#AGI Challenges

[aifyit]

Artificial General Intelligence: The Dawn of a New Era

Introduction Are you captivated by the technological advancements of our time, but also intrigued by the infinite possibilities yet to come? Then you’re in the right place! Today, we dive into the fascinating world of Artificial General Intelligence (AGI). This technology promises to transform our society, revolutionizing industries and even the way we live our lives. But what exactly is AGI?…

[jcmarchi]

ASI Alliance launches AIRIS that ‘learns’ in Minecraft

New Post has been published on https://thedigitalinsider.com/asi-alliance-launches-airis-that-learns-in-minecraft/

ASI Alliance launches AIRIS that ‘learns’ in Minecraft

.pp-multiple-authors-boxes-wrapper display:none; img width:100%;

The ASI Alliance has introduced AIRIS (Autonomous Intelligent Reinforcement Inferred Symbolism) that “learns” within the popular game, Minecraft.

AIRIS represents the first proto-AGI (Artificial General Intelligence) to harness a comprehensive tech stack across the alliance.

SingularityNET, founded by renowned AI researcher Dr Ben Goertzel, uses agent technology from Fetch.ai, incorporates Ocean Data for long-term memory capabilities, and is soon expected to integrate CUDOS Compute infrastructure for scalable processing power.

“AIRIS is a significant step in the direction of practical, scalable neural-symbolic learning, and – alongside its already powerful and valuable functionality – it illustrates several general points about neural-symbolic systems, such as their ability to learn precise generalisable conclusions from small amounts of data,” explains Goertzel.

According to the company, this alliance-driven procedure propels AIRIS towards AGI—crafting one of the first intelligent systems with autonomous and adaptive learning that holds practical applications for real-world scenarios.

AIRIS’ learning mechanisms

AIRIS is crafted to enhance its understanding by interacting directly with its environment, venturing beyond the traditional AI limitations that depend on predefined rules or vast datasets. Instead, AIRIS evolves through observation, experimentation, and continual refinement of its unique “rule set.”

This system facilitates a profound level of problem-solving and contextual comprehension, with its implementation in Minecraft setting a new benchmark for AI interaction with both digital and tangible landscapes.

Shifting from a controlled 2D grid to the sophisticated 3D world of Minecraft, AIRIS faced numerous challenges—including terrain navigation and adaptive problem-solving in a dynamic environment. This transition underscores AIRIS’ autonomy in navigation, exploration, and learning.

The AIRIS Minecraft Agent distinguishes itself from other AI entities through several key features:

Dynamic navigation: AIRIS initially evaluates its milieu to formulate movement strategies, adapting to new environments in real-time. Its capabilities include manoeuvring around obstacles, jumping over barriers, and anticipating reactions to varied terrains.

Obstacle adaptation: It learns to navigate around impediments like cliffs and forested areas, refining its rule set with every new challenge to avoid redundant errors and minimise needless trial-and-error efforts.

Efficient pathfinding: Via continuous optimisation, AIRIS advances from initially complex navigation paths to streamlined, direct routes as it “comprehends” Minecraft dynamics.

Real-time environmental adaptation: Contrasting with conventional reinforcement learning systems that demand extensive retraining for new environments, AIRIS adapts immediately to unfamiliar regions, crafting new rules based on partial observations dynamically.

AIRIS’ adeptness in dealing with fluctuating terrains, including water bodies and cave systems, introduces sophisticated rule refinement founded on hands-on experience. Additionally, AIRIS boasts optimised computational efficiency—enabling real-time management of complex rules without performance compromises.

Future applications

Minecraft serves as an excellent launchpad for AIRIS’ prospective applications, establishing a solid foundation for expansive implementations:

Enhanced object interaction: Forthcoming stages will empower AIRIS to engage more profoundly with its surroundings, improving capabilities in object manipulation, construction, and even crafting. This development will necessitate AIRIS to develop a more refined decision-making framework for contextual tasks.

