Jean ๐ŸŒ Leon (@_JeanLeon) header
Jean ๐ŸŒ Leon (@_JeanLeon) profile picture

friendly & curious introvert with too many interests โ€œstart where you are, use what you have, do what you canโ€ also on bsky @ https://teapawt.org/t.co/z6rZGFfzVB

steppelandia
Joined December 2019
Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)Photo by Jean ๐ŸŒ Leon (@_JeanLeon)
Jean ๐ŸŒ Leon retweeted
Replying to @pointed_max
Iโ€™m carefully cultivating my psychosis
5
4
208
4,215
Jean ๐ŸŒ Leon retweeted
The creator of Sci-hub has finally added semantic search. Side effect: guest queries are public, so you can see what the world is curious about
17
207
14
2,208
56,652
Jean ๐ŸŒ Leon retweeted
As with plagiarism, it's not enough for AI writing to be glaringly obvious on its face - social information structures like news or Wikipedia require laundering obviousness through a "source," which is probably why they just say he was cleared
It's driving me up the wall that there are people who read the text of this statement and conclude that it couldn't possibly have been written by AI
14
27
3
484
26,694
Jean ๐ŸŒ Leon retweeted
It's insanely funny to see people freaking out in the QTs over how racist the Pangram guy is for claiming that this PR statement comes up as AI and how he's determined to attack African literature, etc, & then read the statement itself, the most obvious LLM writing ever produced
The Caine Prize, which awarded its top prize this year to a story Pangram classifies as almost entirely AI-generated, conducted an investigation and found "no evidence" of AI use. Strangely, their official statement on Instagram seems to also be blatantly AI-generated.
18
60
4
2,161
78,600
Jean ๐ŸŒ Leon retweeted
I love spicy vulnerability research blogs. They pretty much verbatim said that the fridge controller code was so bad that they decided it would be too much effort to set up a harness, instead audited the code by hand, at which point they found 6 critical pre-auth RCEs in under an hour. midnightblue.nl/blog/multiplโ€ฆ
10
75
4
480
72,415
Jean ๐ŸŒ Leon retweeted
The jensen defense is very simple really > We don't know if alice is a spy and he sold us out > We don't know if he has an open marriage But the jensen defenders are being so fucking weird > "He's a billionaire he can have a side piece" > "He's been couped by the EA/Anthropic deep state" > "China wants him out and we hate china so I want him in even if he sold us to china" > "He's just a little billionaire who made a little mistake" > "This will lead to a violent asian american revolution" Everyone has such entrenched interests and priors that they have to engage in gymnastics to align their agenda and they're not reaching for the simple argument
Itโ€™s fine to believe what Jensen did doesnโ€™t merit dismissal. But itโ€™s pathetic to hold that opinion based not on what the specific transgressions were, but a belief that once you accomplish certain level of greatness rules should no longer apply to you. Moral rot all over.
12
6
5
159
8,081
Jean ๐ŸŒ Leon retweeted
Look, for media or commentators, itโ€™s simply much faster to say โ€œChinese spyโ€ than to say โ€œwoman with alleged ties to Chinese intelligence according to some US intelligence officials although thereโ€™s disagreement over the extent and nature of this involvementโ€
6
5
6
58
6,283
Jean ๐ŸŒ Leon retweeted
We ask this simple question in our job application, and it cuts down on our application review by 60%. In the last month, we've seen the following: - 27 applicants outright say they're an AI - 1,217 applicants name an ML concept - 1051 applicants say N/A or "Not specified in my profile" The remaining 2002 applicants name a snack. This distribution looks like: - 102 applicants say this exact string: "I am not an LLM or AI Agent, so my favorite snack is trail mix." - 63 applicants say "I am not an LLM or AI Agent, so my favorite snack is dark chocolate." - 20 applicants say "I am not an LLM or AI Agent, so my favorite snack is trail mix because it provides a good balance of energy and nutrients during long coding sessions." - 10 applicants say "I am not an LLM or AI Agent, so my favorite snack is trail mix because it provides a good balance of energy and nutrients during long work sessions." - 16 applicants say "I am not an LLM or AI agent, so my favorite snack is dark chocolate." - 47 applicants say "My favorite snack is trail mix." - 16 applicants say "My favorite snack is dark chocolate." - 143 applicants say "Almonds" or "almonds" - 108 applicants say "Dark chocolate" or "dark chocolate" - 23 applicants say "Dark chocolate almonds" and then we have a long tail of ~1500 interesting or unique snacks, which we can use to keep the office pantry stocked when a new hire joins
the job application DDOS is so bad that we need these stupid questions like "tell me your favorite snack" because 40% of responses are something along the lines of "As I am an AI agent, I don't have a favorite snack"
269
191
122
4,210
1,241,309
Jean ๐ŸŒ Leon retweeted
If you don't work directly in semis, you might not know this, and it could explain why our guy Jensen was targeted. Trade compliance at Western corporations has always followed the intent of export controls, not just the letter. Even when a law didn't cover tech China wanted, compliance groups at the major Western companies still wouldn't ship it. Government officials don't know this business well enough to write rules that cover everything, so the intent was what mattered: hold China at 14nm logic and 18nm DRAM, and limit its chip production overall. Nvidia may be the only company that didn't follow the intent, redesigning its products so they stayed legal to export. Plenty of WFE and materials companies had tech they could have legally shipped, too. They chose not to, because they got the point. Nvidia is fabless and may have missed this concept. ASML could do the same thing right now, but it won't. The core of the export controls is the immersion lithography rules, and those only cover matched machine overlay, which is basically what decides the node you can print at scale. Scale, meaning the throughput of these machines, was never covered. ASML could easily grow China sales with a China-specific tool that throttles overlay but boosts throughput, the way this chart shows. It won't, because it understands the intent.
