Alien_AV

@SobonBet @BigTechnoFuture @ape_detective @bscholl Those are among the highest-paying jobs on the market, intentionally, to attract the global top talent. Not every H1B is like that, but the researchers in question are.

@BigTechnoFuture @ape_detective @bscholl Sorry, I refuse to believe that there are not Americans that can do these things. The Tech Companies just don't want to pay them.

I understand why we don’t want people to come to the US to be criminals, mooch on welfare, open learing centers and otherwise undermine the country. But I don’t understand why we make it harder for motivated, ambitious, hardworking people to come to the land of opportunity.

@smchatter1 @GaryMarcus "produce more". in the end products get sold to people, that's why there's marketing involved 🤷🤷🤦 what's the point of producing if nobody knows they should consume it?

@GaryMarcus The objections of the Pro side tells the tale. To win they need just to produce more, instead it's marketing chest beating. Which is really the whole problem. Good tech marketed as something well beyond what it can do.

can’t believe people assume that success on highly verifiable problems in math (where we don’t even know how many tests were performed and how many might have failed) iautomatically generalize to everything else when there is not a shred of evidence that they do. 🤷‍♂️

@FeanFiltonpair @GaryMarcus You're just salty you didn't do it.

@GaryMarcus People have the wrong perception of those 'breakthroughs'. This is not a rando writing a prompt 'resolve some erdos'. Highly specialized R&D teams scaffolded the LLM so it had the luck to find an existing "answer" (from another math field). Still, impressive 125 pages of CoT

@US_Macauley @allTheYud You're looking past even the "evidence" you're providing. You brought up the topic of relying on data, and are supplying instance after instance of data that doesn't support your conclusion and doesn't even relate to it. Is this intentional?

@Alien_AV @allTheYud Lol, what evidence am I looking past? I didn't bring the topic up.

Every month a new guy discovers LLMs; discovers a skill the current LLMs require to get good results; and writes about the future jobs that will always be available for smart people like HIM, that are SKILLED in using LLMs. The next generation of AIs doesn't need his fancy prompt. The image model goes from needing to type in just the right set of weird words and cryptic sorcerous invocations, to most people being able to type in English what they want and get a pretty good result. There are still tasks that require careful invocation. But they are a much smaller fraction of all the tasks people are trying to do, or you can get a bleh result without the elaborate invocation to get it really good. And to improve on the bleh result you need to be substantially more of an expert than back when the Guy was memorizing a rule about adding "trending on Artstation" to the image prompts, as would always require a human paid to do that. Another generation of AIs comes out. The next generation of Clever Skills is obsolete. Image models just obey the instructions for compositing panels without mixing them up, and you don't need to be an expert to get them to do it right. Another human value-add is gone. A wider set of tasks require no human expert. Now a new Guy notices LLMs have become useful in his field for the first time. He discovers they require SKILL to use CORRECTLY. He posts about how there will always be jobs for humans who are SKILLED in using LLMs like HIM. But it is not an infinite cycle. It is not the same each time it repeats. Now the Guy is a highly paid programmer or a career mathematician in 2026, instead of a graphic artist in 2023. In six months the models will no longer require his vaunted Skills. And by then there will be another Guy. But the process doesn't continue forever. The Guys are coming from fields that were harder and harder for AIs. The brief centaur eras are shorter and shorter. Today it is writers who are laughing at how bad the LLMs are at their job, and who will perhaps soon be posting about how it takes Skill to get an LLM to do their job Correctly. But the models are coming faster, and the eras of kinds of human value-add in each field are shortening. There is a point when you run out of Guys, either because the centaur eras are too short for people to develop SKILLs and post to Twitter about them; or because there are not lands left for AIs to conquer; or because ordinary people are not reassured by some Nobel laureate proclaiming there will always be jobs for Nobel laureates with the SKILLS to prompt robotized biology labs Correctly. But we'll never run out of amateur economists who assert entirely *without* a brief contemporary example that there will always be jobs for humans skilled at operating AIs! We'll run out of professional economists saying it when nobody is paid for that work anymore. I guess we'll also run out of amateur economists when they're dead.

The first graph, again, seems like it ends in 2023, with a projection outwards. You're not trying really hard to rely on data here, unless you think that it means to bring any graphs that associate with the same keywords, regardless of whether they prove, disprove, or are agnostic to the point you make. You obviously want something to be true and are willing to look past the evidence on that, but don't want to be seen looking past the evidence on the first glance. ("It was just a low-effort twitter post", aren't they all?)

