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.

The Princeton researchers calling out ‘AI snake oil’

38 pointsby irtefaover 1 year ago

4 comments

dangover 1 year ago
Related. Others?<p><i>Princeton ‘AI Snake Oil’ authors say GenAI hype has ‘spiraled out of control’</i> - <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=37243354">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=37243354</a> - Aug 2023 (101 comments)<p><i>ML is useful for many things, but not for predicting scientific replicability</i> - <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=37089573">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=37089573</a> - Aug 2023 (37 comments)<p><i>The greatest risk of AI is from the people who control it, not the tech itself</i> - <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=36139852">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=36139852</a> - May 2023 (504 comments)<p><i>GPT-4 and professional benchmarks: the wrong answer to the wrong question</i> - <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=35245626">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=35245626</a> - March 2023 (185 comments)<p><i>ChatGPT is a bullshit generator but it can still be amazingly useful</i> - <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=34618376">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=34618376</a> - Feb 2023 (154 comments)
pedalpeteover 1 year ago
<a href="https:&#x2F;&#x2F;archive.is&#x2F;tqUg2" rel="nofollow noreferrer">https:&#x2F;&#x2F;archive.is&#x2F;tqUg2</a><p>I&#x27;m surprised they are only providing 2 buckets for AI, predictive and generative, and that they say &quot;most of the snake oil is in predictive&quot;.<p>Most of the snake oil being in predictive isn&#x27;t the same as saying most predictive AI is snake oil.<p>With predictive most predictive AI, we can provide a result of how often the AI gets the prediction correct.<p>We&#x27;re in the sleep space, and do sleep stage classification using neural nets. My co-founder decided for fun to see how the model would work in predicting sleep stages, and it was surprisingly good (above 80% coherence with manually scored data if I remember correctly).<p>I think there is a different challenge in things like psychology, where it is very difficult to get a true reading on what is happening with the subject. However, we&#x27;re also building EEG, so maybe we&#x27;ll get some insights there as well (but psychology is not our focus).
Nevermarkover 1 year ago
&gt; One example is non-consensual deepfakes. This is an area where you don’t need information to spread virally for it to cause harm.<p>At some point soon, in a galaxy you are living in, AR headsets are going to let you see everyone naked!<p>Seriously, deep fakes are going to be so easy to find, make and animate that society is going to stop being sensitive to them. There is no other option.<p>Today I can draw anyone I choose doing any odd thing. I can put horns on their head and have them bubble talk every kind of nonsense. Nobody cares. At all. It&#x27;s going to get there for these fakes too, out of sheer fatigue.<p>Let&#x27;s call it &quot;Deep Outrage Exhaustion&quot;.
评论 #37579346 未加载
评论 #37578936 未加载
AISnakeOilover 1 year ago
Finally my time to shine.