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。

Why general artificial intelligence will not be realized

3 点作者 banjo_milkman将近 5 年前

2 条评论

golem14将近 5 年前
It&#x27;s an interesting article. I&#x27;m not sure I agree with it 100% - I would say a lot depends on the larger picture and incentives.<p>Right now, there is a huge incentives to create &quot;narrow AIs&quot; because they help improve productivity and get rid of tedious labor, from facial recognition to harvesting machines. So, none of those AIs has &#x27;skin in the game&#x27; or as the paper calls is &#x27;are in this world&#x27;.<p>However, if there were a concerted push to give AIs &#x27;skin in the game&#x27; and try to solve more general problems (the poker AIs look interesting, for instance), the situation might be different.<p>I&#x27;m thinking along the lines of the essay &quot;Golem XIV&quot; or the TV show &quot;person of interest&quot;. In these SciFi scenarios, governments give machines very high level objectives (&quot;protect the country!&quot;) and skin in the game, which then ultimately leads to reaching &quot;consciousness&quot;.<p>I don&#x27;t see how with current economic incentives, such efforts would be fundable, the narrow AI field is sucking up all the talent and $$$. But if someone were to massively fund efforts like this, I wouldn&#x27;t rule out progress in that direction. Spoken as someone with practically nil education in ML and AI :)
ksaj将近 5 年前
There was a time people thought humans could never build a flying machine. Then we invented helicopters, jets, rockets and airplanes - all flying machines, none of which even attempt to emulate birds and bees.<p>As a person who&#x27;s spent 3&#x2F;4 of his life on subjects like Artificial Life, synthetic psychology, etc, I believe we&#x27;ll eventually reach <i>various</i> types of intelligence. But not until we figure out what exactly produces the effect that we currently define as intelligence, and emulate those instead of trying to build replicas of machines (brains, organs, etc) that already seem to exhibit it.<p>Attempting to build brains is one of many paths to learning what that needed extra spice is, but you can be pretty sure if&#x2F;when we eventually do create an artificial intelligence, it won&#x27;t look or act much like our own, except for the parts we choose to cosmetically appear that way.