https://www.windowscentral.com/microsoft/windows-11/my-goal-is-to-eliminate-every-line-of-c-and-c-from-microsoft-by-2030-microsoft-bets-on-ai-to-finally-modernize-windows

My goal is to eliminate every line of C and C++ from Microsoft by 2030. Our strategy is to combine AI *and* Algorithms to rewrite Microsoft’s largest codebases. Our North Star is “1 engineer, 1 month, 1 million lines of code”. To accomplish this previously unimaginable task, we’ve built a powerful code processing infrastructure. Our algorithmic infrastructure creates a scalable graph over source code at scale. Our AI processing infrastructure then enables us to apply AI agents, guided by algorithms, to make code modifications at scale. The core of this infrastructure is already operating at scale on problems such as code understanding."

wow, *and* algorithms? i didn't think anyone had gotten that far

In the future, I'm going to add "at scale" to the end of all my fortune cookies.

idea: end of year worst of ai awards. "the sloppies"

"Google Gemini Pro: please generate a map of the western united states showing cities of over 250K population"

lol, Oliver Habryka at Lightcone is sending out begging emails, i found it in my spam folder

(This email is going out to approximately everyone who has ever had an account on LessWrong. Don't worry, we will send an email like this at most once a year, and you can permanently unsubscribe from all LessWrong emails here)

declared Lightcone Enemy #1 thanks you for your attention in sending me this missive, Mr Habryka

In 2024, FTX sued us to claw back their donations, and around the same time Open Philanthropy's biggest donor asked them to exit our funding area. We almost went bankrupt.

yes that's because you first tried ignoring FTX instead of talking to them and cutting a deal

that second part means Dustin Moskovitz (the $ behind OpenPhil) is sick of Habryka's shit too

If you want to learn more, I wrote a 13,000-word retrospective over on LessWrong.

no no that's fine thanks

We need to raise $2M this year to continue our operations without major cuts, and at least $1.4M to avoid shutting down. We have so far raised ~$720k.

and you can’t even tap into Moskovitz any more? wow sucks dude. guess you’re just not that effective as altruism goes

And to everyone who donated last year: Thank you so much. I do think humanity's future would be in a non-trivially worse position if we had shut down.

you run an overpriced web hosting company and run conferences for race scientists. my bayesian intuition tells me humanity will probably be fine, or perhaps better off.

I will never forgive Rob Pike for the creation of the shittiest widely adopted programming language since C++, but I very much enjoy this recent thread where he rages about Anthropic.

Remember how slatestarcodex argues that non-violence works better as a method of protest? Turns out the research pointing to that is a bit flawed: https://roarmag.org/essays/chenoweth-stephan-nonviolence-myth/

AI researchers are rapidly embracing AI reviews, with the new Stanford Agentic Reviewer. Surely nothing could possibly go wrong!

Here's the "tech overview" for their website.

Our agentic reviewer provides rapid feedback to researchers on their work to help them to rapidly iterate and improve their research.

The inspiration for this project was a conversation that one of us had with a student (not from Stanford) that had their research paper rejected 6 times over 3 years. They got a round of feedback roughly every 6 months from the peer review process, and this commentary formed the basis for their next round of revisions. The 6 month iteration cycle was painfully slow, and the noisy reviews — which were more focused on judging a paper's worth than providing constructive feedback — gave only a weak signal for where to go next.

How is it, when people try to argue about the magical benefits of AI on a task, it always comes down to arguing "well actually, humans suck at the task too! Look, humans make mistakes!" That seems to be the only way they can justify the fact that AI sucks. At least it spews garbage fast!

(Also, this is a little mean, but if someone's paper got rejected 6 times in a row, perhaps it's time to throw in the towel, accept that the project was never that good in the first place, and try better ideas. Not every idea works out, especially in research.)

When modified to output a 1-10 score by training to mimic ICLR 2025 reviews (which are public), we found that the Spearman correlation (higher is better) between one human reviewer and another is 0.41, whereas the correlation between AI and one human reviewer is 0.42. This suggests the agentic reviewer is approaching human-level performance.

