fional's recent activity
-
Comment on Sound engineers, help me name this vocal sound in ~music
-
Comment on Six months from now this channel stops in ~tech
fional Link ParentI suspect from a YouTube optimization perspective, if you don’t private these videos there’s a good chance it will end up in people’s autoplay or recommendations well after it has outlived its...I suspect from a YouTube optimization perspective, if you don’t private these videos there’s a good chance it will end up in people’s autoplay or recommendations well after it has outlived its usefulness and people dropping out of the video early will effect a negative penalty on the channel in the algorithm.
-
Comment on Report: Potential New York Times lawsuit could force OpenAI to wipe ChatGPT and start over in ~tech
fional LinkI suspect the concern here is going to be search engines: supplementing if not replacing organic search results with ChatGPT/Bing or Bard/Google is a natural use for LLMs. However, there's already...I suspect the concern here is going to be search engines: supplementing if not replacing organic search results with ChatGPT/Bing or Bard/Google is a natural use for LLMs. However, there's already a brewing conflict over Google's increasing use of snippets and other sorts of widgets that profit off of a website's content without giving them traffic. Injecting a chatbot into the top of the page is similar, but even worse, as there won't even be an attribution to the source material.
I think you could make a reasonable analogy to money laundering, wherein you mix small amounts of legally tainted money into an otherwise diverse cash stream and then claim the entire flow to be clean. Likewise, training an LLM allows you to mix small amounts of legally tainted content into a diverse stream and then claim the entire output to be clean.
There's even analogies to structuring--it's a crime to break up large transactions involving tainted money to smaller amounts to avoid arousing suspicion. Prompting your LLM to avoid verbatim repeating its source material feels similar in a way. "Plagiarize that thing, but not too much, ya know?"
At any rate, it's going to be a hot mess. I don't envy whomever is saddled with the task to work through the legal and moral implications.
-
Comment on What are your favorite animes and what do you like about them? in ~anime
fional LinkHmm, in the last few years I’ve been watching a lot more anime, there’s been quite a few standouts, but I’ll throw ODDTAXI into the ring. It’s about a walrus who drives a taxi in a modern day...Hmm, in the last few years I’ve been watching a lot more anime, there’s been quite a few standouts, but I’ll throw ODDTAXI into the ring. It’s about a walrus who drives a taxi in a modern day Japan inhabited by anthropomorphic animal people. The absurdity of the concept is expertly played off what is at its core an exceptionally charming character driven storyline.
-
Comment on Is anyone else just fed up with companies being greedy? in ~talk
fional Link ParentI’ve always taken a stance that the benefits we attribute historically to capitalism are really the benefits of competition, and we should be striving to maximize the quantity and quality of...I’ve always taken a stance that the benefits we attribute historically to capitalism are really the benefits of competition, and we should be striving to maximize the quantity and quality of competition. The “free market” is a wonderful invention but the concept has been corrupted in our era to be equivalent to an unregulated market. Instead, I see capitalism like nuclear energy: properly regulated and there’s wonderful benefits to all, left to run unregulated it will eventually blow its lid off and turn everything in the vicinity into a wasteland.
Pro-competition:
- Too big to fail is too big to exist, creative destruction is a prescribed burn that clears the forest for new growth. A pro-competition stance says that companies should be regulated to be small enough that the failure of any one is at most painful but not catastrophic.
- Income inequality transforms markets from democracies into autocracies. “Rising tides” arguments are bullshit; relative wealth disparity disenfranchises the common person. A pro-competition stance works towards an equal enough wealth distribution that the allocative power of markets is democratic.
- Non-living wages are a trap—if your economic activity demands a person make a Hobson’s choice between a crushing job or risk starvation and death, then you’re barely better than slavery. A pro-competition stance generates a social safety net that fundamentally gives people an alternative to taking jobs that otherwise do not compensate adequately for their downsides.
