131 karma · joined November 27, 2016
Anyways, the new CF AI tool is relatively decent for this purpose, explains the fact that most warnings in CF are harmless, but the lack of standardized guidelines is annoying to say the least.
There's a growing (retarded) perspective, not only here in Romania but in most of the EU, that we must digitalize at all costs; this itself is obviously not a problem, but it usually comes with the removal of the "traditional" options too. The same goes for payments (removal of cash) and basically every aspect of the administration/society where this could be applied. The most recent concrete example for Romania is the mandatory digital signature by ANAF (IRS equivalent) for all companies.
Unlike other regions of the West: NA/WE/rest of the West (SK, JP, etc.), in EE, people (*younger generations, explained below) don't often think about the aspects of confidentiality that much, or if they do, they usually apply it incorrectly (either IRL [filming in public] or online ["I've accepted a draconian TOS that gives the company rights to all of my data, but I haven't actually read any of it, and now I'm invoking my "GDPR rights" against said company."]).
It does not take a genius or a historian to observe that while the older generations do not have this "problem" (they're very skeptical of the gov. [at least in certain aspects], in an almost american fashion — explained by the fact that they've experienced communism), younger generations position themselves in the extreme opposite; they despise the older generations ("boomers") for being luddites and somehow blame them for the gap in civilizational progress between WE and EE. There is a legitimate point here, of course, but it should be noted that the argument is almost always presented in an emotional, overwhelmingly irrational fashion, and almost never includes comparisons between the centralization done by the state in the past [communists] and the new one attempted by current govs [but digitalized]. (This delves into a broader socio-cultural divide between the generations that includes other aspects like religion, traditions, etc.)
The biggest portion of the so-called "IT people" who call for this "digitalization wave" are midwits, "inverse luddites" (they don't care about the implications), and, in smaller numbers, grifters and second-order beneficiaries (on the "digitalization" gov. projects). Obviously, it should be mentioned that, despite progress, nepotism and corruption are still present, and there are cases where gov. contracts are not given to the best candidate.
As a person who's both working in the IT sector and not a boomer (despite the impression of my comment), "we" should be careful about enabling totalitarians in our endeavors."Hackers" is just one reason out of a multitude of other reasons for not abandoning the sacred paper (see the ES+PT incident one year ago).
Yes, under the TOS, what grok is doing is not the "fault" of grok(the reason is the causal factor of the post[enabled by 2 humans: the poster and the prompter]; the human intent is what initiates the generated post, not the bot; just like a gun is shot by a human, not by the strong winds). You could argue it's the fault of the "prompter", but we're going to circle back to the cat & mouse censorship issue. And no, I don't want a less censored grok version that's unable to "bikini a NAS"(which is what I've been fortunate to witness) just because "new internet users" don't understand what the Internet is.(Yes, I know you can obviously fine-tune the model to allow funny generations and deny explicit/spicy generations)
If X would implement what the so-called "moralists" want, it will just turn into Facebook.
And for the "protect the children" folks, it's really disappointing how we're always coming back to this bullsh*t excuse every time a moral issue arises. Blocking grok is a fix both for the person who doesn't want to get edited AND the user who doesn't want to see grok replies(in case the posts don't get the NSFW tag in time).
Ironically, a decent amount of people who want to censor grok are bluesky users, where "lolicorn" and similar dubious degenerate content is being posted non-stop AS HUMAN-MADE content. Or what, just because it's an AI it's suddenly a problem? The fact that you can "strip" someone by tweeting a bot?
And lastly, sex sells. If people haven't figured out that "bikinis", "boobs", and everything related to sex will be what wins the AI/AGI/etc. race (it actually happens for ANY industry), then it's their problem. Dystopian? Sure, but it's not an issue you can win with moral arguments like "don't strip me". You will get stripped down if it created 1M impressions and drives engagement. You will not convince Musk(or any person who makes such a decision) to stop grok from "stripping you", because the alternative is that other non-grok/xAI/etc. entities/people will make the content, drive the engagement, make the money.
