I’ve been wondering about something similar - a system that enforces (or does the heavy lifting) of dividing a large task into smaller sub-tasks so that it’s easy to run/check/test each one independently - even on a fresh model instance if needed.
This is based on the observation that the medium-sized open weight models (~20-35b) are very able to one-shot smaller discrete tasks but seem to lose their way project managing themselves through larger tasks that have multiple steps.
Let me have one more go. I'm 100% not trying to trigger or upset you. This is the chain of logic I'm trying to argue for, in very clear blunt terms.
1. Germany's economy and future is threatened by its demographics, and uniquely so due to the nature of its industry, and the need for skilled workers to maintain its economy.
2. Germany will therefore need to compete to attract these skilled workers against other countries.
3. Germany has strengths but also several weaknesses relative to other countries (tax burden, bureaucracy, public infrastructure, relative cost of living (vs. some not all competitors, note), language barrier and culture around language, etc.) in attracting such skilled workers.
4. To succeed at attracting the very best of the skilled workers it needs, it could/should therefore embark on a bold plan to address all of its weaknesses and make itself the most attractive country for skilled immigrants in the world.
5. Unfortunately for Germany, the standard language of business around the world is English. (If e.g. Chinese people want to learn or work internationally, they speak English professionally, etc.). If an English-speaker has a choice of two countries, one English-speaking and one German-speaking, then if all else is equal, they'll mostly not pick Germany. And some may be put off Germany altogether due to the relative intolerance of non-German speakers.
So, how best to address the language question?
If it was my country at threat, and we needed to attract a large number of a specific type of worker speaking a different language, (amongst other measures) I'd be advocating for removing the language issue as a barrier: for example, starting to teach Mandarin, or Hindi, or whatever language made most sense, in schools. Offering free adult education to learn those languages. Offering direct language support to businesses most needing to attract those skilled foreigners. Etc.
In Germany's case, to address this weakness, I'd advocate adopting English as a co-official language, if not the official language. Pour money and support into language training. Offer tax incentives for individuals and companies who become primarily English-speaking. Offer tax incentives and other support for English-speaking companies to part-relocate. And so on.
(Side note: this isn't a pro-English argument per se. If Germany decides it can more easily recruit the immigrants it needs from China, make Mandarin the co-official language. IDC.)
This post can essentially be distilled down to: yes, Fable's classifier (which is meant to downgrade cybersecurity, biology, or jailbreak attempts to Opus 4.8) is definitely overly sensitive to the point of uselessness.
e.g. a colleague asked Fable to help create an simple app to help calculate the statistics for phase II and III trials. (Ignoring that such things already exist) it passed his request down to Opus, despite only being very marginally, tangentially, somewhat related to biology.
> If you want to make me responsible for the fate of Germany, make me chancellor, or get off my back.
Suggesting that you’re not responsible for your opinions because you don’t have the ultimate power to change the country? Don’t be childish. Maybe if you can’t cope with being disagreed with, stop posting opinions on internet forums?
> What more should I do? Berate other Germans to be "less racist" or "more accepting" until people who simply just might by entitled assholes stop being that? Nah. Over it.
Well, you could be the change you want to see? Part of the problem that others are diagnosing in Germany is the attitude of some Germans; and you’re part of that group, no? Accepting this would be a small step forwards.
> And do you notice how it's just normal to talk shit about Germans and Germany, but when Germans speak _their_ mind, and say hey wait a second, that's rude, and it would be rude if we did that, too -- that's worth nothing? You wouldn't like to be treated that way.
For my part, you’re welcome to speak your mind, but that’s not really the issue. The previous poster and I have both given you evidence-based positions on specific weaknesses that Germany has, and suggested that the government and ultimately the people might want to address them for their own benefit. If you can’t see well-meaning, informed, constructive criticism being anything other than people “talking shit about Germans and Germany”, then… eh… there’s not really any point to discussing further.
(And you’ve still avoided addressing the well-meaning, informed, constructive criticisms.)
The indigenous Swiss have the same attitude; it's just that in the major cities and companies, the concentration of English-speakers is so high, it's not such a problem.
You seem to be completely missing (or avoiding?) the point.
It is widely recognised that Germany is one of the most at-risk countries in the world regarding its aging population, low birth-rate, and need for skilled workers for its economy.