- Regulatory capture lets companies dig a moat around themselves and pull up the drawbridge. Convenient that OpenAI launches a massive AI regulatory lobbying effort months after launching the most successful AI project in recent memory, right? That’s not to say that all regulations are inherently bad—many were written in blood—but we tend to see the immediate negative outcomes being restricted without seeing the long term incentives that have changed. A pro-competition stance regularly evaluates the ongoing costs of regulation against their benefits and how that balance plays out over time.
I don’t know if there’s an official term for this political philosophy; it doesn’t seem to strictly fall into one party or another.
-
Comment on Is anyone else just fed up with companies being greedy? in ~talk
fional Link ParentIt’s also that Discord’s taken nearly a billion dollars in VC money at a valuation of $15b, so almost regardless of the specific details of how, Discord is locked into the road of increasingly...It’s also that Discord’s taken nearly a billion dollars in VC money at a valuation of $15b, so almost regardless of the specific details of how, Discord is locked into the road of increasingly desperate and invasive ways to try to show a return to their investors.
-
Comment on The origins of precision in ~engineering
fional Link ParentIf you ever want to get into the nitty-gritty of how parts are brought into reference tolerances, this is a neat video on how people did it (and still do it!) manually with a carbide scraping...If you ever want to get into the nitty-gritty of how parts are brought into reference tolerances, this is a neat video on how people did it (and still do it!) manually with a carbide scraping tool: https://www.youtube.com/watch?v=QJXqHpSh3SE
-
Comment on The origins of precision in ~engineering
fional Link ParentWithout having seen the video, I’d assume that, given a known flat surface plate, you’d first grind/lap three maybe-squares checking against the plate until they have one flat reference surface....Without having seen the video, I’d assume that, given a known flat surface plate, you’d first grind/lap three maybe-squares checking against the plate until they have one flat reference surface. Then, using the surface plate to constrain them along their known-good axis you would then a/b/c them against each other on their protruding side to bring them collectively into square. You would need at least three for the same reasons you would making the original surface plate.
-
Comment on How trans singers adapt to their changing voices in ~lgbt
fional LinkAhh, as a trans woman, I'm in this bucket! I'm fortunate enough to have a very extended upper range, and can hit a pretty convincing mezzo-soprano, but I'm always aware that I can't really...Ahh, as a trans woman, I'm in this bucket! I'm fortunate enough to have a very extended upper range, and can hit a pretty convincing mezzo-soprano, but I'm always aware that I can't really generate the tonal depth or color I want in that range. I suppose, to some degree, it's a trope that most people can't stand their own voices, but I think being trans adds complications--I'm always aware of what my pre-transition voice sounded like, and it's really hard to not hear elements of it in my voice that might not even register to an external listener. It's probably healthiest to embrace an attitude that one's voice is unique and worthwhile even if it's somewhat atypical, but it's hard not to be frustrated by comparing myself to others.
-
Comment on ‘Self-healing’ Roman concrete could aid modern construction, study suggests in ~science
fional Link ParentI think this vibes with the common engineering idiom I've heard paraphrased as "anyone can build a bridge that will never fall down, but an engineer can build a bridge that will only barely not...I think this vibes with the common engineering idiom I've heard paraphrased as "anyone can build a bridge that will never fall down, but an engineer can build a bridge that will only barely not fall down." Historically the way to build things given a relative lack of material science and structural engineering knowledge was to just massively overbuild with absurd safety factors. Reinforced concrete gets you a gigantic boost in bang-for-your-buck in terms of material strength, but means you're turning "artificial rock" into "artificial rock that critically depends on easily corroded structural members."
You can still do it "the hard way" today, but it means on a fixed budget you'll have one supremely overbuilt bridge instead of five that are "good enough" based on your predicted usage and lifespan. Build the Brooklyn Bridge for horse carriage traffic and you're going to have to awkwardly retrofit it for vehicular traffic in a few decades. Likewise, it's probably unwise to engineer a forever bridge for ICE cars today when who knows what the transit of thirty or forty years from now might look like.