YouTube's system of DMCA takedown(the copyright issue being way more serious legally than what DSA is supposed to protect against) is not perfect and cannot be perfect (proven by the fact that content is unjustly taken down all the time). DSA is just the same, except more vague, more complicated and (imo) ultimately worse.
DSA has an appeal mechanism, with an option for out-of-court settlements, which means you can employ independent fact-checkers (certified by Digital Services Coordinators (DSCs)); the list of certified bodies is, of course, maintained by the European Commission. The problem is that these DSCs are appointed by each country's gov., which means there's potential room for conflict of interests not only at a national level(I find hard to believe appointed DSCs are completely impartial to the gov. that appointed them) but also at an EU-wide level(certified fact-checking bodies who are supposedly not influenced by EC when judging cases pertaining to EU in international cases).
And the Cambridge Analytica "phenomenon" is not really something you can realistically prevent. I'm sure it happens now with some other better firm (Palantir probably), but this is really beside the point. The point is that normal citizens, like you and me, are effectively censored upon suspicion before any burden of proof is provided. Nothing says "protecting democracy" like deleting posts from social media and then finding out the context.
> Individual free speech is not - of course - ethically or politically identical to "free speech" produced by weaponised industrial content farms funded by corporations and foreign actors.
Sure, nobody likes bots/paid shills. But of course, in a normal society, you have to prove those posts are made by actual bots/content farms before taking any action. Otherwise it's just censorship. Election interference always happens, without exceptions, but degrees vary. This is not to say we shouldn't point out when it happens, but to not do censorship against our own citizens because "the models indicate a pattern akin to foreign entities." Patterns are not burdens of proof, and thus employing a "crowdfunded" fact-checking system like Community Notes or the one from YouTube is at least partly the actual solution instead of directly removing content. Under DSA, you can effectively remove content without providing burden of proof regarding the identity of the poster. Platforms must provide a "statement of reasons" (Article 17) to affected users for any removal, including appeal rights, but this does not impose pre-removal identity checks on posters.
Or let me guess, "Trump bad and therefore we should accept DSA/Chat control 2.0/3.0/etc."? Sorry, I don't care. And people who think this is only about the recent X fine are also wrong (this started last year when Thierry Breton started influencing european elections while also boasting about how he can annul such elections without repercussions; you can deduce what I'm talking about by asking an LLM). This is in part US gov. protecting private companies (and thus itself) from fines, sure, but the broader point about censorship within the west applies. Everything that hurts the people making legislation regarding the Internet (or software in general) within the EU should be welcomed with open arms.
EU apologists would rather change the subject and talk about Trump and the polarizing social environment in the US rather than acknowledge that within the EU, there's not even a chance for discourse to be had about any policy(especially the nonexistent free speech) due to the aforementioned laws. The same people will act surprised when extreme positions regarding the EU are adopted by an ever-increasing number of people "until morale improves".
Or maybe it is a problem with the structure that fosters an environment. What comes to my mind is the exceptional case of OpenAI, which started as a nonprofit. Sure, it "ended badly" because of the known drama, but my guess is that besides the money that was poured into it, it thrived because researchers had kind of an "emotional safety net," meaning that they wouldn't be pressured for results as much. Probably the reason some startups perform much better too.
I think career continuity matters, and you don't necessarily get that in the private sector for sure. This discontinuity then leads to practical work discontinuity, which means less work done (which is amplified by the non-decentralized nature of working in private compared to shared science in public, as you've explained).
My bottom line is that the private field could do better, and frankly it's kind of their loss. What I'm curious about is whether a "semi-private" approach is better: a non-profit or some kind of foundation. I guess in practice they're still private, but whether the money part can be "solved" through crowdfunding/some modern methods and whether they're viable long-term remains to be seen. One thing is for sure: a culture appreciative of science will definitely open more doors into novel methods of funding and organizing (maybe in the future these methods could rival the "traditional ways" of public science).
