Embed Notice

HTML Code

<blockquote style="position: relative; padding-left: 55px;"><section><a href="https://mastodon.social/users/shriramk/statuses/116998683256115897">Shriram Krishnamurthi (shriramk@mastodon.social)'s status on Wednesday, 29-Jul-2026 02:10:03 JST</a><a href="https://mastodon.social/@shriramk" title="shriramk@mastodon.social"><img src="https://gnusocial.jp/avatar/79575-48-20221223221604.webp" width="48" height="48" alt="Shriram Krishnamurthi" style="position: absolute; left: 0; top: 0;">Shriram Krishnamurthi</a><div><a href="https://mastodon.social/@shriramk/116998679679492612" rel="in-reply-to">in reply to</a></div></section><article><p>3/ ↵</p></article><footer><a rel="bookmark" href="https://gnusocial.jp/conversation/6583866#notice-12962055">In conversation</a><time datetime="2026-07-29T02:10:03+09:00" title="Wednesday, 29-Jul-2026 02:10:03 JST">about 3 months ago</time> <span>from <span><a href="https://mastodon.social/@shriramk/116998683256115897" rel="external" title="Sent from mastodon.social via ActivityPub">mastodon.social</a></span></span><a href="https://mastodon.social/@shriramk/116998683256115897">permalink</a><h4>Attachments</h4><ol><li><label><a rel="external" href="https://gnusocial.jp/attachment/6295817">&gt; **A bright 22-year old walks into your office and asks your advice about what kind of formal methods they should do a PhD in. The PhD will end in 2031, past the point where many folks in frontier labs believe AGI will be achieved. What would you say to them?**(a) What I always say, e.g., in my FAQ, but even more so: Take a year off to go work for one of the billion-dollar AI+verification startups, because we many never see money like that again, so make the most of it. Make sure you get paid in $, not only in stock.(b) While you're there, study the discrepancy between their Twitter statements and what you actually observe internally.(c) Focus on finding places where human intent matters. As a concrete example, we've got all this amazing verification technology…and yet, I don't even need to ask as complicated a question as cyberphysical systems for us to come up short: I can just ask, "Is this system usable?" And if that's too fuzzy, think about usable security: it's an area that has gotten almost zero attention (I've done literature reviews!), and a place where we have real technical questions but also real trade-offs that people have to make decisions about so that tools can't just operate independently.</a></label><br><a href="https://files.mastodon.social/media_attachments/files/116/998/681/997/258/376/original/9297a93527be0ae7.png" rel="external">https://files.mastodon.social/media_attachments/files/116/998/681/997/258/376/original/9297a93527be0ae7.png</a></li></ol></footer></blockquote>

Corresponding Notice

Embed this notice
Shriram Krishnamurthi (shriramk@mastodon.social)'s status on Wednesday, 29-Jul-2026 02:10:03 JST Shriram Krishnamurthi
in reply to
3/ ↵
In conversationabout 3 months ago from mastodon.socialpermalink
Attachments
1. > **A bright 22-year old walks into your office and asks your advice about what kind of formal methods they should do a PhD in. The PhD will end in 2031, past the point where many folks in frontier labs believe AGI will be achieved. What would you say to them?** (a) What I always say, e.g., in my FAQ, but even more so: Take a year off to go work for one of the billion-dollar AI+verification startups, because we many never see money like that again, so make the most of it. Make sure you get paid in $, not only in stock. (b) While you're there, study the discrepancy between their Twitter statements and what you actually observe internally. (c) Focus on finding places where human intent matters. As a concrete example, we've got all this amazing verification technology…and yet, I don't even need to ask as complicated a question as cyberphysical systems for us to come up short: I can just ask, "Is this system usable?" And if that's too fuzzy, think about usable security: it's an area that has gotten almost zero attention (I've done literature reviews!), and a place where we have real technical questions but also real trade-offs that people have to make decisions about so that tools can't just operate independently.
  https://files.mastodon.social/media_attachments/files/116/998/681/997/258/376/original/9297a93527be0ae7.png

Public

Embed Notice

HTML Code

Corresponding Notice