-
Comment on Beyond the obvious -- keep your tech skills up to date -- what advice would you give someone to future-proof their developer career? in ~comp
fional (edited )LinkI feel like... the more senior I get in my software career, the more I think that software development is in effect extremely short-term sci-fi writing. While you've got to "earn your chops" in...I feel like... the more senior I get in my software career, the more I think that software development is in effect extremely short-term sci-fi writing. While you've got to "earn your chops" in getting a computer to understand your intent, the bulk of effort in a new coding project is to look into the future and be able to articulate that vision with a concise and easy-to-understand "story." At the lowest levels, that story turns into easy-to-reason-about code. At higher levels, it's about being able to convey that story to the developers on your team, in your org, and in your company as a whole.
One result of this is that I find it difficult to truly separate management from coding; it's harder than you think to split a project into "someone tells us what to do" and "we figure out the best way to do that thing." In many cases the real effort of the project is turning some vague high-level goal that, in reality, is deeply confused and self contradicting, into simple analogies and stories that clarify the myriad ways the top-level goal could be interpreted. This process essentially defines what the business is and how it functions, and so as a developer you're taking on "business" responsibilities whether you want to or not. As an aside, I suspect this is behind much of the failure of the "offshoring" trend of the 90s and 00s--if you're unable to push-back or work to change the business that you work for, you have little hope of a successful software implementation.
As this translates to keeping up with a development career, I see several places to grow. Developing writing skills and clarity of thought pays dividends in being able to do "inception" and instill your vision for the future in those around you. The better your ability to communicate, the more you signal to everyone around you that you're the keeper of the vision. On the other hand, you need the ability to generate that vision in the first place. To that end, I think it's beneficial to keep an open mind and explore many areas of interest. You never know where the next brilliant metaphor or analogy for a system will come from. I've pulled from mathematics, engineering control theory, history, sociology, economics... the bigger and broader the toolkit you develop as a person the more likely you are to have some interesting overlap and connection to pull from.
-
Comment on The lonely surfaces of AI-generated images in ~arts
fional LinkI really like the “pattern matching” vs. “pattern breaking” idiom raised in the article. It feels like a nice reframing of “learning the rules so you know when to break them.” I wonder if that...I really like the “pattern matching” vs. “pattern breaking” idiom raised in the article. It feels like a nice reframing of “learning the rules so you know when to break them.”
I wonder if that ties back to a more general behavioral economics framework: we make artists master difficult skills and the ability to effect the current artistic paradigm as a means of signaling to an audience that they’ve put in the effort and are worthy of consideration. But that’s sort of the table stakes for consideration—the genuine “art” vs. “Art” dichotomy is determined by what you choose to do with that audience attention once you’ve got it. What contribution do you make, stylistically, thematically, technique, or otherwise, above and beyond the status quo zeitgeist?
If AI generation nullifies “putting the hours in” or “earning your chops,” by what measure do we as a collective psyche choose to direct our limited attention to? Marketing budget? Nepotistic insider connections to existing success? Random dumb luck? It’s always seemed like a problem, but it feels like it could get worse.
-
Comment on As pay TV subscribers decline faster, pressure builds for streaming profits in ~tv
fional LinkI think the problem here is that streaming took off as the more convenient alternative to piracy--the fact that piracy is less popular today isn't that those technologies have been lost, but that...I think the problem here is that streaming took off as the more convenient alternative to piracy--the fact that piracy is less popular today isn't that those technologies have been lost, but that for $10/mo or whatever, you didn't have to deal with all the hoop-jumping. The more you squeeze the streaming customer base, the more likely it is that there's a sea change back to the old ways.
-
Comment on Twitter is planning to start charging $20 per month for verification. And if the employees building it don’t meet their deadline, they’ll be fired by Elon Musk. in ~tech
fional (edited )LinkI’m curious how this affects employee retention; I know in the short term Elon wants to “trim the fat” and stories like these are a clear sign to run for the hills, but I question if it will be...I’m curious how this affects employee retention; I know in the short term Elon wants to “trim the fat” and stories like these are a clear sign to run for the hills, but I question if it will be worth the long term damage to both Twitter’s engineering reputation and to Musk as an employer.
