Erik Meijer (@headinthebox) profile picture
Joined November 2011
Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)Photo by Erik Meijer (@headinthebox)
My proposal for AI safety formulated as the dual of human rights Punishing an innocent person is worse than letting a guilty person go ==> Universal Declaration of Human Rights: No person may be treated as guilty unless guilt is proven. Enacting unsafe AI acts is worse than rejecting safe AI acts ==> Universal Declaration of AI Acts: No AI act may be treated as safe unless harmlessness is proven. In other words, we should consider AIs to be dangerous unless proven safe.
3
6
1
49
6,156
LeCun’s new company on physical AI with world models [9] looks a lot like our 2014 company on physical AI with world models [1] 😀 See also [2-8] - all references in the reply!
30
25
20
375
115,374
Erik Meijer retweeted
Why I find Yann LeCun to be dishonest and unscholarly, with receipts. Lots and lots of receipts.
72
61
38
733
692,160
AI has many potential advantages and challenges when applied to hardware verification, especially formal verification. Join me for a free to attend webinar where we will discuss AI and hardware verification with panelists Adam Chlipala (MIT professor, formal hardware verification expert), Kanad Basu (leading work on using AI for generating formal properties) and Sean Safarpour (Executive Director at Synopsys, overseeing formal verification tools). workcast.com/register?cpak=6…
2
47
6,032
8
43
6
279
19,573
Erik Meijer retweeted
i wonder what a PIP looks like for a chief AI scientist
160
53
25
1,675
369,904
Erik Meijer retweeted
I was just talking with a friend who's been investing in startups for about 20 years, and we both agreed that one of the weirdest things about the AI boom is seeing journalists writing their usual contrarian stories about how it's a bubble about to burst.
119
91
12
1,809
243,773
Do you like poking holes in your favorite compiler’s type checker? Are you interested in discussing the various sources of unsoundness in type system and verification tools with like-minded PL nerds? The UNSOUND workshop at SPLASH 2024 is for you! Talk proposal deadline is Jul 15
2
14
2
47
6,371
Erik Meijer retweeted
The SWE-agent preprint has finally landed! Check it out at swe-agent.com/paper.pdf
6
71
3
320
37,856
Great opportunity for a fully-funded PhD position in the functional programming lab in Nottingham!
I have a fully funded PhD studentship for a home student at University of Nottingham. If you would like to work with me, in category theory, logic and theoretical computer science, please get in touch. Details on my website: stringdiagram.com/phd-studen…
6
12
4,951
Erik Meijer retweeted
ArtPrompt: ASCII Art-based Jailbreak Attacks against Aligned LLMs "We propose a novel ASCII art-based jailbreak attack. We show that five SOTA LLMs (GPT-3.5, GPT-4, Gemini, Claude, and Llama2) struggle to recognize prompts provided in the form of ASCII art. Based on this observation, we develop the jailbreak attack ArtPrompt, which leverages the poor performance of LLMs in recognizing ASCII art to bypass safety measures and elicit undesired behaviors from LLMs" (non-peer-reviewed) paper: arxiv.org/abs/2402.11753
2
32
6
104
13,736
Erik Meijer retweeted
In the past, innovation was slow, and mastering specific skills was valuable and useful for long periods of time We have now entered the era of exponential growth! The best strategy is to simply chill out and eat an Apple, like Tom Sawyer did, and let others do the hard work... Everything that you can imagine is going to be invented in the next 1-5 years anyway 🤣💤
52
79
14
437
119,659
Interested in a PhD studentship in the Functional Programming Lab in Nottingham? We are currently seeking applications for up to 10 positions. Please share! cs.nott.ac.uk/~pszgmh/10-phd…
24
45
6,426
Our group has a new preprint out, in which we make some very tentative steps towards translating trained neural networks into code: arxiv.org/abs/2402.05110 Quick summary/thoughts 🧵:
9
77
7
467
106,072
Running your own lab (watch the whole 7 minutes, it's priceless)
47
52
58
625
165,869
Haha, mutation, aliasing, and functions strike again! Even in programming languages for pros!
> Zig May Pass Anything By Reference Hmm that doesn't look very fun. lobste.rs/s/et3ivs/zig_may_p…
3
8
2
53
18,748
[reads news about Sam Altman] [closes newspaper] [takes a deep pull from pipe] "Well, see, AI's just cost another person their job."
12
27
1
324
28,851
For years, I had to convince other investors, potential candidates, and customers that AWS would not crush [insert startup name] each year following Re-Invent. Now, it seems like we’ll have to go through that same exercise after each OpenAI dev day.
5
12
1
102
21,146
It's official: we're recruiting in SE! Come join us! recruit.ap.uci.edu/JPF08667 #academia #academicjobs #SoftwareEngineering
10
3
24
7,499