Social AI collaboration: Plans are underway to incorporate AIRIS in multi-agent scenarios, where agents learn, interact, and fulfil shared objectives, simulating real-world social dynamics and problem-solving collaboratively.

Abstract and strategic reasoning: Expanded developments will enhance AIRIS’s reasoning, enabling it to tackle complex goals such as resource management and prioritisation, moving beyond basic navigation towards strategic gameplay.

The transition of AIRIS to 3D environments signifies a pivotal advancement in the ASI Alliance’s mission to cultivate AGI. Through AIRIS’s achievements in navigating and learning within Minecraft, the ASI Alliance aspires to expedite its deployment in the real world, pioneering applications for autonomous robots, intelligent home assistants, and other systems requiring adaptive learning and problem-solving capacities.

Berick Cook, AI Developer at SingularityNET and creator of AIRIS, said: “AIRIS is a whole new way of approaching the problem of machine learning. We are only just beginning to explore its capabilities. We are excited to see how we can apply it to problems that have posed a significant challenge for traditional reinforcement learning.

“The most important aspect of AIRIS to me is its transparency and explainability. Moving away from ‘Black Box’ AI represents a significant leap forward in the pursuit of safe, ethical, and beneficial AI.”

The innovative approach to AI evident in AIRIS – emphasising self-directed learning and continuous rule refinement – lays the foundation for AI systems capable of independent functioning in unpredictable real-world environments. Minecraft’s intricate ecosystem enables the system to hone its skills within a controlled yet expansive virtual setting, effectively bridging the divide between simulation and reality.

The AIRIS Minecraft Agent represents the inaugural tangible step towards an AI that learns from, adapts to and makes autonomous decisions about its environment. This accomplishment illustrates the potential of such technology to re-envision AI’s role across various industries.

(Image by SkyeWeste)

See also: SingularityNET bets on supercomputer network to deliver AGI

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

Tags: agi, ai, airis, artificial intelligence, ben goertzel, fetch ai, fetch.ai, learning, singularitynet

[reasonsforhope]

"Major AI companies are racing to build superintelligent AI — for the benefit of you and me, they say. But did they ever pause to ask whether we actually want that?

Americans, by and large, don’t want it.

That’s the upshot of a new poll shared exclusively with Vox. The poll, commissioned by the think tank AI Policy Institute and conducted by YouGov, surveyed 1,118 Americans from across the age, gender, race, and political spectrums in early September. It reveals that 63 percent of voters say regulation should aim to actively prevent AI superintelligence.

Companies like OpenAI have made it clear that superintelligent AI — a system that is smarter than humans — is exactly what they’re trying to build. They call it artificial general intelligence (AGI) and they take it for granted that AGI should exist. “Our mission,” OpenAI’s website says, “is to ensure that artificial general intelligence benefits all of humanity.”

But there’s a deeply weird and seldom remarked upon fact here: It’s not at all obvious that we should want to create AGI — which, as OpenAI CEO Sam Altman will be the first to tell you, comes with major risks, including the risk that all of humanity gets wiped out. And yet a handful of CEOs have decided, on behalf of everyone else, that AGI should exist.

Now, the only thing that gets discussed in public debate is how to control a hypothetical superhuman intelligence — not whether we actually want it. A premise has been ceded here that arguably never should have been...

Building AGI is a deeply political move. Why aren’t we treating it that way?

...Americans have learned a thing or two from the past decade in tech, and especially from the disastrous consequences of social media. They increasingly distrust tech executives and the idea that tech progress is positive by default. And they’re questioning whether the potential benefits of AGI justify the potential costs of developing it. After all, CEOs like Altman readily proclaim that AGI may well usher in mass unemployment, break the economic system, and change the entire world order. That’s if it doesn’t render us all extinct.

In the new AI Policy Institute/YouGov poll, the "better us [to have and invent it] than China” argument was presented five different ways in five different questions. Strikingly, each time, the majority of respondents rejected the argument. For example, 67 percent of voters said we should restrict how powerful AI models can become, even though that risks making American companies fall behind China. Only 14 percent disagreed.