He’s always had a reputation but it’s feasible to see people falling for the cult of personality if it means working on new rockets at SpaceX. Working on sclerotic bay area big tech is a pretty different environment and working for Twitter suddenly moved near the bottom of the list—making a place a nightmare to work usually means attracting only people who don’t have better options.EDIT: With some thought, I think this might actually be the intended effect. If my boss told me I had a week to push a huge live production change or get fired, I'd tell him to get stuffed and walk. Part of having integrity as a professional is being able to push back on things and getting hit with a non negotiable wall strikes me particularly sourly. That said, it could be a pretty efficient way to weed out anyone in the company who isn't a back-bending sycophant that you can then walk over. Makes me think of military coups.
-
Comment on HBO Max adding ten original ‘Star Trek’ movies, bringing back all eight ‘Harry Potter’ films in ~movies
fional LinkI find this a bit hilarious in that if there was ever a reason I would subscribe to Paramount+ it would be as a one-stop-shop for everything Star Trek related (alas the newer offerings largely...I find this a bit hilarious in that if there was ever a reason I would subscribe to Paramount+ it would be as a one-stop-shop for everything Star Trek related (alas the newer offerings largely fail to impress, and DVD box sets of the classics are a one time expenditure). If they're scattering that offering out, I just don't see the value at all.
-
Comment on What's your unpopular opinion or idiosyncrasy about video games or games in general? in ~games
fional Link ParentI think the intrinsic problem is deeper and at the core there's always a conflict in prioritizing the game developer's story versus the player's story. Ultimately, a compelling story arises when...I think the intrinsic problem is deeper and at the core there's always a conflict in prioritizing the game developer's story versus the player's story. Ultimately, a compelling story arises when characters make interesting decisions with consequences, but "consequences" is almost directly equivalent to budget. At the end of the day, you can broadly bucket them into a few categories:
-
Fully prioritize developer agency: This is probably one of the more common strategies today, and boils down to "watching a movie with periodic gameplay breaks." This is the easiest route, you only have to develop a single timeline of content, but it's often incredibly unsatisfying narratively because you are, by definition, excluding the player from any consequential decisions. It further often drives you to avoid putting the main character into situations where they might actually make consequential decisions. Otherwise if the MC does something without the player's involvment, it can be REALLY frustrating (looking at you, Spec Ops: The Line). Examples: Most of the later Final Fantasy games (post-FFX, excluding MMOs). Uncharted, God of War, etc. style games.
-
Fully prioritize player agency: If you have a game whose gameplay mechanics intrinsically support consequential decision making, you don't need a story, the players will craft their own stories through the decisions they make. For example, "I was hauling a billion ISK worth of goods through null space when the Goonswarm capital ships started warping in and I was like ohhh fuuuu" (EVE Online), "I almost finished my fortress when a creeper blew a hole into magma and flooded the entire bottom floor!" (Minecraft), etc.
-
Multiple endings: This is a halfhearted attempt to shoehorn player agency into the first story model. You can't really change anything about the game itself, but some token decision you make towards the end gives you a different epilogue! This can help, but it usually suffers for one of two reasons. If you make the choice early in the game, you have the narrative dissonance of that choice not actually seemingly affecting anything for the rest of the game until the end scenes. This is often the problem with morality systems--you can press the "savior" button or the "be a jerk" button and your character portrait will change colors but it doesn't really do anything narratively because you still have to get to the same place in the end. Alternatively, you end up deferring the choice until the very tail end of the game at which it feels like a capricious decision (Mass Effect, Deus Ex: HR).
-
The tour guide: This actually can work, and is the route some of my favorite RPGs take. In this model, the main character is still essentially at the top-level a boring player stand-in, but can intercede within modular side-quest content. The decisions taken within those modules are consequential, but because they are bounded within a specific side-quest domain, they limit the amount of consequence-sprawl that can happen. The main arc of the story is still linear, though. Examples: Fallout NV, Chrono Trigger, Deus Ex).
