[lkm]

Title VII of the US Civil Rights Act governs discrimination in hiring. Prior studies found very limited adverse impact in algorithmic hiring data as a whole. We surface previously-overlooked adverse impact by studying positions separately. Black applicants are the most likely to be adversely impacted: 30% of Black applicants apply to at least one position that demonstrates adverse impact against Black applicants. In terms of total effect, Asian applicants experience the largest shortfall: if Asians were selected at the same rate as the most selected racial group for each position, then 29000 additional Asian applications would be recommended.

Algorithmic Monocultures in Hiring

[lkm]

What’s going on today is like if — instead of getting shady, threatening, and litigious — the tobacco companies had just purchased CBS, purged the staff at 60 Minutes, and hired a bunch of pro-cigarette stooges to replace them.

Daring Fireball: 'The Insider'

[lkm]

It's also interesting in that it hints at a real dollar value for what Uber is getting out of these tools. If we assume two actively used tools per engineer that's $3,000 * 12 = $36,000 cap per engineer per year. Levels.fyi lists the median yearly compensation package for Uber software engineers in the USA at $330,000. That means each employee's AI spending cap is ~11% of that median compensation package.

That seems about consistent with the benefit LLMs provide for real development work. Uber Caps Usage of AI Tools Like Claude Code to Manage Costs

[lkm]

Ukrainian officials say the Army of Drones Bonus system, or ePoints, is the only program of its kind in the world. Units earn points for each Russian soldier they incapacitate or kill and each weapon, vehicle or piece of military equipment they destroy. Points may be redeemed in an online government marketplace for more drones, with which to target more Russian forces.

Ukraine turns war into a video game, with points and prizes for kills - The Washington Post

[lkm]

They cut down elms to build asylums for people driven mad by the cutting down of elms.

Accommodation - Futility Closet

[lkm]

I used to describe myself as American, but that’s becoming less true with time. I’m from the America of the 80s, 90s and early 2000s. But that place is long gone. It’s not like that anymore.

Geography is four-dimensional | Derek Sivers

[lkm]

What’s going on? I think this shows how easy it is to fool yourself about the quality of prompting. When the model is already pretty good at a task, you can give it a very elaborate prompt without impacting performance. It’ll still be pretty good, except this time it’s good because of what you did. This is particularly true if you’re iterating with the model and asking it “what should I add to the prompt” for each mistake. Models will happily make up stories for you about their own reasoning processes, and will almost always say “yes, that helped a lot!” when you ask them if a particular prompt tweak made things better. The only way to actually know is by constructing some kind of benchmark.

The famous o3 "GeoGuessr" prompt did not work

[lkm]

1 square yard costs 18 pence. Therefore 1 yard costs \sqrt{18} , or 4.243, pence. Therefore 1 foot costs 4.243 ÷ 3 = 1.414 pence. Therefore 1 square foot costs 1.4142 = 2 pence.

"A Dimensional Howler" - Futility Closet

[lkm]

The undersigned organisations condemn Meta’s recent decision to restrict the Facebook and Instagram accounts of independent NGOs, researchers, and civil society figures from reaching audiences in Saudi Arabia and the United Arab Emirates (UAE). This is the latest in a pattern of major technology companies, including Meta, acting as enforcement arms for repressive governments in the Gulf. We call on Meta to uphold its human rights responsibilities and ensure that users’ content is not arbitrarily restricted.

Meta blocks human rights accounts from reaching audiences in Saudi Arabia and the UAE

[lkm]

Microsoft’s GitHub was compromised when a Microsoft developer using Microsoft VSCode installed a rogue extension from Microsoft’s VSCode extension library, which is moderated and hosted by Microsoft.

GitHub (@github): "1/ We are sharing additional details regarding our investigation into unauthorized access to GitHub's internal repositories. Yesterday we detected and contained a compromise of an employee device involving a poisoned VS Code extension. We removed the malicious extension version, isolated the endpoint, and began incident response immediately." | XCancel

[lkm]

Before this first Mythos report, we had already scanned curl with several different very capable AI powered tools (I mean in addition to running a number of “normal” static code analyzers all the time, using the pickiest compiler options and doing fuzzing on it for years etc). Primarily AISLE, Zeropath and OpenAI’s Codex Security have been used to scrutinize the code with AI. These tools and the analyses they have done have triggered somewhere between two and three hundred bugfixes merged in curl through-out the recent 8-10 months or so. A bunch of the findings these AI tools reported were confirmed vulnerabilities and have been published as CVEs. Probably a dozen or more. (...) The report concluded it found five “Confirmed security vulnerabilities”. I think using the term confirmed is a little amusing when the AI says it confidently by itself. Yes, the AI thinks they are confirmed, but the curl security team has a slightly different take. Five issues felt like nothing as we had expected an extensive list. Once my curl security team fellows and I had poked on the this short list for a number of hours and dug into the details, we had trimmed the list down and were left with one confirmed vulnerability. The other four were three false positives (they highlighted shortcomings that are documented in API documentation) and the fourth we deemed “just a bug”.

Mythos finds a curl vulnerability | daniel.haxx.se

[lkm]

Every time I prompted Claude for a feature, it delivered. Perfectly. The fleet view worked on the first try. Log streaming worked. Mouse support worked. The problem is that each feature was implemented in the context of "make this work right now" without any awareness of the 49 other features sharing the same state.

Im going back to writing code by hand | k10s devlog

[lkm]

“Yes, you can just demand that the LLM not make errors,” journalist Karl Bode joked in a mocking Bluesky post. “That’s definitely how the technology works.” “I know this isn’t a unique observation but these gentlemen are in absolutely no way remarkable outside of their good fortune,” Bode added. “Marc Andreessen putting ‘you are a world class expert in all domains’ and ‘don’t hallucinate’ in his custom prompt really demonstrates the calibre of the people steering the ship,” another user added.

The world's dumbest people are controlling the world's most dangerous technologies. Marc Andreessen Mocked for Accidentally Revealing That He Seems to Have a Deep Misunderstanding of How AI Actually Works

[lkm]

This recall itself involves a potentially serious safety issue, namely that the truck’s wheels could fall off. However, the problem isn’t with the steel wheels themselves; it’s with the brake rotors. The holes for the wheel studs in the rotor can begin to crack due to “road perturbations” and stress on the hub assembly during cornering. If the owner doesn’t notice the cracking, which can easily be hidden behind the wheels, it could eventually lead to total failure, causing the wheel studs to separate from the hub and the wheel to come off.

Cool cool cool cool. Only 173? Tesla Cybertruck RWD Recall Reveals Shockingly Low Sales

[lkm]

This is the part of the phenomenon I find hardest to write about. The tool did not make him a worse colleague. It made him able to impersonate, for months, a discipline he had never trained in, and the impersonation was good enough that the institutional incentives all bent toward letting him continue. Perhaps it’s a failure of management, but I have been finding management to be so eager to embrace AI that they’re willing to accept the risk.

Appearing Productive in The Workplace — No One's Happy

[lkm]

$25 billion is similar to: (...) Spending by U.S.A.I.D. before it was disbanded.

Daring Fireball: The Pentagon Pegs the Cost of the Iran War, So Far, at $25 Billion

[lkm]

Well, what I mean by Education is learning the rules of this mighty game. In other words, education is the instruction of the intellect in the laws of Nature, under which name I include not merely things and their forces, but men and their ways; and the fashioning of the affections and of the will into an earnest and loving desire to move in harmony with those laws. For me, education means neither more nor less than this.

Thomas Huxley, via Training - Futility Closet