Actually, now all my concerns are now completely gone. They found that one number is bigger than another number, so I take back all of my counterarguments. I now have full faith that this is going to work out.

Reviews are AI generated, and may contain errors.

We had built this for researchers seeking feedback on their work. If you are a reviewer for a conference, we discourage using this in any way that violates the policies of that conference.

Of course, we need the mandatory disclaimers that will definitely be enforced. No reviewer will ever be a lazy bum and use this AI for their actual conference reviews.

Sean Munger, my favorite history YouTuber, has released a 3-hour long video on technology cultists from railroads all the way to LLMs. I have not watched this yet but it is probably full of delicious sneers.

I posted about Eliezer hating on OpenPhil for having too long AGI timelines last week. He has continued to rage in the comments and replies to his call out post. It turns out, he also hates AI 2027!

https://www.lesswrong.com/posts/ZpguaocJ4y7E3ccuw/contradict-my-take-on-openphil-s-past-ai-beliefs?commentId=3GhNaRbdGto7JrzFT

I looked at "AI 2027" as a title and shook my head about how that was sacrificing credibility come 2027 on the altar of pretending to be a prophet and picking up some short-term gains at the expense of more cooperative actors. I didn't bother pushing back because I didn't expect that to have any effect. I have been yelling at people to shut up about trading their stupid little timelines as if they were astrological signs for as long as that's been a practice (it has now been replaced by trading made-up numbers for p(doom)).

When we say it, we are sneering, but when Eliezer calls them stupid little timelines and compares them to astrological signs it is a top quality lesswrong comment! Also a reminder for everyone that I don't think we need: Eliezer is a major contributor to the rationalist attitude of venerating super-forecasters and super-predictors and promoting the idea that rational smart well informed people should be able to put together super accurate predictions!

So to recap: long timelines are bad and mean you are a stuffy bureaucracy obsessed with credibility, but short timelines are bad also and going to expend the doomer's crediblity, you should clearly just agree with Eliezer's views, which don't include any hard timelines or P(doom)s! (As cringey as they are, at least they are committing to predictions in a way that can be falsified.)

Also, the mention about sacrificing credibility make me think Eliezer is intentionally willfully playing the game of avoiding hard predictions to keep the grift going (as opposed to self-deluding about reasons not to explain a hard timeline or at least put out some firm P()s ).

Starting this Stubsack off, here's Baldur Bjarnason lamenting how tech as a community has gone down the shitter.

elsewhere on lemmy, a piece from the atlantic (be warned: they quote lasker/cremieux for some reason) on new shiny glp-1 agonist that you can order off telegram from some random ass chinese lab:

The tests, insofar as they are reliable, do flag problems. According to Finnrick Analytics, a start-up that provides free peptide tests and publicly shares the results, 10 percent of the retatrutide samples it has tested in the past 60 days had issues of sterility, purity, or incorrect dosing. Two other peptide-testing labs, Trustpointe and Janoshik, have said in interviews with Rory Hester, a.k.a. PepTok on YouTube, that they see, respectively, an overall fail rate of 20 percent and a 3 to 5 percent fail rate for sterility alone across all peptides.

isn't dear leader EY taking this? it's still not approved yet, so it's not available on normal market, and because it's peptide it's i.m. only. also, side effects not just for this one, but for entire class include anhedonia, which must be very rational thing to risk without medical need. chat, what's your p(infected sore on EY's ass)

odium symposium christmas bonus episode: we watched and reviewed Sean Hannity's straight-to-Rumble 2023 Christmas comedy "Jingle Smells."

https://www.patreon.com/posts/146610014

Gemini helps a guy increase his google cloud bill 18x

https://news.ycombinator.com/item?id=46378065

Sunday rant post. I really dislike that so many people are now adopting 'electrons' when they mean power (it is good as a 'this person drinks the coolaid' shibboleth however).

And I was amused to hear people go 'AI (by which they meant the recent llm stuff) malware creating will be a risk in the future, look at the drug discovery that AI is already doing', wonder if drug discovery people have said 'look how great drug discovery will be in the future, look at all the malware development AI is already doing'.

Ask HN: Can you patent prompts?

https://news.ycombinator.com/item?id=46365116