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。

AI is not alien, it's us

4 点作者 pbw超过 2 年前

2 条评论

jschveibinz超过 2 年前
I think this is really important. Here is my take:<p>Right now in our economy, we heavily depend upon the limited experience and expertise of a single individual to solve a problem, create something new or make something “trendy.” What if this model was changed?<p>AI uses data from “everywhere” to train. AI requires human input. Lots and lots of human input. These data are “previous” work that are publicly available, dated and not necessarily “in tune” with the times. But the results generated from this broad set of public sources is biased-and not necessarily in the “right” direction for business applications.<p>It is conceivable that in the near future, employees will be told to focus on generating unique content—like puzzle pieces for a puzzle that hasn’t yet been designed-that will serve the main interests of the business employer, and this output data (content) will be used to train the “private” AI to help “move” the results in a direction that is more useful and competitive for the business.<p>Think: “active business intelligence.” Looking forward rather than backward.<p>Of course, if and when AI is capable of non-linear, creative, cross-domain thinking, we may have reached our undoing. But until then, I feel cautiously optimistic.<p>I don’t think that we have even begun to see the possibilities of AI in the near future economy. In 2023, it’s still just a toy, albeit the most amazing toy ever built.
scott-smith_us超过 2 年前
It&#x27;s common to train UIs on a mountain of human-generated data. For some AI systems, that will influence things that the AI knows, and the relationships between things that it knows.<p>The point of saying that an AI is &quot;alien&quot; is to remind everyone that AIs don&#x27;t think AT ALL like humans do. That&#x27;s certainly true with current AI technologies, and likely to remain true for the next decade or two.<p>It&#x27;s dangerous for people to project human psychology onto AI systems in the same way that it&#x27;s dangerous for prison staff to project themselves onto sociopathic residents.<p>Being trained entirely on data from and about humans doesn&#x27;t imply that the AI&#x27;s thought process will be <i>anything like</i> a human&#x27;s. More likely, the AI will be very good at responding or interacting like a human.