Kevin Lacker @lacker

Working on math + AI at https://t.co/u95v5xIPQ4 and telescope software at https://t.co/Yx0Z8UFXOE. Formerly: Parse cofounder, Facebook, Google lacker.io Piedmont, California Joined March 2008

Tweets

18K
Followers

7K
Following

1K
Likes

23K

Kevin Lacker @lacker

a day ago

Wrote a post about Stanislaw Lem, how the AI in his science fiction is similar to Fable, and which of his books you should read. lacker.substack.com/p/stanislaw-le…

0 1 3 974 0

View Details

Kevin Lacker @lacker

4 days ago

@tribbloid @roboguy20 yeah… the user base is changing, it’s moving from theory to practice

0 0 1 22 0

View Details

@roboguy20 @tribbloid this is an interesting point, about the difficulty of finding a representation that is both efficient and easily provable. I feel like I have run into this myself many times but never expressed it this concisely

1 0 2 25 0

View Details

Kevin Lacker @lacker

5 days ago

@slatestarcodex @Scholars_Stage Truman also got annoyed by Oppenheimer angsting about the downsides of nukes, maybe there's a parallel to Trump and Dario

0 0 1 184 0

View Details

Kevin Lacker @lacker

6 days ago

@zooko I think nix might be resistant, because they (accidentally?) have the most secure philosophy, of making it a huge pain to ever update any dependency

1 0 1 24 0

View Details

Kevin Lacker @lacker

6 days ago

Wrote a post about cybersecurity, why Fable is slightly overrated and we should be paying more attention to Shai-Hulud. lacker.substack.com/p/cybersecurit…

2 2 14 3K 5

View Details

Kevin Lacker @lacker

6 days ago

@zooko the irony is that the biggest problems are coming in JavaScript, a language designed from the ground up to run in a sandbox

1 0 1 24 0

View Details

Kevin Lacker @lacker

6 days ago

Fable doesn’t actually have to be impossible to jailbreak. It just has to be more annoying for hackers to circumvent its controls, than to use the inferior Chinese models.

0 0 3 382 0

View Details

Kevin Lacker @lacker

6 days ago

@tszzl @WestsideLAGuy the tech aura is real. but not everyone has the eyes to see.

0 0 2 625 0

View Details

Kevin Lacker @lacker

6 days ago

@joseph_h_garvin @Aron_Adler how about for example, de bruijn indexing. fairly easy to represent "is this de bruijn index valid" as a dependent type. but proofs suck, you're constantly shifting and dealing with partial data that itself isn't obviously typed in a useful way

1 0 2 28 2

View Details

Kevin Lacker @lacker

6 days ago

@TheZvi @teortaxesTex Depends on the type of security - it is logically possible to write bug-free code, but fraud, impersonation, and other social engineering do not seem to have perfect defenses.

0 0 3 253 0

View Details

Kevin Lacker @lacker

7 days ago

@perrymetzger Maybe we need a formal semantics for Rust that is not 100% accurate to the official compiler, along with some escape hatch like "unsafe". Kind of a chicken and egg where you want to show it's useful before official Rust can commit to it.

0 0 1 518 0

View Details

Kevin Lacker @lacker

7 days ago

In other words... Trump is demanding that we "Pause AI"? This sure sounds a lot like "Stop the development of the most powerful general AI systems until we know how to make them safe." Well now we have it, a pause. Let's see if it makes anyone safer.

Polymarket @Polymarket

7 days ago

NEW: The White House is reportedly demanding Anthropic make Fable 5 impossible to jailbreak before rerelease — which security experts say “can’t be done.”

333 279 5K 1.3M 399

0 0 5 350 0

View Details

Kevin Lacker @lacker

a week ago

@aidan_mclau @tszzl the “icky feeling” seems fixable. what if it integrated right into your consciousness, bit by bit, so it all felt completely natural, and you could control the whole process. like you’re just tuning your own brain.

0 0 2 198 0

View Details

Kevin Lacker @lacker

a week ago

@dbasch I feel like most billionaires are not like the “famous billionaires”, you wouldn’t know their name, and the main way their life is different from a mere centimillionaire is more foundation type work and some different top end luxuries.

0 0 2 203 0

View Details

Scott Kominers @skominers

a week ago

In the last week, multiple colleagues have expressed concern that generative AI will somehow destroy society's esteem for mathematics and mathematicians. Contrariwise, I conjecture the opposite. I've never seen the level of public fascination with math that we're seeing right now — weekly articles in major outlets; people outside the field teaching themselves arithmetic combinatorics; heck, even the owner of a local café recently asked me to explain the unit distance conjecture. If anything, this seems likely to renew students' excitement, uncover new applications, and open new frontiers — all of which should inure to the benefit of the field and of mathematicians. QED?

33 57 319 162K 76

View Details

Kevin Lacker @lacker

a week ago

@satnam6502 @dearmadisonblue like I love watching the NFL. if you were to tell me it's stupid and pointlessly violent, well, I'm not going to argue you're wrong. I still like it though. I don't like Marvel movies but I wouldn't argue that the NFL is aesthetically superior, you just like what you like.

0 0 0 7 0

View Details

Kevin Lacker @lacker

a week ago

@satnam6502 @dearmadisonblue well, people have different taste in art and that's fine. I think it is natural especially if you consume a ton of something to get bored by it. maybe I consume *so* much AI writing for work that I find it cliched for general writing.

1 0 0 25 0

View Details

Kevin Lacker @lacker

a week ago

@TheZvi @lionellevine It just takes a while! Here's an example: "Five ideas died, each sharpening the problem (the weights can be taken nonnegative; the wall is cell-independent; the input is an evaluation, not a bound)." In the task I gave it, there are no "weights", "walls" or "cells".