@Alien_AV @allTheYud As I provided elsewhere, you can look at more recent data for those too. x.com/US_Macauley/st…

From his behavior before it was absolutely not surprising that this is the kind of a pivot that he'd do. (It's luck that there's any pivot at all.) A strong personality can come forward and say: I admit I can be wrong, and was wrong, that's the power of human intelligence, to understand when one is mistaken, correct it, and come away smarter. A weak person can't fathom showing they were wrong even for a second. Maybe they're afraid that their pretence of being "an important smart guy" will be exposed.

At this point its just kinda sad to see him go like this.

GOALPOST MOVED. "LLM must solve major conjecture without lean or data augmentation using lean. Otherwise its neuro symbolic AI, just as I predicted years ago. Gotcha."

2/3 graphs you provided aren't good, because the expected drop in translator employment was not supposed to happen within the timeframe they describe. 3 years time difference is full generations of evolution in current models capabilities. Don't you think it's a problem with your method? As an empiricist, you shouldn't be allowing such a sloppy approach to data in your arguments?

@Alien_AV @allTheYud The first one is through 2025, I provided other more recent data in other graphs. But FRED is the most reliable.

I'll quote yourself back to you again, last time was in the very post you're replying to: "It would be up to the anti-Nazi Germans to condemn & disassociate themselves with Nazi Germany", and "The problem is many Jews are silent". You *are* assigning collective blame for Germans and Jews. And then you're trying to weasel out of it. Well, the words are on the page. I'm not the one making normative statements at all, I was just asking you if your expectations apply universally (unsurprisingly, they didn't). The only thing left for you to do is identify (personally, with yourself) where the gap comes from.

@Alien_AV @JonahPlatt I am rejecting collective blame for anyone, be them Jews or Muslims. You are demanding it from Muslims but reject it for Jews/Israel. The inconsistency is from you. The principle is simple: collective guilt is either wrong for everyone or for no one.

I constantly see people on this platform blame Israel for making Jews' lives worse. Here's your reminder of what happened within days after October 7th — before Israel had even finished cleaning up their dead, while hundreds of innocent men, women, and children were being held hostage in tunnels underground. • Vandals smashed a kosher restaurant in London. • A synagogue was desecrated in Porto. • Hundreds outside the Sydney Opera House screaming, "Gas the Jews." • Crowds in Geneva chanting "Khaybar, Khaybar, ya yahud!" — a reference to a 7th-century massacre of Jews in Saudi Arabia. • A rabbi's home in British Columbia was covered with Nazi graffiti. • A synagogue was set ablaze in Tunisia. • The only mezuzah-marked door in a Paris apartment building was doused with gasoline and set on fire. • Molotov cocktails were hurled at a synagogue in Berlin. This is what happened before Israel launched a single airstrike in response. The idea that anti-Jew hate is Israel's fault — that Jews would be safe if Israel just behaved differently — is not just wrong. Blaming Jews for hate against us is the oldest lie in the book. They don't hate us because of what Israel does. They hate us already and need something convenient to blame it on.

@US_Macauley @allTheYud The graphs end at 2022 🤷

@Alien_AV @allTheYud How do they not? Hiring has not fallen to zero. Translation services and graphic design have still been going, though certainly impacted by AI developments.

@koanchuk @liron @jeremyscahill Does he have a name? Like "Yahya Sinwar" or something? Or is he just your personal "face of Palestine"? (If so, is there a reason to select Ben Gvir as ones personal face of Israel, and not, say, the Jewish neighbors of the guy from the plane who he famously got along with?)

@liron @jeremyscahill the guy I sat next to on the plane once, who told me that he was getting along famously together with his Jewish neighbours until one day some extremists showed up and decided to ruin his life for no good reason

Narrator: He is, in fact, the face of Israel.

@gidonsaar Ben Gvir belongs in prison for treason.

You knowingly caused harm to our State in this disgraceful display - and not for the first time. You have undone tremendous, professional, and successful efforts made by so many people - from IDF soldiers to Foreign Ministry staff and many others. No, you are not the face of Israel.

@US_Macauley @allTheYud You're not serious. This is not the data that supports your conclusions. It's hypocritical to say "you're not following the data" and to post graphs that don't contain the data in question.