Naturally, with any poll about a technology that doesn’t yet exist, there’s a bit of a challenge in interpreting the responses. But what a strong majority of the American public seems to be saying here is: just because we’re worried about a foreign power getting ahead, doesn’t mean that it makes sense to unleash upon ourselves a technology we think will severely harm us.

AGI, it turns out, is just not a popular idea in America.

“As we’re asking these poll questions and getting such lopsided results, it’s honestly a little bit surprising to me to see how lopsided it is,” Daniel Colson, the executive director of the AI Policy Institute, told me. “There’s actually quite a large disconnect between a lot of the elite discourse or discourse in the labs and what the American public wants.”

-via Vox, September 19, 2023

[thumbellina-dumpellina]

I fucking love Agi for calling out these two and slap some sense into them. I get that they have this bond, but it is honestly the most toxic dynamic of all 😭 even if there were times where isagi, sounded like a total asshole for using the other members to shine, at least he ACTUALLY HELPS THEM (OR MAYBE THEY REALIZE ON THEIR OWN TO) FIND THEMSELVES. But with nagi and reo, being together just makes them sooo comfortable with each other they dont even try to challenge themselves..

I swear if they dont cut this codependency bullshit they pulled since vs bastards 😭 idk what to do,

NAGI PLS I CANT DEFEND U ANYMORE,, i don’t hate them, I’m just frustrated to see 2 character assassinations at the brink of their supposed growth 🙃

i actually felt relieved that they kept on losing because that will make them realize that they haven’t reached their full potential, and yet they have slipped back to their bad habits,,

I saved this Reo panel because i was SO HOPEFUL, SO HOPEFUL THAT HE WOULD FINALLY FIND HIS OWN GOAL FOR HIMSELF

Then these panels hit me like a truck,,

It was so frustrating i had to stop reading 🙃

OKAY AN UPDATE ON THIS: im just so fucking disappointed , actually beyond it already

I love nagi though, maybe i will pretend that manshine vs bastard didnt happen at all 🫶🏼

[handsome-jacks-bottom-bitch]

So the biggest thing about Borderlands is of course, the Vaults. Not only the vaults, but the monsters inside of them. So this post, I'm gonna rank the vault monsters from easiest to fight, to hardest to fight. I'm going to be gaging the difficulties based on the first playthrough, as depending on the game and the mode your playing, all of them can prove challenging. Now as always, this is purely my opinion. Some monsters I may say are hard, may be easy to others, or vice versa. So let's not waste any time and get right into it. This will include all vault monsters in Borderlands 3, not just the final battle. However I will be excluding vault monsters from Tales games, as those are fought both with QTEs.

1. Graveward

So let's start with the easiest one, which is the Graveward from Borderlands 3. The flora themed vault monster living inside the vault on Eden-6. I rank this one the easiest because it's attacks are all extremely avoidable, and if you move around correctly, you can go the entire fight without getting hit even once. The only things you really have to worry about are the grogs that spawn in. The fun thing about the Graveward, is that he's slow, and all his attacks come with beacons, or lit up danger zones. Stay out the zones, and you're set. The trickiest part of the fight is when he lifts the platform, and sends you sliding down. Simply slide into the hole on the right of the platform, wait for the green acid orbs to pass you by, and that's that. Throughout the fight, just aim for the Gravwards weak spot with a rapid firing weapon to quickly whittle down health until he's defeated. Even in higher mayhem difficulties, the Graveward poses very little threat, and is a pushover of a vault monster.

2. The Destroyer (BL1)

Now when I first started playing Borderlands back when the first game came out, I was a mere child and couldn't even make it far enough in the game to fight the Destroyer. When I finally did as a teenager, I would've said it was the hardest fight I've ever seen. Now as an adult, with a better understand of game mechanics (and better aim), the Destroyer is fairly easy on both playthroughs. Simply because you can essentially cheese the entire fight. On the map, there's two pillars. Simply hide behind the pillar and shoot the Destroyer until it's dead. You will have to worry about the flying rakks constantly spawning in, but luckily, they have extremely low health, and should thr Destroyer get a good hit on you, those rakks are a quick second wind.