Everything can be "sold", especially in today's age with the new methods of discoverability. But I would argue scientists don't need to "sell" something in the capitalist sense. They need to link the hope of a new discovery to inventors, innovators and entrepreneurs. Sure, some things might "fail" to continue by failing to adjust to the markets, or some scientific discoveries might be used for bad things (ethically), but this is (1) both inevitable and (2) the responsibility of the scientists & the people buying the end product/service. If I'm not mistaken, most bad/evil/etc. discoveries were made by scientists working FOR the government/king/etc. throughout history. If anything, democratizing science through the capitalist markets seems like a more beneficial way to develop self-sustaining science. The key thing is transparency, which can be less present in the private sector, especially when corruption is involved(assuming transparency is demanded by the gov.).
People talk about changing laws or technical solutions, but the inconvenient truth is that technically literate people should peer-pressure nearby friends/family/etc. into being more aware of such possibilities. I've done so, to the extent that some people find it ether borderline schizophrenic/paranoid (to my "luck", I live in an ex-communist country, where most people are usually skeptical in many contexts with strangers; so this group of people is relatively small).
People who know better bear a responsibility towards helping others who don't; towards those who are too kind (or naive) for their own good; Even though I'm the "tech guy" in my close circles (family, friends,etc.) like many here, I often do the >opposite< of what other "pro-technologists" do these days: I don't encourage people, especially the older generation OR the more tech-illiterate ones to use more technology, because it is obvious that doing so "injects" another vector of attacks into their lives. More often these days this is not possible, everything gets digitalized to the detriment of such groups, but this also delves into the politics of keeping "older options" (cash, paper trails, etc.) available even if digitalization happens. Often times the older options are more secure, though obviously less convenient.
This is a non-solution, yes, but it is the correct way to approach this (imo), as more and more places LEGALLY force digitalization of different institutions(banking, gov. agencies, etc.) which inherently either add, or worse, completely shift the risk into virtual spaces. This is why a "legal" solution is more often than not either a slow one or a completely pointless one. It will always be an arms-race between scammers(which operate more effectively[in theory] due to their decentralized nature) and the gov./banks/etc., which operate in a more centralized fashion, thus demanding and imposing more control above all included parties. A legal way will always demand more than it's worth.
I digress from my shift into politics, but bottom line is this: don't let your peers/family/closed ones get into these situations. If you have "an authoritative" voice regarding tech, use it to first cultivate awareness regarding dangers, before cultivating hype/or anything else. (Obviously not talking about anyone specifically, but the whole "geeksphere" as a whole)
Good luck to you and your family.
Secondly: the death of EU cannot come soon enough. 10 years ago euroskeptics like me were wrongfully called "russophiles" (lmao), even though I'm from a country that is constantly threatened by Russia with drones, propaganda, etc. (RO for one's curiosity). Ironically enough, for those coming from ex-communists countries, EU sure looks increasingly like USSR, but with blue instead of red. It's infinitely better than communism, sure, but the optics and path of EU resemble those of USSR in it's "wellbeing of workers"(and other socio-cultural issues) propaganda phase(the irony is Russia here, obviously).
History never repeats, but it rhymes. And a calcified supranational institution delves into authoritarianism in the later stages of its existence. Reform never happens, and if it does, it’s at face value. It’s much “easier” (for the people in those institutions) to double-down on the status quo position rather than reform; and obviously it becomes increasingly harder for the vast majority of people to voice their opinions or concerns, especially those not aligned with the status quo. (* Key difference here is NATO which is US-led and EU which is still Western Europe-led; US is still a functional democracy unlike EU institutions)
Although it’s not really a very complex topic and the causal factors are relatively simple (at least to identify, solutions are much harder to propose [mainly due to the mentioned constant double-downing]), it would take a long time to explain/convince why the existence of the “current” EU is detrimental to Europeans (at least to the people not aligned with the status-quo). So-called benefits stopped at the common market treaty (EEC) iteration of the EU. Security is not and should not be in the EU’s purview, we have NATO for that. And the most obvious issues that the EU keeps worsening are socio-cultural positions that either (1) dilute the differences between different nations (2) [in case (1) was not a “problem” due to shared values] directly propose completely different values and/or positions. There’s no “objective morality” debate to be had here, democracy does not inherently mean choosing the most “scientific”/“moral”/(any other metric) position in policy: it simply means choosing what the majority of people want (If you want to change policy: change people’s minds). The double-downing of the “EU regime” is usually defended with the rhetoric that it does so in the name of “democracy”, “morality”, “tolerance”, “objective wellbeing of society”, etc. but there’s no mechanism for true democracy if the EU undermines/punishes the will of individual nations on the pretext that “it does not conform to EU-wide proposed policy”. This leads to “multi-step” issues and other regional conflicts between interests of nations (the winning bloc [usually the wealthier one] gets to basically impose policy on the smaller one).
