TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Watching the Generative AI Hype Bubble Deflate

4 点作者 erlend_sh5 个月前

1 comment

ilamont5 个月前
Serious question: Does anyone reading this completely trust unedited Generative AI text output to represent themselves as a person or their company? That is, you would sign a letter or post social media content that was actually 100% created by Generative AI, without any review or editing?<p>Hallucinations are only part of the problem. I recently fed a PDF manuscript (nonfiction) into Adobe&#x27;s &quot;AI Assistant&quot; and asked it to generate a summary. Adobe completely botched it, mistaking marketing text appearing on a single page of the frontmatter as the main focus of the entire book. The title page, table of contents (which clearly identify the topic), and the chapters were ignored.<p>For my YouTube channels, AI-generated suggested responses now appear in the comment moderation area, with many of them totally off the mark of what I would say in terms of tone and topic.<p>But at least I can reject this output. Some companies are using AI summaries by default with no human oversight, such as Amazon rewriting detail pages with AI without input from sellers, or customer service bots which make stuff up.