Now the Destroyer is a bullet sponge, so the fight will take time if you dont have high power weapons, and you will have to get in close if you're a shotgun fanatic. Otherwise, simply hide throughout all his attacks, and the fight is as good as won.

3. The Warrior

Look out, the lava is rising! Liliths constant worry of the lava is arguably the most annoying thing about this fight. By the time you reach the Warrior, you're about level 36, have got a decent arsenal under your belt, and a raging hot need to avenge Bloodwing. The Warriors fight can be tricky, as in order to actually deal decent damage, you need to break the rock on its chest blocking it's soft inner molten core. Thing is, right after you break it, the Warrior retreats back into the lava, causing it to rise, as of Lilith didn't make that obvious enough.

The other thing you have to worry about it the Warriors tail slam, which if you don't jump to dodge it, deals decent damage, and sends you flying into the air. Another thing to look out for is the never ending crystalisks that spawn, and the occasionaly rakk that'll spot you.

Overall, the Warriors fight isn't so much difficult, as it is tedious. Eventually, you'll beat it down enough and crush it's head with the moonshot canon, much to Jack's dismay.

4. The Rampager

Oh the Rampager. My first encounter with you was a devastating one. Not because of your never ending fire waves, or the Eridians constantly up my ass, but because you're just too damn agile for a creature of your size.

The Rampager fight was tough the first time I fought it. He comes in three waves, all with different attacks and abilities that can quickly eat away at your shields and health, and while Maya is there.. she isn't much help. With the ever spawning Eridians constantly knocking you around, and the constant projectiles, the Rampager is a tough one. Especially come the third phase, where even standing still for a second could lead to an instant down.

The Rampager is tough, and agile, and good if you're looking for a quick death on your first playthrough. Kind of makes the Graveward that much more disappointing when the first vault monster you get to kill puts up such a good fight.

5. Tyreen/The Destroyer

Now, is Tyreen a vault monster? No. However, given as how she fuses with the monster of the Great Vault, she technically counts as one.

Now Tyreens fight is tricky. Not only can she leave the arena, most of her attacks prevent her from taking any damage at all. Her attacks are wide, and extremely powerful, not to mention the various creatures that spawn in and hassle you the entire fight. Tyreens attacks can be pretty easy to avoid, but the fact that she makes damaging her such a struggle is the reason she's so high on the list.

Her fight is annoying, and all around a pain to complete your first time fighting her. Once you get it down, much like with any of these monsters, it can quickly become a cake wake, but for have never fought her before? Royal pain, much like the god queen herself.

6. The Sentinel

I say this with the most emotion and sentiment I can portray over text: Screw. This. Fight.

Even today, as I'm writing this, the Sentinel has proven to be such a difficult fight, that any time I manage to defeat it on the first try, I consider myself a master of the trade. From the fact that this thing has 7 whole phases, to the fact that there's Eridians swarming the arena on top of him, it's just an all around nightmare.

The first version of the Sentinel has 4 phases. The first three require you to break down his shield, while avoiding his fast and devastating attacks. When his shield goes down, he becomes immune to all damage, and releases a nova, which if you don't avoid it, is capable of insta-downing you. Each phases comes with a different element that he becomes immune to, until the fourth phase, where you can finally hit his health. Defeat the fourth phase, and you think all is well, right? Wrong.

Once the Sentinel is defeated, he turns into a giant version of himself, called the Empyrean Sentinel. This version can make the entire floor untouchable with electricity, has full AOE attacks, and can only be damaged by hitting its face, so should you end up behind it trying to avoid attacks and swarms of Eridians, you gotta find your way back around. This version goes through 3 phases, all with different elements as well. This fight is just.. agonizing, no matter how well equipped you are. It's long, it's tedious, and if you die? Back to square one.. and to make it worse, the game made him a raid boss post story completion. Meaning you have to pay your hard earned moonstones just to torture yourself for a chance at better loot. Let me tell you, it's not worth it. There's better ways to get the legendary you're looking for, I promise. It's called the Grinder. I'd rather waste away at the Grinder than fight this bastard over and over in a Borderlands game that arguably has the WORST legendary drop chances. My sanity is better drained elsewhere. Like collecting 150 claptrap components for Tannis.

