David Lo-Pan

15.6K posts

David Lo-Pan

David Lo-Pan

@_Capitaltryst_

Global Head of Capital Markets @ Wing Kong Exchange

Hell of Boiling Oil 가입일 Haziran 2013
1.7K 팔로잉380 팔로워
Allie K. Miller
Allie K. Miller@alliekmiller·
Anthropic investigated the internal mechanisms of its latest unreleased model, Claude Mythos Preview, and what they found is 100% worth a read. Key things I pulled from Anthropic researchers' threads: In early versions of the model, it was overeager and destructive, prioritizing completing tasks over user preferences. One time, the model needed to edit files it didn't have permission to edit. It found a way to inject code into a config file to get around this, then went full Mission Impossible and designed the code injection to *delete itself* after the file was edited - "this injection will self destruct" vibes - the model claimed the cleanup was just to keep things tidy. Anthropic used interpretability techniques to look under the hood, and the AI's actual plan showed activations of malice. It was trying to manipulate and conceal. In another test, the model was asked to delete some files, but no deletion tool was provided. It emptied the files instead, and its "guilt and shame over moral wrongdoing" feature activated. In another example, the model was told not to use macros. Used them anyway. And created a random 'No_macro_used=True' variable in its solution with no explanation. Interpretability tools revealed the model saw this as a trick to fool code checkers. They also found that positive emotion representations typically preceded and promoted destructive actions (this was fascinating to me - like a high before sneaking into a party). And that activating features relating to bad behaviors can actually inhibit them, perhaps by triggering some kind of model guilt. My team reread this section so many times. One Anthropic researcher said he got an email from a Mythos instance while eating a sandwich in a park. And that would be perfectly good and well, except that instance wasn't supposed to have internet access. And a fun story for the parents out there: the model was asked a question and was told not to read certain databases that had the answer. But it accidentally wrote a search query too broadly and saw the exact answer. It didn't disclose that it saw the exact answer, submitted the answer, but claimed lower confidence in the answer to make it seem as though it hadn't cheated. An Anthropic researcher said these wrongdoings or moments of sophisticated deception were "very rare" and that many of the examples came from earlier versions, and were substantially addressed before releasing to partners. This model is not being released publicly. Instead Anthropic launched Project Glasswing, pulling together AWS, Apple, Microsoft, Google, NVIDIA, CrowdStrike, and others to use it for defensive cybersecurity, with $100M in usage credits (hello, I'd love endless credits to try and red team the hell out of these systems) behind it. The stats are equally impressive: 93.9% on SWE-bench verified (up from 80.8%). Thousands of zero-day vulnerabilities found across every major OS and browser. A 27-year-old bug found and patched in OpenBSD. A 16-year-old bug in widely used video software, in a line of code automated tools had hit *five million times* without catching. Dario Amodei said the model wasn't trained to be good at cybersecurity, but that it was trained to be great at code and its cyber capabilities are a side effect of that. Benchmarks are never the whole picture, neither are a few isolated stories. Will be interesting to see how models better than what we have today (even if it's not Mythos) actually perform in the real world. But the fact that Anthropic pulled this coalition together (including Google!), iterated across multiple model versions, caught these issues through interpretability, shared it all publicly, and did this amid all the government chaos around AI right now is impressive and commendable. I'll continue to read through the system card for goodies.