I’m quite off-topic on the subject of privacy, but my key point is this: don’t expect things to get better; or at least not without a huge cost. Recent pullbacks from the EU regarding the DSA/AI Act/GDPR are done so out of necessity: the EU is losing ground massively (= money) in the tech space due to stupid policies made by dumb bureaucrats. Half-assed “reforms” like these will not make huge improvements (mainly due to the unchanged fiscal policies [which are going to get worse: upcoming euro stablecoin]) but will keep the EU afloat amongst those who can’t see the sinking ship. Oh, you like privacy? Well, expect it to get worse, as eID is surely coming for all citizens in the name of safety (which has been eroded due to “our” [i.e. regime’s] stupid policies).
Finally, as I foresee some will keep replying with “Russian <something>”, let me just say this: pray the downfall of the EU won’t be an opportunity for Russia to do anything. At this rate in the regional conflict, Russia doesn’t look good long-term, but if history has tried to teach us anything, it is that Russia’s unpredictable.
Besides the obvious compromise in quality that companies would have to make to appease the 'karens' of society (not to mention the additional compliance and regulatory burden imposed on new companies), wouldn't it be simpler to just have users take a basic 'TOS test' when creating an account? Sure, it's inconvenient, but at least the company would be legally protected. The purpose is obviously not to protect companies, but to move the spotlight towards the real causal factors.
No matter how simple the TOS acceptance process becomes, people will still find a way to blame the product or company, ignoring the core issue of how someone got into a mental state where they use LLMs to cause self-harm. I don't see people suing rope manufacturing companies for facilitating suicide.
Ads & referrals are already in the works, and people are generally tolerant of those. But, as with any company, appearances matter. ChatGPT will definitely lose users at the slightest possibility of having non-sanitized content served to more morally sensible groups.
Frustratingly enough, i had already done the "re-enrollment" a long time ago(basically when they announced it was mandatory), but it seems like that was pointless(hopefully not).
I saw some prompts about birdhouse, re-did the enrollment, and badly enough (I think i dug my own hole with this one) it asked to remove the other 2FA option (SMS), to which I clicked yes.
This might sound bad but I sincerely hope X fixes it somehow, and all the keys enrolled/re-(re-[etc])-enrolled are not lost, especially those that were not added today. It might be a good idea (in practice, bad for security) to disable this new "https://x.com/account/access?flow=two-factor-security-key-po..." garbage fully, as I don't see myself contacting X support anytime soon(for obvious reasons).
If it was incompetence, one could argue that nothing should happen, perhaps an apology or some useless corporate article. But it was malice, and to deny that is (imo) the real issue here. (I'm not saying you're doing it, I'm just saying some people do/did it)
I'm curious if you think people outside US/Western Europe (like me; greetings from EE[we've seen such edits in our communist period, fyi]) who disagree with the assessment that the BBC is "centrist with a slight leftward skew" are far-right with obvious biases? And if so, on what grounds? Most people who say BBC is propaganda(like me) don't consume MSM at all(or, in the case of US, stick to Fox or something). To say all alternative media I consume(which you'd be correct in assuming) is "skewing far-right" is to, ironically, behave like the ones you're pointing to. It's also incorrect: alternative media is infinitely more diverse today, even after all the reshuffling/restructuring in the past 10-12 months(which culled a decent amount of the left-leaning alternative media) than MSM.