If one performs a SWOT analysis of Germany from the perspective of a potential skilled immigrant, alongside the undoubted strengths, there are relative weaknesses - e.g. high taxes, awful bureaucracy (especially for outsiders and entrepreneurs), low levels of digitalisation, failing public infrastructure, and a language barrier vs. the otherwise-English-speaking professional world.
A rational response to this would be to address each weakness at all levels, from government to culture to individuals, because Germany is in competition with many other countries for the same skilled workers.
In short, put a process in place to make Germany the most skilled-immigrant-friendly country in world. And a piece of this process would be to be less intransigent around language - be much more accepting of people who want to settle in Germany and contribute positively to its future but don't want to spend large amounts of time learning a difficult new language which is irrelevant almost anywhere else.
> If someone acts entitled and assumes I have to accommodate their laziness forever, not just until they'll get settled in, I won't like them, and there won't be much warmth between us. If that dooms the nation, that's totally fine.
This is a very odd position: you're literally saying you'd rather have your country fail, fall apart, die, than get over your hurt feelings about people not wanting to speak the language fluently. (If Germany ceases to exist as a meaningful country in a few generations, few will care about speaking German then either.)
Individual Germans on the whole are very efficient, possibly because they have to be to survive the awful sclerotic bureaucracy in Germany.
The country is a Darwinian training system. Possibly.
There's a spectrum of possible explanation, from "this is a model training artifact which for now they correct via the harness" through to "this is deliberate, and creates a constantly moving target to make third-party harnesses less efficient for lock-in purposes".
I'd not discount the adversarial end of the spectrum.
If we take the noise about Mythos' capabilities as read, then releasing it freely into the world could result in chaos, as attackers find myriad new vulnerabilities and use them, and code owners frantically hunt for them and fix any that are exploited. (Noting, of course, how legendarily quick and agile large corporations aren't, compared to motivated individuals or small groups.). Eventually, given unfettered access to Mythos and sufficient time, things would settle down again once everything was patched, but who knows what would happen in the process?
So I suspect this has less to do with the underlying ethics or logic, and more to do with Anthropic not wanting to be held responsible for unleashing a potential period of chaos onto the world.
Of course, if someone has access to a tool that can find vulnerabilities in code, the process is identical whether the ultimate intent is to fix or exploit them (which may be Hegseth's underlying logic?). So to avoid this 'world chaos' scenario, Anthropic needs to somehow restrict Mythos access, avoiding bad players. And the only heuristics available at scale are either task-based assessment by AI (with downgrading of anything marginally risky to older models) or selection of trusted organisations by humans.
(By the by, to your point, it would also make sense to expand Glasswing to open source maintainers, at scale. I can't tell to what extent this has been part of that project?)
Yes. It would occasionally work once or twice, then stop. Just wasn't reliable. I have no idea what I was doing wrong, but:
1) I've installed -> played -> struggled with trains -> gave up and uninstalled the game about four times now.
2) The last time I tried every setup and combination I could think of, and also every tutorial, forum post, and Youtube video I could find, to no avail.
Having read what you wrote several times, I'm not sure whether you're agreeing or disagreeing with me. FWIW I agree with you 100% - we should judge "evil" on clear intent, impact, actions, etc.
I agree that Nestle's actions are evil, based on an assessment of their actual actions and not just the fact that they have a strategy to deal with competition.
Let's go back a step. You said Anthropic are "not trustworthy" and appeared to support this by saying "The CEO has recently started taking a stance against local AI". You then said they are doing "evil things".
I'm totally prepared to accept that any company --Anthropic or otherwise-- is doing evil things, but it takes a little more justification than someone saying so, or that "The CEO has recently started taking a stance against local AI".
I couldn’t get them to work reliably at all in an automated manner. I think I beat the level that needed them by mostly manually operating them, and then got fed up.
> Since 2.0 it's quite straightforward using interrupts
AFAIK Apple’s “services” revenue is a little over a quarter of their total; everything else is hardware, dominated by the iPhone. Mac hardware is <10% of total revenue.
iPhones are largely locked to their App Store so no risk there. Macs (currently) aren’t locked to the App Store - and I’d guess that Mac App Store usage is middling as a result.
Which is to say, I doubt that a marginal Mac App Store revenue hit from a small proportion of users switching to Linux over MacOS is the driver for not supporting Linux development. I’d guess it’s more about an inflexible company culture and maybe not wanting to extend their area of responsibility and risk.
This is based on the observation that the medium-sized open weight models (~20-35b) are very able to one-shot smaller discrete tasks but seem to lose their way project managing themselves through larger tasks that have multiple steps.