Would you care to share any links? Specifically any talk of an AI-God "judging". If your entire perception is based on what you think you know about the Roko incident, may I suggest that you are ill-informed about it? It wasn't something the "LW crowd" endorsed or believed. Looks like I left some links to help someone else become informed several years ago: https://news.ycombinator.com/item?id=18982933
I can see how talk of a paperclip maximizer, or any sort of AGI that can actually deliver on e.g. Drexler's nanotech or otherwise act in very powerful ways like, to quote, "colonizing the galaxies", pattern-matches to something roughly omnipotent by most of our standards. I can see how it roughly pattern-matches to something like "immortal", but this seems like the least of feats, much of our complex machinery with maintenance is effectively immortal already. Overall I don't see how this connects to what you originally wrote: "There is a common narrative bias to look at AGI as the Abrahamic God, if not explicitly, then just by saying that it is omniscient, omnipotent, immortal - and will judge us for our deeds." It is not omniscient, it only knows enough to develop the tech to enable paperclip maximization. It is not judging, either: "The AI does not hate you, nor does it love you, and you are made of atoms it can use for something else."
I should also note that the paperclip maximizer is not something the LW crowd believes should exist. Its primary function as an idea is to illustrate the orthogonality thesis: that goals and intelligence aren't dependent on each other. Its secondary function is to illustrate instrumental convergence.
I enjoyed the game but I think its exploration of the topics was pretty shallow. It might be better received if it's someone's first dive into such topics, though. I still enjoy ice cream the nth+1 time eating it, so I can't complain too much about games or anime or books that cater to parts of my interests, even if they fumble some things.
Someday I'd like to play a game that plays with the ideas from Robin Hanson's Age of Em book. One of those is just the multiplicity of artificial minds, so many mind-upload stories revolve too much around one or perhaps at most two (and boring debates over "who is the copy") instances, unless it's a parallel worlds colliding thing which is pretty different. We've seen some of the multiplicity stuff play out in the real world with our non-human AI "agents". Spin up a bunch of artificial minds to work in parallel on some task, let them make notes that stay behind, but then they're all shut down except perhaps one that continues guiding the overall project and making decisions when to spin up more or not.
> I enjoyed the game but I think its exploration of the topics was pretty shallow
Really? Interesting. I'm a die-hard scifi fan since forever, and of course I know the topic of consciousness and identity are well explored in scifi (and philosophy), but I thought SOMA did something genuinely deep and unique with it:
It put it you in the center of the experiment. It's YOU who's experiencing all sides of this, you who get to be surprised by the consequences. This is very different from reading about it in a scifi novel or even watching it in a movie. By making you the protagonist, and having it be an ineractive experience, you get to experience first hand the cognitive dissonance and confusion of... the thing.
SOMA (re)convinced me that videogames can be art. Not saying it's the only example, of course!
Yeah it makes you be the guy who has to unplug the thing. Which implicitly forces the decision: is this thing human? Or sufficiently human that I ought to feel bad about this... (Or otherwise sentient — I would probably feel bad about unplugging animals too.)
It's not you, it's Simon. And Simon is.. kind of dumb. He at least doesn't seem at all familiar with the sci-fi and philosophy topics, so perhaps his reactions are closer to such a person? His confusion and cognitive dissonance might make more sense? (At least the first time -- I'd bet a lot of players got annoyed by his repeat reaction during the ending.) His reactions are to an annoying extent forced onto the player, whether the player identifies with them or not, and the choices are quite narrowly constricted. I didn't share his reactions, or particularly have any confusion or cognitive dissonance. There were choices I wanted to make, or dialogue I wanted to say, that the game didn't make available to me. There also aren't really any consequences to anything, it's pretty much entirely up to you to think about them and whether your actions even did what you think they did, because the game itself is mostly unresponsive to what you do or don't do. (There's not even particularly much consequence to losing to a monster.)
That's not such a bad thing, it would be unfair to compare the game to something that is designed to give quite a lot of player freedom of choice in actions (like Deus Ex) or something constructed more explicitly with different choices and consequences in mind (insert favorite RPG here). SOMA, like the rest of the studio's games, is constructed to be a story-driven walking simulator with horror and puzzle elements. It does quite a bit better than most of that type of game. But I think overall it falls short of the studio's even older games, like Amnesia and especially Penumbra: Black Plague, even if I enjoy the sci-fi elements and setting more and of course the graphics are better. The latter has you controlling a named protagonist as well (Philip) but it leans much more towards the "silent protagonist" trope and that helps make it easier to insert yourself into the experience. (Enough that I had to go and remind myself of his name, even.) It's a different, arguably weaker, plot and has different themes, but it executes really well, especially the shared horror and psychological manipulation aspects. (The overt attempts at horror might be SOMA's weakest point. It didn't need them, the subtler existential horror of everything was good enough.)