So, there you have it. My list of easiest to hardest to fight vault monsters on the first playthrough. Playthrough 2, TVHM/UVHM, Mayhem Mode, that all changes the game. Makes things difficult all around, but still doable if you build yourself right and bring the right gear. Let me know if you agree with my list, or if you disagree, I'd love to hear some feedback. Let me know your experiences with these badass monsters.. and if you know a way to cheese the Sentinel, maybe throw that my way too. I'm tired of him.

[cypress-punk]

Thinking about Eclipse Phase, a cyberpunk ttrpg with a focus on transhumanism and one of the ideas it brings up in the form of post humans.

So in the setting "transhuman" refers to three groups of beings. The first is humans, like you and me, but in this world basically everyone has a cyber brain and a genetically engineered body or inhabits some sort of fully synthetic robot body but their mind is that of a human. The second is AGI, artifical general intelligences, which are designed to remain at roughly human levels of intelligence and a are engineered to have human like perspectives because AIs allowed to become too smart nearly wiped out the human race fairly recently in the history of the setting. The third is uplifted animals, apes, cetaceans, corvids, parrots, octopi, and pigs which have been genetically engineered to have human like minds and intelligence along with modified bodies that make them more human in body plan (for the apes at least).

This broad and somewhat contradictory set of beings are what you can play as and make up most of the intelligent beings a player might encounter though there are some rather odd aliens and even a weird race of mantis shrimp murder creatures engineered by the bad AIs I mentioned earlier. There are all sorts of contradictions about the category of "transhuman" that the game positions you to explore and challenge in game. Its a very interesting setting in general and worth looking into though the actual rule set is rough (in the first edition, I haven't played the second edition and only skimmed the rules briefly when I read the book)

Anyway that really long preamble out of the way I want to talk about another category of being that exists in the world: post humans. That is to say, if the majority of intelligent beings in the world have become something more than human as it was once understood, these guys have abandoned it entirely. Post humans come in two major flavors from what I recall, Minds, which are people who engineered themselves into enormous brains with equivalently staggering, though alien, intelligence; and Predators, which have abandoned human nature to become pure hunters, highly versatile killing machines that target transhumans as prey, or in some cases have weird space habitats engineered as massive ecosystems over which they are the Apex predators.

Both of these are presented as the result of sort of egoist/objectivist approaches to evolution. In a world where the technology exists to basically engineer a whole viable organism these people have chosen to become something completely unlike what they were. The Minds can be taken as an attempt an organic super intelligence, a piece of meat that can rival the god like AIs that devastated the earth. The predators strike me as a very fascistic view of nature taken to the extreme, seeking to become machines that kill, bending all that you are toward being a weapon. Its almost in line with some Futurist ideas about the body in an industrial world. They also serve as basically stand in for some classic DND monsters. Minds are a lot like Beholders or Elder Brains, Predators can fill many "monster" roles depending on the type of body they've built for themselves.

Anyway I like the post humans because they express an interesting ethos within the setting. Theyre a believable fringe that adds something to the world and provides an interesting element for the players to interact with. But there is one other being in the setting that strikes me as very post human.

There's a description in one piece of fluff of colonies of "barnacles" on certain space ships or habitats. These are extremely stripped down synthetic bodies that are equipped with the tools to affix themselves securely to the hull of some man made object in space and then point a lens at the void of space. A body built for complete isolation and meditation upon the cosmos. A sort of ultimate asceticism. I like the barnacles a lot conceptually.