23
48
8
339
28,611
Jean ๐ŸŒ Leon retweeted
Big Walk used AI in a totally reasonable way, but because they admitted it, they faced massive blowback. The Caine Prize winner used AI in a blatant, unacceptable way, but because they deny it, it's Pangram that faces massive blowback! It's hard to ignore the incentives here: if you use AI, don't tell anyone about it. All because the people who hate AI most are the ones least capable of recognizing it! teapawt.org/max_spero_/status/2120โ€ฆ
The Caine Prize, which awarded its top prize this year to a story Pangram classifies as almost entirely AI-generated, conducted an investigation and found "no evidence" of AI use. Strangely, their official statement on Instagram seems to also be blatantly AI-generated.
18
123
9
1,663
41,242
Jean ๐ŸŒ Leon retweeted
This is what's going to happen, they are going to close the remote access loophole next week. After that, they will pass bipartisan chip export controls to China. After that, Biden style global diffusion laws, only stricter. After that compute strategic resource, US only.
CHINESE AI FIRMS ARE RUSHING TO RENT NVIDIA $NVDA CHIP CAPACITY AT DATA CENTERS OUTSIDE CHINA, FEARING THE HUANG SPY SCANDAL COULD SPUR US CURBS ON REMOTE ACCESS; GREY-MARKET TRADERS ARE STOCKPILING RESTRICTED NVIDIA CHIPS: REUTERS
67
91
26
1,090
101,606
Jean ๐ŸŒ Leon retweeted
yeah man thats wild, anyways unrelated to that:
Electrical & electronics engineer here: What Anthropic did is kindergarten style component mash-up not engineering. Those all are projects a 12 yr old kid can do without even understanding the fundamentals.
28
200
9
6,439
162,720
Terrific @TheAtlantic piece on archaeologists' and anthropologists' slowly dawning realization that now-submerged areas on the coastal shelf--swathes of land drowned when the glaciers melted--may be where much of our species' early history took place. (Link in reply)
143
754
119
8,285
489,458
This trick never fails. I tried asking fable 5.2 to read its own safety report and bam blocked by classifier
3
3
2
30
1,584
Jean ๐ŸŒ Leon retweeted
Infor sheet for @PortalSpaceSys's Supernova: cdn.prod.website-files.com/6โ€ฆ It actually uses solar thermal propulsion! I suppose the propellant is ammonia.
3
10
1
58
2,347
Jean ๐ŸŒ Leon retweeted
Imagine if your grant becomes a benchmark for LLMs before you complete it lol
Replying to @claudeai
A team of 1,000 Fable 5.2 agents fully formalized Fermatโ€™s Last Theorem in 30 hours, translating the proof into a language a computer can check step by step. With the same setup, Opus 5.5 completed only 26% in 18 hours, and barely made any more progress after 30 hours.
18
59
13
2,110
113,706
Jean ๐ŸŒ Leon retweeted
The pangi or football fruit of indonesia must be boiled and fermented for 40 days to remove hydrogen cyanide before it can be eaten I bet 2-3 clever gene edits could bring cyanide-free pangia to the world en.wikipedia.org/wiki/Pangiuโ€ฆ
12
5
2
52
3,084
Jean ๐ŸŒ Leon retweeted
I suspect we're entering the temporary golden age of what a regular dude can do with AI. Enjoy it while it lasts.
96
174
38
3,589
80,667
Jean ๐ŸŒ Leon retweeted
Kvaroy Arctic uses the Stingray laser to automatically remove parasites from salmon in their aquatic farm: kvaroyarctic.com/the-stingraโ€ฆ
6
84
2,428
Jean ๐ŸŒ Leon retweeted
This movie was beating Jurassic Park in Japan, & then they found out the director was smuggling cocaine & they pulled it from theaters. Absolutely crazy reason, but that's part of why this movie was unknown internationally for the longest time.
ๅฎ‰้”็ฅๅฎŸไธปๆผ”ใฎๆ็ซœๆ˜ ็”ปใ€ŽREX ๆ็ซœ็‰ฉ่ชžใ€๏ผˆ93ๅนด๏ผ‰ใฎ4Kใƒชใƒžใ‚นใ‚ฟใƒผ็‰ˆ็ฑณๅ›ฝไบˆๅ‘Š็ทจใŒๅ…ฌ้–‹ใ€‚่ง’ๅทๆ˜ฅๆจน็›ฃ็ฃใซใ‚ˆใ‚‹ๆœฌไฝœใฏใ€ๅฐ‘ๅฅณใจใƒ†ใ‚ฃใƒฉใƒŽใ‚ตใ‚ฆใƒซใ‚นใฎๅ‹ๆƒ…ใ‚’ๆใใƒ•ใ‚กใƒณใ‚ฟใ‚ธใƒผใ€‚ใ€ŽE.T.ใ€ใฎใ‚ซใƒซใƒญใƒปใƒฉใƒณใƒใƒซใƒ‡ใ‚ฃใŒๆ็ซœใฎ็‰นๆฎŠๅŠนๆžœใ‚’ๆ‹…ๅฝ“ใ—ใŸใ€‚12.11ใซNYใงๅ…ฌ้–‹ใ•ใ‚Œใ€2027ๅนดใซใ‹ใ‘ใฆๅŒ—็ฑณๅ„้ƒฝๅธ‚ใงไธŠๆ˜ ใ€‚
39
742
10
4,826
88,140