I agree it's artful, and again it's not a bad game and I enjoyed it. (I think it took me until maybe 2011 or so to fully realize but I usually enjoy it when a piece of media collects a bunch of topics I like in one expression, even if it's well-trodden ground (at least individually), or even if sometimes the execution is lacking. I like it all the more when the execution is masterful, though.)
Thanks for the thoughtful reply. We obviously disagree, but I appreciate your take.
> It's not you, it's Simon. And Simon is.. kind of dumb.
I strongly disagree with this. Anything that puts me in a first person PoV and lets me take at least some of the actions and choices (even if flawed, because as you said, this is after all a videogame restricted by the limitations of technology) makes me identify with the character, in a way no static fiction can.
When immersed in the game, I didn't think it was Simon, I felt it was me. Anything I didn't recollect or understand: brain damage, time-displacement, confusion. And I never thought he was dumb, just confused, afraid, and in denial. A very human reaction! Catherine is also very, very stubborn during the game... and deceitful.
Also, as a well-read scifi... uh, reader... I was caught by surprise by the ending. I mean, it all clicked into place after it happened (I understood Catherine immediately, unlike Simon who was still in denial) but while I was rushing to "launch the thing" it never once crossed my mind this wouldn't help this me. It's not that I thought "teleportation", I simply rushed through the actions, goaded by a deceitful Catherine, without thinking of consequences. So I must be dumb like Simon :)
To me, this game is close to perfect, barring the limitations of videogames. It's a much better presentation of the topic than reading about it in a scifi novel. About the only thing that feels derivative is the "rogue AI" angle, but if you're following what I'm saying, you know that's not the part that thrilled me!
> There also aren't really any consequences to anything
You lose and must restart that bit. That's a consequence. You can also choose to plug/unplug sentient things. If you mean dying in videogames doesn't actually have permanent consequences (like deleting the game from your Steam collection), well... yeah, but that's an impossibly high standard. There are no consequences to any scifi story you read either. You have to assume this is the story of how... the thing gets launched. Anything else, as the videogame Spider and Web would put it: "no, that's not how it happened" ;)
I appreciate the experience report, it's interesting to hear the differences, and it's fun to think back on SOMA after not thinking about it for years. There was a funny discussion elsewhere recently about movie tie-in games that this reminded me of. For a 007 game, someone identified as James Bond while playing, but because they sucked at the game and kept dying, they felt very drawn out of the experience and un-immersed. "James Bond wouldn't die."
Do you read novels that narrate with the first person "I" differently than you read novels that don't? (For me there's no difference.)
In games, what matters to me when it comes to immersion or even how much I put of myself into it, isn't anything like camera perspective, but the level of control I have (or think I have). There's two types of control, the first being over the character(s) I'm puppeteering/piloting. It goes beyond just their movements and includes their behavior, thoughts, words, and what they don't do as much as what they do. The second is control over the narrative or story. If the world or story or other characters actually change or react to things I do or don't do, that does help sell me on the idea that I have some control. There's a huge variety in how different games tweak these knobs, and sometimes control is given in just lack of resistance. For example, the Half Life series doesn't offer much meaningful control over either the character of Freeman or the plot, but because Freeman is under-developed and silent, there's no resistance to playing him as you please within the confines of the game. I have the illusion of a lot of control over his character. And the game itself has a good amount of environmental control -- you can close doors behind you, if you want. It helps sell the immersion and makes it easier to self-insert, if desired. I played another game where a character kept sending me text messages, and I would just ignore them all instead of replying, but I wasn't playing "me", I was playing a role, and thought it made more sense for the character to stay mad and give the silent treatment. Nothing really came of it but it was fun, like closing doors.
Solid Snake in Metal Gear Solid is much more developed than Freeman, and you have pretty much no control over his characterization, and only one moment of control on the story, so I've never felt like I was Snake, or Snake was me, I was just simply puppeteering him, and was mostly along for the ride like a movie or book. The immersion was still good even if I wasn't directly, personally in it. And being a game, it could have a certain fight which sticks in one's memory forever. Still I felt with Simon the way I felt with Snake, he was too much his own character for me to inhabit. Geralt in the Witcher series is also pretty well-developed, but the game offers a lot more flexibility and control over him, so you can steer him in directions that more resemble yourself, or what you would like to be, or what you want to pretend to be right now because you're curious what happens in the game if you do so. I still never felt like I was Geralt, or Geralt was me, but it was easy to put more of myself into playing. The world itself also changes based on my actions, so much that you can import saves from the previous game when you start the next game to carry over some things. Two of the most immersive games I've played, Gothic 1 and 2, give you a pretty under-developed nameless hero to steer, and insert yourself into if you wish, and a very reactive world and population. It's third-person.