-
Actually build the game and story together: this is the hardest approach because your story isn't something bolted onto an existing game, but both evolving in conjunction over the span of development. Sometimes, this means exploring a story that doubles down on the intrinsic lack of player agency by echoing it thematically in the setting and plot (Outer Wilds, Majora's Mask). Sometimes you play up the frivolity of the entire idea of agency and consequences (The Stanley Parable). Finally, sometimes you actually do give the player agency and double down on that choice being consequential (Undertale is the biggest example that comes to mind here). This is the rarest route because even with a single choice, you're often essentially forking the universe and developing multiple games in parallel.
That was a lot of words. Ultimately, I don't think game writing is bad because we undervalue narrative or writers--heck, I think most AAA games secretly wish they were movies or novels--but because AAA gameplay expectations and budgetary requirements preclude the necessary sacrifices to compellingly blend narrative and gameplay.
-
-
Comment on Are billionaires a market failure? And if not market, are they social failure? in ~finance
fional LinkI've always looked at market-based capital allocation as a system that functions best away from extremes. A certain sect of political thinking has associated the concept of "free market" as one...I've always looked at market-based capital allocation as a system that functions best away from extremes. A certain sect of political thinking has associated the concept of "free market" as one that is as free of regulation and intervention as possible, and this has resulted in unchecked growth in monopoly power at the company and personal level. When "too big to fail" sees zombie corporations limping along on market power and bailouts, when the little regulation that does get passed only serves to empower firms large enough to afford costly compliance mechanisms, and when tax law and loose monetary policy encourages stock buybacks and market concentration, it's hard to call the present world a good implementation of free market capitalism.
Instead, I see the original envisioning of a "free market" as one that maximizes competition, where there are few barriers to entry for new aspiring players, and relatively few mechanisms to prop up failing players that unfairly bias incumbent firms. Under this approach, we'd make policy to that effect:
- A laborer is a market player, and so a strong social net increases competition and free market effects. This means not being slaved to a corporate gig for healthcare, and having a viable alternative to working an untenable job that isn't "be funneled into the homelessness/crime death spiral." This might be better labor standards, a higher minimum wage, or some form of basic income. A company should not be able to unilaterally force unsustainable wages, schedules, or working conditions because they are the monopsony buyer of labor in a particular market.
- Companies that are "too big to fail" are "too big to exist" and should be broken up. We should be on constant lookout for firms whose structural implications grant its leadership to act with impunity knowing a bailout will save them.
- The legal system should not weight the relative depth-of-pockets of disputants in outcome. Today, it's very easy for a big fish to eat a little one just by virtue of the zero-sum arms race that is having an army of lawyers.
- Higher top-end tax rates. Just as a company can achieve monopoly power that allows it to drag on unproductively for decades without contributing value, there are levels of capital concentration where no single person can legitimately claim to be able to productively allocate said capital. This isn't "Harrison Bergeron," it's okay for wealth inequality to exist, but the range between the poorest and richest could be significantly compressed before people collapse into black holes of unassailable fortune.
Ultimately, I see capitalism a bit like running a nuclear reactor. There's a range in the middle where everything is great, but if there's too little (e.g. not enough societal infrastructure to support fair contract enforcement or market making), or if there's too much (markets become so influential that their gravity distorts the societal space around them), then either economic activity stops or blows itself up.