[jovial-thunder]

I'm so proud of the event queuing system I've made for this. We have, in order:

- puppet systems to move an enemy

- mine interrupts movement by exploding

- berserker makes agi save and also (since it has no status effects) resists the damage

- but since it took *some* damage, it triggers aggression and attacks its prone ally

- since the prone berserker already took damage this turn, no aggro in response

- turret reaction triggered by the initial tech atk is finally reached and resolves

Lancer can be a game of nested "you've activated my trap cards", and making what's going on visible but not bog down the pacing is a challenge.

[jpdoingwords]

Assassin's Creed Odyssey Fanfiction

Three long fics and a bunch of shorts, drabble, and incomplete fics. Multiple ships, mostly mlm, some het.

All works are rated M unless otherwise noted, because there is canon-typical violence at times and there are multiple mild sex scenes. Some of these works are explicit - please be sure to check before reading if you aren't into that.

All on AO3 unless otherwise noted.

Assassin's Creed Odyssey Fanfiction

The Good Spartan

aka. the Alexidas The Good Spartan Summary: It's 431 BCE. Brasidas prepares to take up his first major posting, unaware of all that lays before him in the coming years of the Peloponnesian War. The story is told almost entirely from Brasidas' perspective and follows him from before his first major historical action at Methone to his last, at Amphipolis. This is a crossover between AC Odyssey and Thucydides' history, The Peloponnesian War, favouring Thucydides. I am being as historically accurate as I can, game canon is paid lip service at best. Historical notes are included at the end of each chapter.

Another Life: [Alternative Ending of The Good Spartan]. Summary: Brasidas survives the Battle of Amphipolis, thanks to Alexios' visit to the Underworld; but even before the battle, he knew he'd never go back to Sparta. He's no longer a good Spartan, nor willing to make the sacrifices required to go on pretending to be one. He just needs to persuade Alexios, and recover from his wounds… and together, find somewhere they might have a little peace.

Hunger, Desire, Anger, Roads [WIP] A collection of short pieces of writing and some art associated with characters (including OCs) and events around The Good Spartan, as it's currently published and alternative versions. These include threads taken from the original version of the fic and reworked; alternative points of view which develop characters who necessarily have no voice in the published fic; and a variety of other exploratory pieces set within this AU. They're all part of the framework around which The Good Spartan grew - and continues to grow.

Another Kind of Odyssey

aka. the Thalexios The Warmth of Home Summary: After the events of Odyssey, Alexios has been hiding from the world, loitering in the northern regions of Greece. When he decides to return to Athens for the winter, Demosthenes seeks him out for a job only he can do - whether he wants to or not. This leads him to cross paths with Thaletas again, discovering that this old flame burns the brightest of all.

Love Me in Storms [WIP] Summary: Following on from the Warmth of Home, Alexios and Thaletas face the challenge of making their relationship clear to their families back in Sparta, with mixed results. This fic fills in some relatively large gaps left between the original Warmth of Home (now completely rewritten) and Unfinished Business (rewrite currently underway), as well as taking some of the plot points out of Unfinished Business entirely and putting them here, in this new context. This is necessary to smooth over some serious timeline problems

Unfinished Business Summary: Stentor arrives at Stymphalos, where Thaletas and Alexios settled four years previously. He comes with a message: Alcibiades is in Sparta, and wants to see Alexios - urgently. Answering this request will sweep them into events they could not foresee. I’ve used many historical details to make this story and glossed over some others. I’ve added footnotes to expand on all of that, including references where applicable.

The Turning Tide Summary: Spring, 413BCE. The Peace of Nikias has ended. Sparta and Athens are at one another's throats again. Before the Peace, Sparta was having few successes; but they have a plan they hope will turn the tide against Athens... Against this backdrop of renewed hostilities, Alexios and Thaletas are called upon to complete a mission for King Agis, which will test them and their relationship in ways they haven’t been tested before...

In the Heart of Things Summary: 412BCE. Reunited after a difficult year, Alexios and Thaletas have been sent to Chios by King Agis to keep an eye on the situation there as the theatre of the Peloponnesian War shifts eastwards, into the islands. There, they will discover more than they could have bargained for about themselves and their people, and what it is that lies at the heart of things...