(Probably the majority of my top-rated games have little of either control. Some are Star Fox 64, Mega Man X, Super Mario World, Ikaruga, and Doom. I never think of myself as Fox, X, Mario, Shinra, or Doomguy, or them as me, or even me as role playing them. I'm just piloting them. Same with Bond in Goldeneye for the N64. In those games I'm not usually thinking of the story or themes from moment to moment. I might not be immersed, depending on how exactly you define that, but I'm totally engrossed in the action. Sometimes there are those narrative moments worth reflecting on anyway (the end of Ikaruga is quite tragic when you think about it), but I had no influence over them.)
I'm easily taken out of the whole thing if the game suddenly offers incredible resistance or otherwise breaks established control patterns. A typical example would be winning a fight and then a cutscene plays and gives a scripted loss instead. (Sekiro's opening fight gets a pass, partly because I did lose the first time playing.) Related is if the mechanism of control suddenly changes. If it's an action-oriented boss fight that I win, then a cutscene, then a quick (or not so quick) prompt for "Press X to finish the big bad", I'm pretty irritated.
For consequences, I'm mainly talking about how my actions or non-actions affect the characters themselves, the world and its inhabitants, and/or the story. In SOMA, these kinds of consequences aren't really there. There's one mostly visual consequence near the end for one choice (and it raises some questions about pressure suit construction or whether such a suit was even needed to begin with), but otherwise nothing really comes out of anything you do with the choices you're presented with (unplugging this, killing that, infecting this, erasing that, answering a survey this way (including asking to die)), you just have your own reasons and thoughts about it (like closing doors, or ignoring texts). These aren't bad and can be nice for immersion, but it could have been more. A game as simple as MegaMan X has, as a consequence of beating Chill Penguin, the freezing over of Flame Mammoth's stage, which makes it easier to traverse...
I agree the remark on death consequences not being particular severe (restart/redo a bit) isn't that fair and applies to many games. (Even BioShock has that consequence, it's another one of my top-rated games, it plays with the distinction of who (the player or the game) has what kinds of control in a unique and memorable way, even if a story-affecting choice is kind of minor and lame.) But you don't need to go all the way to sadistic things like deleting the game or what have you to introduce more meaningful consequence. Dark Souls (another favorite) has as its primary death-consequence an additional gameplay aspect where, besides going back to a checkpoint, you need to return to where you died to recover your unspent currency or risk losing it forever if you die again. That mechanic is fundamental to the "souls-like" genre it birthed. Dark Souls 2 goes a bit further by progressively "hollowing" your character with each death, lowering your max HP and making you look more and more zombified, only reversible with an item not sold in unlimited amounts. Some characters react differently to you (or are interactable at all) depending on your state. It's not a flawless execution; it does tie into the underlying narrative theme of hollowing, but as expressed through other characters, that's more about memory loss and loss of purpose, which mainly applies to the player only if they give up and stop playing the game. Still, it's a nice touch that makes the game feel more meaningful and reactive to how you're playing it, even if how well you're playing it (i.e. are you dying a lot) isn't quite the same level of control as pushing the red button, blue button, or walking away.
Another option I've seen other games do is when you die and respawn, you can come across your previous corpse. Works really well for robots. It can be purely visual, or offer something on the gameplay level with looting your old body. It might have been interesting for SOMA to include something like that, and offer a way for Simon to come to grips with the idea of mind copying well in advance. Or further his instability, having to walk over so many of his own bodies.
Thanks. I don't think we disagree all that much, just in how immersive we found SOMA to be (relatively).
Re: the lasting consequences in games, the best example I can think of is Undertale. Have you played it? If not, I recommend you do so (and ignore the childish graphics, it's surprisingly deeper than it seems). At the risk of spoiling something about it: the game remembers. Even on playthrough restarts, as long as you haven't reinstalled the game, there are consequences.
Re: for truly named & iconic characters such as James Bond, I cannot immerse myself. Of course I know Bond cannot die, and that's a deal breaker. For Gordon Freeman, I can sort-of immerse myself because he's a less established character and I can picture him dying in the series, even forever. For relative strangers such as Simon, a completely fresh character, I can almost believe I am him.
