797 karma · joined December 14, 2017
I think this is a common behavior by inexperienced people in new spaces, but is quickly thinned out as people move up in skill. The most consistent sign of an expert seems to be a cool “smoothness” to their choices that achieves good results with the least effort possible.
Spam filtration at all endpoints has become so effective and options to unsubscribe sufficiently effective that the impact as a user to receiving repeated contact is almost nil, often helps as it can be easy to miss contact on one or more platforms if one’s busy, and it definitely doesn’t feel comparable to the harms of stalking.
It would therefore seem they’re done growing and are planning to go into maintenance mode.
[1] https://headphonesaddict.com/world-of-warcraft-player-count/
Having worked on very long-running projects, testing or well-funded QA doesn’t tend to save you from organizational issues.
What typically seems to happen is tests rot, as tests often seem to have a shelf life. Eventually some subset of the tests start to die - usually because of some combination of dependency issues, API expectations changes, security updates, account and credential expirations, machine endpoint and state updates, and so on - and because the results from the test no longer indicate correctness of the program, and the marginal business value of fixing one individual broken test is typically very low, they often either get shut off entirely, or are “forced to pass” even if they ought to be an error.
Repeat for a decade or two and there quickly start being “the tests we actually trust”, and “the tests we’re too busy to actually fix or clean up.”
Which tests are the good ones and bad ones quickly become tribal knowledge that gets lost with job and role changes, and at some point, the mass of “tests that are lying that they’re working” and “tests we no longer care to figure out if they’re telling the truth that they’re failing” itself becomes accidentally load-bearing.
Not to say that KeePassXC isn’t useful if you want even more fine-grained controls, but it seems like in the
> Use password in browser
Use case, KeePass would actually weaken the security guarantee by adding a second component you need to trust.
If the base game isn’t enough to keep you interested anymore because it’s gotten too simple, space exploration is a good hand grenade to mix things up.
If the app NextDoor is any indication, going local doesn’t seem to help either, its as negative and fractious as any other online community.
At some point, one gets the feeling “the medium is the message” and it’s a flaw inherent to all online social media.
1. It typically carves out new canon about its universe, and
2. Closes old loose ends
Canon when added haphazardly can:
A. retroactively create in-universe plot-holes (if X existed in the universe, then Y happening in the originals made no sense, they would have known about it),
B. remove magic and mystery from an existing universe (famously in Star Wars’ case, quite literally, because now “The Force” is a bunch of little biological creatures) or
C. otherwise trivialize what were previously key moments (someone who was portrayed as a big bad previously was later just shown to be some powerless, boring low-level lackey who was irrelevant)
Additionally, loose end patching is infectious to older entries of a story:
A. When an older work presents a mystery, when you have a solution in a later work and that solution is lame, it taints the original by the knowledge that the loose end is, forever after, a broken promise by the writers.
B. It typically crowds out the potential of better solutions later (typically in film, franchises have to reboot to backtrack on canon),
C. prevents the solution from being the potentially better “that’s just where the story ends, use your imagination for what happens next!”, and
D. makes it retroactively hard as a fan to justify recommending the series to new fans, because if they feel passionate too, you know there’s no light at the end of the tunnel for incoming fans, since you know how it ends and know that it’s not rewarding.
This therefore to me seems like a thinly-veiled variant on “if you have nothing to hide you have nothing to fear”. A world of constant recording is not a better one.
From a concept stage, it has attractive similarities to the way people learn in real life (rewarded for successful learnings, punished for failure), and although we know similarities to nature don’t guarantee better results than alternatives (for example, our modern airplane does not “flap” its wings the way a bird does), natural solutions will be continually looked to as a starting point and tool to try on new problems.
Additionally, RL gives you a good start on unclear-how-to-address problems. In spaces where it’s not clear where to begin optimizing besides taking actions and seeing how they do judged against some metric, reinforcement learning often provides a good mental and code framework for attacking these problems.
>There was a paper recently suggesting that you can use a preference learning objective directly
Doing a very quick skim, it looks like that paper is arguing rather than giving rewards or punishments based on preferences, you can just build a predictive classifier for the kinds of responses humans prefer. It seems interesting, though I wonder the extent to which you still have to occasionally do that reinforcement learning to generate relevant data for evaluating the classifier.