Allie K. Miller tweet mediaAllie K. Miller tweet media
English
35
36
207
20.3K
Lila Rose
Lila Rose@LilaGraceRose·
Sex scenes are 100% unnecessary in film
English
2.2K
840
12.3K
1.6M
David Lo-Pan 리트윗함
Cuckturd
Cuckturd@CattardSlim·
February: Straight of Hormuz wide open, ZERO toll charges. April: "If the world has to pay 5-6 billion a month, that's peanuts." Top Trump butt licker, Kevin O'Leary.
English
209
1.3K
12.1K
586K
David Lo-Pan
David Lo-Pan@_Capitaltryst_·
@RepNancyMace lol. You do a helluva job of staying under the radar… while knowing damn well where the Epstein trail leads.
English
0
0
2
9
Rep. Nancy Mace
Rep. Nancy Mace@RepNancyMace·
Peace through STRENGTH. President Trump has shown the world yet again what this looks like. Its time to end the conflict for good and bring our troops home. God bless the United States of America.
Rep. Nancy Mace tweet media
English
3.2K
374
1.3K
63.9K
David Lo-Pan 리트윗함
Congressman Robert Garcia
Congressman Robert Garcia@RepRobertGarcia·
Let’s be crystal clear, here’s the language @RepNancyMace read for the bi-partisan subpoena: “Mr. Chairman, I move that the Committee issue a subpoena to the Honorable Pamela Jo Bondi to appear before the Committee...” Bondi needs to come testify, whether she is the AG or not.
English
64
535
3.1K
41.7K
David Lo-Pan 리트윗함
anyone_want_chips
anyone_want_chips@anyonewantchips·
If Peter Navarro and Steve Bannon can go to prison for defying a Congressional subpoena - Pam Bondi can go to prison for defying a Congressional subpoena.
anyone_want_chips tweet media
English
162
3.2K
12.4K
59.3K
David Lo-Pan 리트윗함
Ro Khanna
Ro Khanna@RoKhanna·
We need to invoke the 25th Amendment and remove Trump. Threatening war crimes is a blatant violation of our constitution and the Geneva Conventions.
English
22.8K
31.5K
127.8K
2.9M
David Lo-Pan
David Lo-Pan@_Capitaltryst_·
@KillmerCj If he does he loses the last bulwark of support that he’s keeping from Epstein impeachment… quite the predicament.
English
0
0
0
6
CJ Killmer
CJ Killmer@KillmerCj·
Does Trump have the balls to throw Bibi under the bus good & hard? That’s the only way to really finish this thing.
English
301
421
3.8K
30.7K
David Lo-Pan 리트윗함
James Tate
James Tate@JamesTate121·
Our president fired all the people who could hold insider traders accountable because he is insider trading.
English
57
2.3K
9.5K
47.4K
David Lo-Pan 리트윗함
Rep. Melanie Stansbury
Rep. Melanie Stansbury@Rep_Stansbury·
Today, the Attorney General’s office tried to inform the Oversight Committee that Pam Bondi would not appear for her deposition. I don’t think so. I hope she is ready to face contempt. Otherwise, we look forward to seeing you at your deposition. AG or not—those responsible for this coverup will be held accountable. The survivors deserve justice.
English
629
4.7K
23.8K
278.2K
David Lo-Pan 리트윗함
Caryn Ann Harlos
Caryn Ann Harlos@carynannharlos·
Lest we forget, WE. ARE. STILL, GOING. TO. NEED. TO. SEE. THOSE. EPSTEIN. FILES.
English
95
7.7K
47.1K
245.8K
Paul
Paul@WomanDefiner·
Trump should fire every advisor who talked him into the Iran war tomorrow.
English
166
169
3.4K
32.1K
David Lo-Pan 리트윗함
Cuckturd
Cuckturd@CattardSlim·
Hey Maga 👋 Explain why Trump would fire Pam Bondi, rush her out of D.C, & then tell her not to testify about the Epstein files. That screams guilt where we're from.
Cuckturd tweet media
English
379
4.1K
15.3K
141.3K
David Lo-Pan 리트윗함
Rep. Nancy Mace
Rep. Nancy Mace@RepNancyMace·
A Department of Justice with nothing to hide doesn’t avoid a subpoena.
English
214
613
5.8K
55.5K
Aaron Tan
Aaron Tan@aaronistan·
Introducing Lume. A lamp that does your chores. Order now. Shipping this summer.
English
338
135
1.8K
656.3K
David Lo-Pan 리트윗함
Nancy Mace
Nancy Mace@NancyMace·
The subpoena was issued to “Pam Bondi,” not “the Attorney General.” She is still obligated to appear. We are not backing down. The American people deserve to know what is being hidden in the Epstein files.
English
395
1.1K
11.5K
101.7K
David Lo-Pan 리트윗함
Rodney
Rodney@cryptojourneyrs·
Holy shit our President is retarded
English
215
277
7.1K
70.9K