Like holocaust deniers, like nazis, etc.? More and more buzz-words and labels broadly and freely used in order to dilute their respective meanings.
Our parents, grand-parents, etc. did not teach us good virtues "by shaming" (i.e by telling us we're X/Y/Z buzzwords), but either through example(of what not to do) or through rationality (explanations until comprehension). It's no wonder we have rising "X/Y/Z" sentiments when, ironically, the people who seemingly advocate against "these bad boys" resume to just label and categorize individuals instead putting the effort to either educate, explain, reference. Low-effort means low results, and virtuous traits are definitely not gained through complacency.
You solve things with dialogue (not monologue*). And if those people who use these buzzwords don't like dialogue "because some guy proved it" [it's actually irrelevant if the referenced fact is true or not] and dismiss discourse shamelessly, then they're doing more damage to their own narrative. Nobody likes being told what to do(this includes knowing/believing/etc.), and facts matter as much as they're being understood.
> Early Friday, Musk said that xAI would release its first AI system — presumably Grok — to a “select group” on Saturday, November 4. But in a follow-up tweet tonight, Musk said all subscribers to X’s recently launched Premium Plus plan, which costs $16 per month for ad-free access to X, will get access to Grok “once it’s out of early beta.”
The lower tiers getting in the deal as well sounded too good to be. But still, at 16$/month if grok is anything like cgpt4 it's worth it, imo. What I'm gonna be interested in is Musk's "promise" that it will not be censored/lobotomized, like chatgpt was(and is). I'm not sure what technology they're using such that grok uses realtime X data (if anyone has ideas, feel free to share), though i'm assuming this is something like Bard(which from my experience does a similar thing) and the knowledge isn't "trained" per se in the model.
Anyways, this sounds too obsessive, I actually still somewhat like NPR and they're still one of the more sensible sources out of the "legacy media" outlets. My points still stands though: MSM has been trending downwards for a while now; And even though data might not fully reflect the overall sentiment, you also have to take into account that: population grows, median of age usually rises, etc. (*) It also needs to be said: it's not only about NPR and podcasts filled this void, which is very understandable since it's the same medium(audio) but more targeted, personal.
[1] https://www.pewresearch.org/journalism/fact-sheet/public-bro...
[2] https://www.statista.com/statistics/614143/npr-weekly-audien...
I'm aware of those frankly tragic episodes. But ask yourself this: is the problem the fact that simply misinfo was put on FB or the fact that there wasn't enough counter-balancing information available? I'm not trying to be sarcastic or downplay the situation, but the fact is that a lot of places suffer "from misinfo" because the only info available is misinfo. The human mind is inherently a curious mind. And as i've stated in my first comment, people tend to follow info from trains of thought that make sense(are logical)[yes, even if we're talking about religion/etc.): the contents of the information matters, how cohesive it is, how it attacks the false claims. I would argue that a big lie does not even need a "big truth" to be debunked, it only needs a well-thought question to instill the skepticism required for the lie to not be taken as plain truth.
Closing thoughts: in third-world regions: the ones you've mentioned, russia, "china"(on the freedom of information/liberty front), etc. the issue of misinformation is not really that misinformation exists; but actually that it is mainly the only information that exists, and there's not enough balancing "truth"/skepticism. You might think i'm silly or that i'm arguing for misinfo(that wouldn't suit me by the way), but I live in an ex-communist country: I've had some relative contact on how the centralized government used to push actual misinformation and how people took it as blind truth. After the said government(s) failed, people usually became more skeptical about all kinds of information(a good thing to do). Those people are the generation of my parents for example, a dying breed, and it should not come as a surprise that every subsequent generation that came after and took all news as "facts" became usually dumber, less rational, and more naive.