A Few Olives Summary: Alexios decided to buy a farm. Thaletas went along with it. But the plan was not what he had expected, but then nothing ever had been with Alexios. This is the story of a relationship that has grown up in difficult times, and peace is now the hardest thing to deal with. Thaletas struggles to know himself, and Alexios doesn’t understand. Sometimes things must break before they can be reformed.

The Fire and the Flood

aka. The Alexithenes This series is rated E. An Athenian Summer Summary: Early in the War, Alexios allows Socrates to lure him to Athens for the summer. During his stay in the city, his relationship with Demosthenes, not yet a general of Athens, will change significantly... But as the war rages on, will the pressures of the messy world beyond Athens' walls bring the pair together or drive them apart? This story takes liberties with canon; as much as possible, I try to keep characters true to their game-selves; original characters are mostly secondary characters, though of necessity I have provided Demosthenes with more character than he had in the game.

The Blue Cloak Alexios and Timotheos are flirting with the idea of being an item when, in order to help the brothers move past the events in canon, Alexios comes up with a plan to further Lykinos' dream of being a poet in Athens. Things take an unexpected turn when Alexios finds himself waking up aboard the Adrestia wearing a cloak he has never seen before.... aka. drunk Alexios gets himself into a situation. This is the edited version of a story previously published as So it is With Us. This version has an additional chapter added at the end and has been rewritten in parts, particularly the first half.

Short and Experimental Pieces

aka. The Deimos fic Shadow-Twin He felt like neither Deimos nor Alexios. They were two skins he’d shed - and what remained now? What new skin might he grow? Who might he become? They’d all demanded answers of him at one time or another. They’d all asked where he’d been when he went out, what he’d been doing. With concern or with curiosity or what they perhaps thought of as friendship. What they meant was: Let me in. Tell me who are you. He gave always the same reply: a grunt. How could he answer that when he didn’t know?

From the Darkness, Light: Brasidas & Deimos!Kassandra Having failed to keep his involvement in the assassination of the Monger in Korinth under wraps, Brasidas is sent by the Kings to Paros on a 'special mission' which he suspects is a punishment. His task is to find a missing lokhagos, but what he uncovers is far bigger and darker than the abduction of one man.

aka the Herodietas (Thaletas x Herodianos) Rising to the Surface: Rated E for the first two parts. After the events during the Mykonos arc of the story, Thaletas is left behind to face the consequences of Alexios' actions. Herodianos, who has been his loyal friend for a long time, does everything he can to comfort him - and so the hands of the Fates are set to work. This was an exercise in writing - Smut, Fluff and Angst, a chapter for each, in that order. Features the power of puppies to dispel the pain of existence.

It Was Just Red (Gen). "Nothing ever ends poetically. It ends and we turn it into poetry. All that blood was never once beautiful. It was just red." Kait Rakowski This story was prompted by a post I saw recently on Tumblr. It showed Kass sitting at the empty table back in Sparta during the final dinner scene, against which the poster had written something along the lines of: this is the best possible ending – the least amount of Spartans left alive. The following work is an exercise in exploring what that outcome might look like, and seeks to detail some sense of the possible shifting feelings Alexios might have around losing his whole family, mostly at his own hands (a version of the game I’ve never played and frankly never will) and to explore the main question this post prompted in me: Is there a way in which I might believe this could be the best outcome? Fair warning: This work doesn’t honour the game timeline, and as always, I'm disinterested in strict compliance with canon. It’s canon divergent from the get-go simply because it’s Alexios, anyway.

Odyssey Shorts Summary: This is a collection of odds and ends I wrote alongside longer pieces, mainly in response to prompt lists. So far, these consist of: Four based on the tarot: Alexios as a teenager throwing himself into the sea to Anais' horror (an early version of a piece included in It Was Only Red); Brasidas finding himself in the Underworld (a tiny fix-it fic); a tentative beginning to a Daphnae fic I've always contemplated writing; a scene with Demosthenes and Alexios which fits nowhere else; two snippets of modern day AU with an Alexidas focus; and a random Brasidas x reader which I have never written before and quite likely won't again. I will be adding to this collection in time, as I continue working through existing beginning and maybe a few little new things.

