学习（RLAIF）的训练后对齐微调。该更新可帮助开发者便捷地对GPT-NeoX进行针对性优化，提升模型输出与人类偏好、AI反馈标准的匹配度，属于AI大模型训练与对齐方向的研究进展。