>The EU is asking Twitter to define their own moderation policies and enforce them. The EU wants a lot of things that aren't feasible. This is one of them. Elon stated plenty of times that the platform should not censor or moderate more speech than necessary. This position turns out to be "harmful" in the eyes of EU, because like the bureaucrats they are, they need everything under control or determined to be under scrutiny a priori(by labeling speech by certain criterias). Imagine a new form of content that does not fall under any current regulation/moderation policies. By default it should not be problematic, because it's not "illegal". Thus my point: what EU deems misinfo/illegal is a slippery slope and will never, ever be satisfied. Unless of course we include a whole lot of nothing-speech.
>So which is it? You want government to do this or private companies?
Neither, or both. Depends if you think my answer is actual "moderation"(or censorship/removal of speech) or providing context in the form of fact-checking(community notes or whatever). If governments want to "fact-check" they should go ahead and do it. Otherwise if they simply just demand that the private company or they themselves(through some intermediaries) want speech removed, that's plain text censorship.[A little note: it's worse than that, you de facto have an actual fascist collusion between the state and private enterprises].
It's 2023 and people still don't understand the premise of government censorship. Guess what: deleting misinformation is still censorship, it does not attack the issue itself; it only probably antagonizes the people who put said (false or not) information, and 'downgrades' the filter of the people who are actually interested in the truth. The information being false or not is actually irrelevant, it's the process of censorship that's damaging. We've been over this with covid already. You don't even have to be an optimist to notice that, most often than not, truth eventually comes out. And what's more interesting is that the more open and free a medium of information/discussion is, the >faster< it gets out. This is because even though humans are impulsive and emotional, we're still rational beings. Fabricating things not rooted in reality is harder and impossible to maintain.
In the last 15-20 years the internet became less and less anonymous, and yet those problems still exist and they're a central issue. While it's mostly a correlation and definitely not a causal factor (because internet adoption was non-existent back then compared to now, amongst others), it still begs the (rhetorical) question of why the pressure against anonimity.(See past and current abuses in this regard by governments/empires/etc). I'm semi-jokingly talking about a conspiracy here, because i've used both anonymous and 'very verified' platforms, and most of the time the misinformation happens on the latter. This is especially true since the facebook days, because the platform itself gives the vibe of credibility (alongside the user/entity posting it).
Trying to combat misinformation in this way is and will remain a cat&mouse game because there will always be actual bad actors which will try to impersonate/immitate the good ones. Put it like this: you have the same people walking on 2 streets: on the first one they hear Biden/Trump/Macron/etc. saying a fake thing, spreading misinformation; on the next: a random hobo saying the same thing. Which one will have the worse impact? While I'm not sure there have been done such studies/experiments, past "anecdata" tells me the influential person successfully fools a higher percentage of those people. While you could say "but once exposed, he's recognized as a fraud" and that's entirely true: we then return to my point of people trying to impersonate/fake credibility or grift the issue by saying unquantifiable or things that just cannot be entirely fact-checked (without projecting or speculation): those actors do more damage because they appear credible.
I fully agree though that there are certain aspects that need to have a 0 tolerance policy (CP and similar things) even when anonymous. And with regards to flagging anonymous users as such: would be interesting if any social network tries to make the experiment of having semi/fully anonymous modes: because honestly that would be just one of the few actual solutions to combat polariation on social media: by encouraging more free & honest discussion (even if there's 90% chance it becomes less civil).
And I'm sorry for the upcoming little rant, but whoever thinks they're anonymous while using a Meta(or any Big Tech platform, really) product is an idiot, tech literate or not. Not even places like 4chan have true anonimity, depending on the place & jurisdiction we're talking about[remember the case of the guy making a call to violence(illegal) that got arrested]. The 'traditional' web is not anonymous at all:not only the underlying protocol(s) is/are inherently not anonymous by design, but you add insane surveillance and you can eventually crack anything. Even things like TOR/others are not truly anonymous, and the US regime proved that if they want to find you, they will, assuming they have jurisdiction.
Coming back: I don't quite get why people talk about free speech in this context. Not only S230 is a broken f&ckfest but we're also talking about a non-US place. What's more hilarious is that even if we would have talked about the US, defamation (w/ calls to violence & other speech not protected by 1A) is still illegal.