TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

DeepMind AI is as fast as humans at solving previously unseen tasks

158 pointsby antonddover 2 years ago

17 comments

mshake2over 2 years ago
Maybe they should make a persuasive demo if they want people to care. ChatGPT isn't so hot right now because boring PR articles were written about its abilities. It's hot because you can try it and see for yourself. Basic showmanship.
评论 #34753121 未加载
评论 #34752813 未加载
评论 #34753527 未加载
评论 #34754590 未加载
评论 #34754528 未加载
评论 #34753500 未加载
评论 #34759693 未加载
评论 #34753223 未加载
评论 #34754929 未加载
评论 #34754330 未加载
评论 #34754294 未加载
greatpostmanover 2 years ago
Google pumping out PR to get some damage control on chatgpt. No evidence in the public of any of these claims. Show us the models google
评论 #34754543 未加载
评论 #34754143 未加载
评论 #34753200 未加载
V__over 2 years ago
&gt; solve tasks that involve navigating, planning and manipulating objects.<p>Without some real examples, this is just PR talk. I could just as well say that &quot;the Stockfish AI&quot; is able to solve previously unseen tasks on a virtual board, moving pieces around, thinking ahead to navigate around and outsmart any human and non-human opponent. Doesn&#x27;t say much.
评论 #34755302 未加载
tehsauceover 2 years ago
Here&#x27;s a video summary of the research: <a href="https:&#x2F;&#x2F;youtu.be&#x2F;U93bUQ1roiw" rel="nofollow">https:&#x2F;&#x2F;youtu.be&#x2F;U93bUQ1roiw</a>
评论 #34757056 未加载
Wowfunhappyover 2 years ago
If some software&#x2F;AI is able to do something, I <i>expect</i> it to be as fast or faster than a human because you can basically always throw more computing power at the problem.<p>The important questions are (1) can the computer do it in the first place and (2) at what financial cost?
评论 #34754565 未加载
6gvONxR4sf7oover 2 years ago
Here’s the original article that this is reporting on: <a href="https:&#x2F;&#x2F;sites.google.com&#x2F;view&#x2F;adaptive-agent&#x2F;" rel="nofollow">https:&#x2F;&#x2F;sites.google.com&#x2F;view&#x2F;adaptive-agent&#x2F;</a>
lettergramover 2 years ago
So in a contrived environment where the AI is given the ability to train on its attributes (aka same simulation framework, different layout). It can effectively explore the layout and complete tasks at the human level.<p>Perhaps there’s only a few search algorithms and humans (and now this AI) are optimized to do it as well.<p>One of my shower thoughts are that we’re effectively recreating the wheel with AI. Yes it can be immensely powerful and outperform humans for specific tasks. However, humans are generalized and evolved to interact with the real world independently. I imagine an AI would effectively match a human in this regard someday, but it’s worse as it can’t generate energy from eating some plants or animals (ie naturally occurring solar generated food). The interconnected nature will let AI have an advantage for many tasks, no doubt, but humans are pretty well optimized for the real world. Imo it’ll be hard to beat.
评论 #34752699 未加载
评论 #34752996 未加载
评论 #34753616 未加载
评论 #34752725 未加载
评论 #34752991 未加载
habiburover 2 years ago
The key part :<p>&gt; The AI, called Adaptive Agent or AdA, works in a 3D virtual world where it is asked to solve tasks that involve navigating, planning and manipulating objects.<p>Not in real world.
评论 #34754044 未加载
asimover 2 years ago
It&#x27;s funny. The way deepmind is going about it, these AI agents are going to live in a total virtual world in which we then enter and participate. It&#x27;s not clear that AI will actually immerse itself in the real world. While we find general uses for differently trained models, the whole robots being everywhere thing seems less likely than say a 3D metaverse we participate in using AR glasses so we can see these digital creatures. Could be totally wrong, could argue for the other case, it mostly doesn&#x27;t matter. I just don&#x27;t think our interactions with this technology will be physical.
评论 #34754370 未加载
评论 #34753273 未加载
评论 #34753347 未加载
margorczynskiover 2 years ago
It&#x27;s behind a paywall, any way to go past it? Also would be great if there was an actual paper or more serious article about it. I see a lot of this type of articles recently and most of them are wrong&#x2F;hype&#x2F;etc. and the only valuable part are the links to the sources.<p>EDIT: Found these with Google:<p><a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2301.07608" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2301.07608</a><p><a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=U93bUQ1roiw">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=U93bUQ1roiw</a>
评论 #34752918 未加载
RcouF1uZ4gsCover 2 years ago
I think DeepMind is really good at solving “spherical cow” problems. They can model a very simplified model of a small aspect of the real world and solve it brilliantly.<p>However, there is a big question of how much the results are actually transferable to the real world.
faeriechanglingover 2 years ago
I don&#x27;t believe them because they won&#x27;t release it to the public and let us see for ourselves and that says everything.
antonddover 2 years ago
Just realised the article is behind a soft paywall. Here’s a LinkedIn post by a Deepmind engineer working on ADa: <a href="https:&#x2F;&#x2F;www.linkedin.com&#x2F;pulse&#x2F;ada-ai-adapts-fast-humans-edward-hughes" rel="nofollow">https:&#x2F;&#x2F;www.linkedin.com&#x2F;pulse&#x2F;ada-ai-adapts-fast-humans-edw...</a>
panda88888over 2 years ago
Looks like currently the agents interact in the simulated world only. Would be interesting when combined with robots from Boston Dynamics or similar and see how it learns to adapt and operate under real world environment and constraints.
bigape911over 2 years ago
Depending on the problem, I expect it to be much much faster especially foor physics and math problems.
bigape911over 2 years ago
It&#x27;s sad to see the great minds at Google fail to capitalize on their tech
mawadevover 2 years ago
I personally don&#x27;t believe a single thing about chatgpt and any ai that pops up around it. To me it&#x27;s the same as google search - instead of truly enriching our lives, it simply made us dependent and then more lazy to the point where we rather search instead of thinking for a brief moment. Now that google became unusable, ai is supposed to step in or what...<p>I hope the volumes of worthless content it generates destroy the internet as we know it, so we can finally move on to better things and start building communities again.
评论 #34753989 未加载
评论 #34753901 未加载
评论 #34760880 未加载
评论 #34753833 未加载