• Model is “whatever you need to store persistently, how you represent that data, how should the data be structured and stored when it’s at rest (eg in a database, nosql environment, or big data query system), how you are expected to query that data when you need it, mediates common CRUD data query operations on the store, and broadly handles giving back the results of any stored data query. Intent of the model is to be a black box the controller can talk to whenever it needs something from “storage”. If you did it right, you can change the entire underlying data store and endpoint, and as long as the API for the model is preserved, the controllers don’t even notice.
• Controllers mediate “turning prepared or cached model data queries into something to give back to the user” and “listening for the user wanting to do something and responding appropriately by fetching needed model data or server resources”. Handles moving between pages, incoming API calls, last mile data filtration (where last mile is based on milliseconds to do it, >25ish or so and it probably belongs on the model) and cleanup if any is needed, and is first line of contact (and defense) for anything a user’s triggered.
• Views are whatever the user can visibly interact with and see, and provide buttons, links and interactibles which hook into controller calls to change things or give back data
I’ve found the above model typically has good separation of concerns, usually avoids most fights of “does this belong on the model or controller”, and you can usually cleanly parallelize work between multiple people on a small team if you use ~~waterfall~~ scrum to agree on a data model for each endpoint and needed functions before starting.
For (;;) { print(“hahaha you didn’t say the magic word!”); }
In another loop would also prevent termination without program shutdowns.
Money makes a lot of problems go away - with this kind of money, one could buy a Hawaiian 10-20 million dollar home in a gated luxury community where access is gated and staff routinely patrol, pay the fully loaded cost of two 100$/hr personal armed guards for a 24/7 security detail ($876,800 per year each), and still have something like 100 million to spend on whatever you want before you die.
If you’re a jilted businessman, sure, maybe you’re still willing to try something insane like hiring assassins or trying to get control of your security staff out of pure spite, but frankly, for every individual at the company besides perhaps the largest shareholders (who are actually the most risk-adverse people with regard to the danger of such a plan becoming public), it’s not actually their money and they’d rather just move on because it doesn’t affect their paycheck.
My guess is they would see a parallel between other also-not-really-telegraph-wire technologies on which others were convicted of wire fraud, and a Starlink connection.
This means it probably wouldn’t legally be treated as appreciably different from any other current physical or wireless connection.
I’ve always found it incredibly uncanny navigating through 3D environments in VR without any kind of force feedback, or any sensation of weight or touch.
First failure mode is that the metric for two bills being “the same” will be too narrow, in which case the rule will be ineffective due to legislators playing with the punctuation or paragraph structure, or
Second failure mode is that the rule will be too broad and this becomes a weapon which would increase gridlock even further.
By providing parties and non-majority fringe groups within them the option to “damn bills by weak praise” into the future, legislators working adversarially could propose a bill they don’t want to pass, veto it, argue any successor bills are “the same bill”, and now, they’ve effectively enshrined the state of the issue for the next X years.
As someone without much familiarity, this alternative seems circuitous if Google results are what you actually want.
That sounds like it could be useful in legal and policing contexts. One complaint I’ve often heard about body cameras in policing is the possibility of footage that was created and did indeed exist being “lost”, with the implication being that legal entities intentionally destroyed evidence of either the footage existing, or its contents.
To the extent that this centralized authority issue applies to textual data or metadata, decentralized records of messaging times and contents would potentially be an invaluable tool of transparency and promoting systemic confidence in a visible shared audit trail.
If my company already uses spreadsheets and teams of people skilled at manipulating them, it might not be very hard to convince my manager that good AI use cases (quick formatting, intelligent auto-completes, intention detection from sheet structure) could probably speed us up by 20% - but convincing management of the value of a full system re-design by comparison would likely be a very tough sell, if not viewed as an insult or sign that I'm not a team player.
Just because the future may prove these methods antiquated doesn't mean management is liable or likely to change their tune today.
That process was invaluable for making normal bullshitters into bullshit artists -- how will we train our elite human bullshitters in the modern age of language models?
For traffic court, perhaps, but these AI tools don’t seem guaranteed to not make this problem worse at broader scales.
These AIs will also be available to large firms, who are equally incentivized to use AI for augmenting argumentation through existing lawyers, but will also be incentivized to train their own walled garden powerful models in a way that poorer clients still would likely not have access to, and which individuals and smaller firms will not have the resources to train themselves.
