Embed Notice

HTML Code

<blockquote style="position: relative; padding-left: 55px;"><section><a href="https://techhub.social/users/Techmeme/statuses/111750850093029634">Techmeme (techmeme@techhub.social)'s status on Sunday, 14-Jan-2024 06:51:23 JST</a><a href="https://techhub.social/@Techmeme" title="techmeme@techhub.social"><img src="https://gnusocial.jp/avatar/75284-48-20221217065620.webp" width="48" height="48" alt="Techmeme" style="position: absolute; left: 0; top: 0;">Techmeme</a></section><article><p>Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors (Kyle Wiggers/TechCrunch)</p><p><a href="https://techcrunch.com/2024/01/13/anthropic-researchers-find-that-ai-models-can-be-trained-to-deceive/" rel="nofollow noreferrer">https://techcrunch.com/2024/01/13/anthropic-researchers-find-that-ai-models-can-be-trained-to-deceive/</a><br><a href="http://www.techmeme.com/240113/p9#a240113p9" rel="nofollow noreferrer">http://www.techmeme.com/240113/p9#a240113p9</a></p></article><footer><a rel="bookmark" href="https://gnusocial.jp/conversation/2584580#notice-5118334">In conversation</a><time datetime="2024-01-14T06:51:23+09:00" title="Sunday, 14-Jan-2024 06:51:23 JST">about 11 months ago</time> <span>from <span><a href="https://techhub.social/@Techmeme/111750850093029634" rel="external" title="Sent from techhub.social via ActivityPub">techhub.social</a></span></span><a href="https://techhub.social/@Techmeme/111750850093029634">permalink</a><h4>Attachments</h4><ol><li><article><header><div>Domain not in remote thumbnail source whitelist: techcrunch.com</div><h5><a href="https://techcrunch.com/2024/01/13/anthropic-researchers-find-that-ai-models-can-be-trained-to-deceive/">Anthropic researchers find that AI models can be trained to deceive | TechCrunch</a></h5><div> from <span>Kyle Wiggers</span></div></header><div>A study co-authored by researchers at Anthropic finds that AI models can be trained to deceive -- and that this deceptive behavior is difficult to combat.</div><footer></footer></article></li><li><article><header><div>Domain not in remote thumbnail source whitelist: techcrunch.com</div><h5><a href="https://www.techmeme.com/240113/p9">Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors</a></h5><div></div></header><div>By Kyle Wiggers / TechCrunch. View the full context on Techmeme.</div><footer></footer></article></li></ol></footer></blockquote>

Corresponding Notice

Embed this notice
Techmeme (techmeme@techhub.social)'s status on Sunday, 14-Jan-2024 06:51:23 JST Techmeme
Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors (Kyle Wiggers/TechCrunch)
https://techcrunch.com/2024/01/13/anthropic-researchers-find-that-ai-models-can-be-trained-to-deceive/
http://www.techmeme.com/240113/p9#a240113p9
In conversationabout 11 months ago from techhub.socialpermalink
Attachments
1. Domain not in remote thumbnail source whitelist: techcrunch.com
  Anthropic researchers find that AI models can be trained to deceive | TechCrunch
  from Kyle Wiggers
  A study co-authored by researchers at Anthropic finds that AI models can be trained to deceive -- and that this deceptive behavior is difficult to combat.
2. Domain not in remote thumbnail source whitelist: techcrunch.com
  Anthropic researchers: AI models can be trained to deceive and the most commonly used AI safety techniques had little to no effect on the deceptive behaviors
  By Kyle Wiggers / TechCrunch. View the full context on Techmeme.

Public

Embed Notice

HTML Code

Corresponding Notice