-
Comment on What have you learned from working in tech? in ~talk
fional LinkSoftware development (and technological development), particularly at higher levels of abstraction, feels like creation when in reality it is decisionmaking. At the beginning of a project (or...Software development (and technological development), particularly at higher levels of abstraction, feels like creation when in reality it is decisionmaking. At the beginning of a project (or startup), the world is a superposition of possibility, but every new piece of software, be it written from scratch or adopted from an external source, is a quantum entanglement that collapses your future possibilities. Interfaces get adopted, data formats become entrenched, conventions get established--these processes are not necessarily one-way but you will either pay back the initial time spent manyfold when circumstances change, or the whole endeavor collapses under its own weight. Better decisionmaking up front can quicken or prolong that process, but it is inevitable so long as your software lives in a changing environment (not that that's always the case, see the 20 year old machines running Win95 in airgapped installations). Part of my despondence with the tech industry is that the entrenchment of giant companies means that this exponential investment in prolonging can go on for MUCH longer (at much greater expense), where in the 90s or 2000s we'd just see companies replaced by other companies that could start with better-informed groundwork.
Corollary to the above: the hardest software development projects are rarely hard due to a technical challenge, but rather because people with decisionmaking authority are unable or unwilling to make decisions regarding their business or processes. Manual processes can accommodate significant slop in dissonant or outright conflicting decisions, by individual humans exercising judgment and massaging things at the edges (see, for example, how the healthcare system runs on fax machines because EHR systems just can't get along). Automating those processes, likely because the human fudge-factor is expensive, means trying to broker a peace between two or more warring conceptual factions. It's satisfying when it works, but it's a hard path to follow.
Software architecture is a hybrid of science fiction and journalism. Your success or failure stems from your ability to visualize and worldbuild in a (near) future fiction, and then bring it back to the present and document it clearly--that might be documents, that might be screenshots, that might be prototypes--such that you can get everyone involved living in that same future world. The further people (or teams') conception of that future world drift apart, the more problems you will have.
-
Comment on What are you battling with right now? in ~talk
fional Link ParentI am struggling in the same vein, albeit a decade earlier in my career. I’ve always embraced my identity as the “computer nerd” growing up, and it certainly lead me to a lucrative career, but the...I am struggling in the same vein, albeit a decade earlier in my career. I’ve always embraced my identity as the “computer nerd” growing up, and it certainly lead me to a lucrative career, but the further along I get, the more the job centers around navigating internecine corporate firedrills and trying to excavate megacorps out of the holes left by my forebearers.
I’ve been circling burnout for years, and increasingly find myself in a destructive loop of anxiety, procrastination, and questioning my own competency—surely anyone looking at me should tell I am not delivering results, so I fear for my job and try to push myself harder, but pushing harder just increases the stress without producing results.
Alas, I live in a VHCOL part of the world and switching careers at this point is so high risk that I cannot ever seriously consider it. I also feel the need to invalidate myself because I realize how precariously most people are living today. -
Comment on <deleted topic> in ~games
fional Link ParentI guess, as a counterpoint, I think when you strip away all the mechanisms of communication—be it game design or storytelling or music or photography—I think there’s a qualitative difference...I guess, as a counterpoint, I think when you strip away all the mechanisms of communication—be it game design or storytelling or music or photography—I think there’s a qualitative difference between something that was created by an author to convey some message or experience, and something that is pure mechanism, no message.
For example, I think modern AI experiments like GPT-3 and DALL-E are converging on mastery of their respective mediums, but it’s a real fear that a GPT-4 or GPT-5 could drown the internet in philosophical “p-zombies” that can say everything while saying nothing. Think like the Elsagate/Spiderman youtube videos that were autogenerated, yet occupied children for countless hours.
I see a few possible routes: you can have a human author that uses procedural generation as a force multiplier to do more with less, but you will ultimately run into the limits of their necessarily finite vision on what they want to convey. Alternatively, AI can become so good to effectively attain the level of personhood necessary to convey messages, in which case “procedurally generated” becomes “hand-authored by a sentient being coerced into doing so” which is a bit tragic. Or the p-zombie AI can get good enough that we’re all watching the adult version of Elsa/Spiderman, which is also kinda scary.
The future is weird. :)
To me, it sounds like overly heavily handed autotuning. I’m not sure if the little vibratos were in the original recording and then clamped down in post, or if they’re artificial vibrato added by the autotune plugin, but either way they sound weird.