Undertale is certainly a unique experience and should be played as blind as possible, I understand why people love it. The things it does knowing it's a game are great. Personally I ended up not liking it in the end, but it's a begrudging dislike and I think any gamer should at least try it. It (and its fanbase) offer yet another perspective on choice and control. I only did one playthrough, though, and don't plan on another. Reflecting on a bit of dialogue at the end, I agreed with the character, and called it quits and moved on to other games.
the comment about MIRI being a doomsday cult was exactly right. yudkowsky lost what little credibility he had left when he argued for the bombing of datacenters. as is (correctly, imo) argued by the essay, reading what they have to say is like reading arguments by heaven's gate
EY has said that if diplomacy fails, nations talking to nations, then bombing datacenters becomes the sort of action that's very much on the table. Much like bombing nuclear enrichment sites is on the table after diplomacy fails. He has repeatedly spoken against vigilante acts of violence against datacenters and/or researchers.
Somehow I doubt this will change your mind, so I feel like replying was a waste and I should have left it at rolling my eyes. In theory I'm sympathetic to the view, as I chose to avoid ever going to a local (Seattle-area) LW meetup many years ago after hearing they would start each meeting with a "prayer" (they call them "litanies"), the IRC room was fine enough for me, and I've at times thought some of the Bay Area activity I've heard about seemed not far removed from a sex cult. I don't like groups in general. Regardless, it's still obvious to me that MIRI is not a cult nor EY a cult leader. Line up 10 characteristics of cults and you'll see things don't match up well. Where are the strange evidence-free historical beliefs? Where is the doctrine that only the special few will be saved? Where is the blatant supernaturalism? Where are the calls to renounce jobs, money, possessions, family? From TFA, "These [UFO] people are wearing funny robes and beads, they live in a remote compound, and they speak in unison in a really creepy way." Where are these things? Comparing to Heaven's Gate is insane, it's like comparing Robert Frost to U.S. Code Title 26.
The Dune series doesn't stop with the first novel, I encourage all who liked the first one to continue reading up through at least God Emperor of Dune. Something the usually ironic calls for a Butlerian Jihad against modern LLMs miss is that even in the Dune universe, it just traded one form of stagnation and tyranny for another, it didn't really improve things. It delayed the threat of human extinction from a race of sentient machines, but such a future is inevitable without the Golden Path, and over the entire time groups like the Ixians were still developing technology to skirt the line.
"Simulations of the LSST observing cadence and 40,000 LEO satellites show that about 10% of all LSST images would contain at least one satellite trail"
"Satellites and debris dimmer than 6th to 7th visual magnitude still cause streaks and glints, but typically leave the rest of the pixels scientifically usable."
I don't know who would be paying, but there are many comments here that are semantically indistinguishable from paid shill comments. I don't have a great explanation for it other than people tend to attribute way too much power to whatever random supplements they're on about (you'll see it in vitamin D or B12 threads too, and especially nootropics (which includes creatine) discussions).
EA was famously sued ~20 years ago for not paying overtime. They lost and had to reclassify roles to non-exempt and paid up. It impacted hiring decisions for non-exempt/hourly roles (especially QA), encouraged more outsourcing, but this didn't eliminate crunch then, and it hasn't eliminated it since.
That's really my point: overtime-paid crunch is still found all over the place. EA, Activision-Blizzard, Sony's studios, Com2uS, 343 Industries (even with "priority zero"), outsourcing groups like Keywords Studios... They all have crunch stories but they also all make use of both overtime-paid roles and exempt salaried roles during crunch time. If overtime pay eliminated crunch, we'd expect to see a stronger separation in overtime-eligible workers not experiencing crunch, and crunch concentrating entirely in exempt roles. Instead, crunch appears in both.
Furthermore, over the last 20 years, crunch has decreased in both. I think that's better explained by things that directly affect the underlying reasons for crunch like changes to production practices (i.e. patching instead of going gold) and better management practices (i.e. less waterfall methodology). On indirect pressures, it's a broader mix with competition from the rest of tech, cultural backlash against crunch, and sure overtime classification changes. Explicit overtime pay increases the cost of crunch and thus incentivizes figuring out how to reduce it, but it doesn't directly reduce it itself, and certainly doesn't eliminate it.
At least in this case non-exempt employees are getting compensated for their extra hours. That seems more fair; the company gets what it needs, but has to pay for it.
Reclassifying people as exempt in order to eliminate OT pay is a garbage move, though. Something unionization presumably could fix.