These kinds of AI models could very easily serve to entrench and raise the cost of a defense by making it so you not only need a lawyer, you need a lawyer-backed-by-a-firm-with-a-LLM to be competitive at trial - making all existing problems even worse.
Devils advocate, because increasing isolation of the necessarily public information a governmental service manages from the government organization servicing it is a good thing.
The role of title transfers could itself be argued to be a form of governmental overreach. For the mere act of interpersonal buying and selling, why ought there be a database managing that?
If we agreed such a database must exist (and horses and carriages seem to have done fine without one), wouldn’t it be preferable for individuals to be able to directly interface with that system interpersonally, rather than through a governmental intermediary?
If we agree there must be a governmental intermediary, wouldn’t we prefer such an intermediary to solely pay for and maintain dumb infrastructure they are not expected to waste taxpayer time and money manipulating, rather than giving them control over the underlying transactions and wasting a great deal of both?
I don’t agree with this theory, as I think it’s rather likely SBF will be viewed as a flight risk.
• His heavy involvement in the cryptocurrency space makes it likely he has significant moveable digital cash reserves that cannot be easily seized and which would greatly aid an escape.
• He has shown himself to be extremely mobile and was seemingly already attempting to get some distance from the USA, given that he was arrested in the Bahamas
• The gravity of sentences charges he’s potentially facing are vast, and he has many incentives to flee prosecution if it means potentially dodging spending 10-20++ years in jail
I therefore think it’s very unlikely he gets bail.
Fortunately, someone’s made sure to get deepfake “Biden” and “Trump”’s opinion on the rapper Chief Keef, which does the job quite handily [1][2]
[1] https://twitter.com/DailyLoud/status/1597678029508968448?s=2...
Spitballing a few ways it could be? Right now all reddit users have the same rights and privileges to use any subreddit. I expect this to change and all these rights are going to get parcelled out, including:
• Pay-to-access subreddits which would have “curated content” or education materials in them
• Subreddit subscriptions which would allow or require ongoing monthly payments to have subscriber permissions granting special feature access.
It can come with post visibility increases (you can upsell this), both algorithmic and graphically on the interface, options for more powerful voting rights on polls and posts, gated posting rights etc., and would be upsold to creators as a value-add they’d push on their audience
• User “blue checkmarks” - you pay money for some kind of proof of identity
• Paid “super” upvoting - perhaps “pay 10 bucks, and we add 100 upvotes to your posts, no questions asked”
• Charge for subreddit creation or growth - limited mod slots, limited subreddits per person, and no creating subreddits over a certain size without paying up
The above are the “nice changes”, you can go meaner but I’d be worried about giving them ideas.
You may need to present evidence and give her ways to tell that this is in fact not Elon, because she currently does not have the tools to recognize a situation where she is being scammed. Methods might include explaining that the situation of "a media famous billionaire is personally interacting with her" is too good to be true, especially without extremely good evidence that it's true which she has not received, that intuitively, someone with a net worth of 181 billion dollars would not be asking her for 5k, and that the site she was using just got closed down for scamming its customers out of their money.
Use any rhetorical tools in your argument toolbox to give her the intuition that she's been had, how she was had, and explain that these kind of scams happen thousands of times a day and explicitly target the elderly, so she's not to blame for what happened.
Once you're on the same page, help cut the scammer out. Block their email, block their social media, block their phone number, virus scan her computer and add an ongoing scanner on her machine, update her phone, install an ad-blocker on her browser, and if you have the savvy to do it, go into her router and block that scam site at the DNS level so she can't get back.
• Website today seem much more aware of how search would ultimately seek to summarize and off-site catalog their content to siphon away incoming users than when Google started, so you don’t have the benefit of naïve counter-parties anymore who will quietly welcome you scraping them.
• The web’s way bigger and faster now than when Google started so keeping results up-to-date is Sisyphean
• If you only have a search engine, you will have to pay for almost literally any other non-search data stream you’d want to use to augment search - locations and business info, traffic, routing, media names and metadata, stock prices, and so on.
• While power users here often complain, it seems that whatever metrics google is using have found that human-curated results and normal non-ML approaches like basic keyword search are not preferred and ML-based approaches have been more successful, which likely cost a lot in compute to build and definitely comes with a hiring talent premium for being a niche skill
The moat is money, and lots of it.