@allTheYud Flat or slowly declining over scale of years. Even if you just look at AI adopting firms, there isn't a massive drop. Just a *gradual* decrease in previous growing junior hiring.

You said "it would encourage aggression against all Germans. It would be up to the anti-Nazi Germans to condemn & disassociate themselves with Nazi Germany", and "The problem is many Jews are silent". But this doesn't extend to Muslims and Islamic terrorism for some reason. As if your position is not based on principles, but rather on special treatment for specific religions/ethnicities. You don't even notice how you turn 180° inside a single conversation. You really should think through what your principles are. "Muslims good, Jews bad" is not a good moral position, but at least you can be honest and state it upfront.

@Alien_AV @JonahPlatt Oh sure, Muslims must hold weekly anti‑terror marches or they "don’t count"… but when Christian or far‑right extremists attack people, suddenly everyone forgets how to demand street protests. Funny how the outrage meter only works in one direction.

The important takeaways I got from my time as a dev team lead were that delegation is inevitable, "code quality" will never match my idealistic expectation, and I will be owning code that I didn't make, and often code I never looked at. Best I can do is put systems in place that reduce the risk, and help produce quality result from the huge amounts of code I could never manage to personally vet. Many of the systems that mitigate human-sourced risk, translate well into mitigating AI-sourced risk.

@Alien_AV @JoelAbenhaim @brunoborges company activity, knowledge, and control over their prod to ai.

As you write code by hand, you intrinsically review it. As you gain experience by writing (and thus reviewing), your code gets better. Your reviewing skill gets better. You spot your own mistakes faster, and address them as you write and rewrite the code. AI written code is a code dump. Reviewing AI code dumps is as hard as reviewing a very long PR someone sent your way. Expecting humane understanding of AI written code is unfathomable in many cases. Only using AI to review AI generated code seems plausible.

@Zapata18791919 @JonahPlatt One would think they'd see some Muslim anti-Islamic-terrorism protests, if your expectation from Muslims to condemn and disassociate from said terror would hold.

@Alien_AV @JonahPlatt Wtf are you talking about?Most Muslims do condemn extremist terrorism. Large majorities in global surveys reject attacks on civilians& mainstream Islamic scholarship forbids such violence.Extremist groups represent a tiny minority &are widely seen as violating Islamic principles

For example, when people talk about guardrails, that usually shows they're not in the right state of mind. Usually that amounts to just asking the model real hard to write good code (and the model fully forgets even this naive request after ~250k tokens). Test harnesses, multi-tier automated review, AI-driven "manual testing", and evolving the mutual understanding of the desired goal (instead of one-shot ad-hoc prompting), all improve the code generation result, to the level that I barely look at the code, and spend 95% of the effort on the structure.

@Alien_AV @JoelAbenhaim @brunoborges 1. Which skill? 2. No because I use it too and know very well what it can do with good context but also it's limits, so I wasn't really surprised to find slop. You absolutely need to review the code seriously OR accept slop (which is sometimes ok). That's just sota rn 🤷‍♂️

@jaggy_twit @JoelAbenhaim @brunoborges Did it occur to you that it might have been a skill issue?

@JoelAbenhaim @brunoborges I work with a guy like you. Talks about guardrails and contextand high quality. Today I had to make a little modification so dumb that I thought "let's do it by hand so I can check this guardrailed quality output". Absolute slop. The only skill you're truly developing is lying to

@Zapata18791919 @JonahPlatt Is it up to all Muslims to condemn and disassociate themselves from islamic terror? You seem to have missed that part for some reason.

@Alien_AV @JonahPlatt Sure, many ignorant people associate the 1.9 billion Muslims in the world with terrorism, when in reality they only number in the tens of thousands. Many of these groups like ISIS & Al Qaeda are creations of the west, armed, trained & funded to be used as terrorist proxy forces.

@Zapata18791919 @JonahPlatt Can you do Islamists next?

@Alien_AV @JonahPlatt Of course not. If Nazis continously associated themselves with all Germans, it would encourage aggression against all Germans. It would be up to the anti-Nazi Germans to condemn & disassociate themselves with Nazi Germany.

@haider1 Skill issue.

Creator of C++, Bjarne Stroustrup: AI-generated code isn't ready — it generates more bugs, more bloat, more security holes, and is nearly impossible to validate "senior developers are already retiring rather than deal with it" The problem is that even a small prompt change can shift the entire codebase in unpredictable ways