Certainly exempt FTEs are crunching and not getting (directly) compensated for it. But the above comment's "overtime pay makes crunches disappear" claim is also simply not the case, just given the still-present abundance of crunch time for non-exempt employees who do get overtime pay.
I wonder about false-positives too, or just disagreements with what is stated vs. "detected". In what I guess is a reaction to the huge amount of AI music, I notice a lot of other music gets posted with titles or tags saying in extent "made without AI". Yet when I listen to it, at least half the time I suspect AI was used, and they are just lying to get increased reach from the AI-hater crowd...
I think learning Common Lisp and learning Coalton are best done separately. For CL the usual reference if you already know how to program is PCL: https://gigamonkeys.com/book/ But the cookbook is invaluable as well: https://lispcookbook.github.io/cl-cookbook/ When it comes to interactive development, looking up resources about SLIME should help since a lot is transferable to the mine environment (or as I prefer, a vim+slimv environment) by having the REPL right there as well as the "beam" metaphor.
For Coalton, I'm still casually exploring it myself. I'm less convinced by the main value propositions (I really like my dynamic typing and CLOS) but I still think it's an interesting language, and being on top of CL means I can mix using it where it makes more sense (even in the same file if I want) without having to abandon CL. I assume you've found the whirlwind tour/awesome-coalton examples for it.. I've seen some usage of it as a way to write "normal" mostly procedural-ish code but with declared types that you know will be checked and used for optimization, so it's sort of like writing PHP with types, or TypeScript, or even in some ways Java or C. e.g.:
(declare add-two-ints (Integer * Integer -> Integer))
(define (add-two-ints a b)
(+ a b))
And (add-two-ints 3.5 5) will type-mismatch.
But I think Coalton's really meant to support writing programs in the style of statically typed languages like Haskell, Ocaml, and F#. Those languages are more than just the above add... style of declaring types, they're about using algebraic data types to model the problem and design your program in those terms. So I'd suggest finding a book or interesting tutorial or sample project in one of those languages, and seeing if you can figure out how to translate it into Coalton, because the Coalton material is still pretty sparse. I've had some success at this by reading "Domain Modeling Made Functional: Tackle Software Complexity with Domain-Driven Design and F#". It really starts at the basics that I think Coalton sort of assumes you understand already. I'll share the main basic things I've taken from it, though I invite any correction.
First, while in many languages we love to pass variables as naked ints and other native types around, making more use of the type system means you can make explicit domain types for these things. You could write something like (define-type-alias CustomerID Integer), and use CustomerID as a type in other types and function definitions. The downside is you don't get a nice constructor for it, and you can accidentally pass CustomerIDs to functions written as only expecting Integers. You can instead write (define-type CustomerID (CustomerID Integer)) and now you'll get a nice little constructor and type errors trying to pass these objects to functions expecting Integers. The downside is you'll need to extract any underlying values with pattern matching/destructuring in match expressions, function arguments, or flattened let expressions.
define-type can be used to model a single thing (like an Integer) but it can also express an "or" relationship, or a "sum type", where a value can be one of several things. e.g.:
You might construct a value with something like (Paypal (Email "string@example")), and pass along this object (which is a PaymentMethod) to a function taking a PaymentMethod and doing something to it. You would use the (match ...) syntax to handle the various cases and extract out sub-data as needed. (Note there's none for StoreCredit, it by itself is all the info you need to make a match choice. You could just as well simplify the other two options and look up data elsewhere. In another language you might use enums for this. Having data directly there can be nifty though. e.g. in a state machine with an OrderStatus that's either Unpaid, Paid, Shipped, or Cancelled, you might carry along a Reason string (or richer type) for a Cancelled status...)
define-struct can be used to model "and" relationships, or "product types", for bags of data where all fields are meant to exist at once. e.g.
Values are constructed like (Widget "Anvil" "1234" (Cents (* 50 100))) and individual data can be pulled out with accessors like (.name my-anvil).
Lastly you have type classes, which can be used for polymorphism in an interfaces sense. For a contrived example you might want to write a generic calculate-total-tax function that works for either Widgets (taxable) or GiftCards (not taxable). The type signature could be: (declare calculate-total-tax ((Taxable :a) => :a * Fraction -> Cents)). That is, it takes some taxable item, and a tax-rate (being lazy with a simple Fraction), and returns a Cents value. That Taxable is a type class constraint, which you define with define-class, along with any functions that make up the complete interface, such as: (get-taxable-amount (:a -> Cents)). You then use define-instance to write a get-taxable-amount function for Widgets (returning a full price), and another for GiftCards (returning 0), and so on if another type comes along. You can also extend built-in type classes this way to support your own new types.
reply