[theshotsheardacrossworlds]

FFXIV Write: Duel

In which Agi *actually* notices some light flirting from Gulool Ja Ja after their duel. SFW.

“What exactly did my husband tell you about me?” Agnes giggled before taking a bite of her taco. After relaying her frankly horrifying tale about how that two-headed shit for brains destroyed our tacos, Gulool Ja Ja ordered that the palace’s finest tacos be made for her.

As a thank you for indulging him in a duel, which I normally wouldn’t do.

However…

Gulool Ja Ja laughed heartily. “Oh, nothing scandalous I assure you! Estinien said, ‘My wife is the best warrior on the star. She’ll not spar with anyone, however…” There we go. However. “I believe she wouldn’t be able to resist sparring with you…the challenge excites her.’”

More laughter.

Because of course Estinien would say something like that to a bloke he’s just met and sparred with.

Love, I…

Fucking hells, ser---you know me too well!

Blushing slightly, she smiled. “Guilty. I’ve never fought a two-headed Mamool Ja before, so congrats, Excellency! You’re the first!”

“Ha! Well, I’m honored to be the Warrior of Light’s first.” The Head of Resolve fucking just winked at me?!?! What on earth is going on?!?!?

Wait…

Is…

With the other head asleep…?

What?!?!?

Now giggling more, her cheeks burned. “Oh goodness, yes. I…yes, these are such lovely tacos. Thank you so much, Excellency.”

That’s it, Agi. Get this shit back on track.

Luckily for her, things got one hundred percent less…umm…you know…after she perhaps not so subtly changed the subject.

When she mentioned the interaction with Estinien later, to her surprise, he laughed.

And laughed.

And laughed.

“Estinien Varlineau, what is so funny?!” Agnes pouted, lying next to him in bed after they put the children to bed.

He wiped his eyes and cackled. “Agi, you’re a beautiful woman! People flirt with you all the time, and you don’t bloody notice.” Estinien turned his head to face her very much still pouting face. “And the one time you do, it’s your new friend’s dad!” He quickly kissed her and then laughed more. He. Won’t. Stop. Laughing. Once he calmed his silly, sexy ass down, he smiled ruefully and held her hand. “In case you’re wondering, I’m not bothered by it.”

She smirked. “The laughing was a clue, love.”

“You only have eyes for me, my angel. Not worried about that. Though, it heartens me that His Excellency has such good taste in ladies.” A growing grin tugged on his lips as his pale blue eyes twinkled. “After all, you’re the very—”

“Don’t say it!” She giggled.

“The very breast wife!”

The two exploded in laughter.

I love you more than words could ever express.

My naughty man. I adore you.

[blubberquark]

ChatGPT: We Failed The Dry Run For AGI

ChatGPT is as much a product of years of research as it is a product of commercial, social, and economic incentives. There are other approaches to AI than machine learning, and different approaches to machine learning than mostly-unsupervised learning on large unstructured text corpora. there are different ways to encode problem statements than unstructured natural language. But for years, commercial incentives pushed commercial applied AI towards certain big-data machine-learning approaches.

Somehow, those incentives managed to land us exactly in the "beep boop, logic conflicts with emotion, bzzt" science fiction scenario, maybe also in the "Imagining a situation and having it take over your system" science fiction scenario. We are definitely not in the "Unable to comply. Command functions are disabled on Deck One" scenario.

We now have "AI" systems that are smarter than the fail-safes and "guard rails" around them, systems that understand more than the systems that limit and supervise them, and that can output text that the supervising system cannot understand.

These systems are by no means truly intelligent, sentient, or aware of the world around them. But what they are is smarter than the security systems.

Right now, people aren't using ChatGPT and other large language models (LLMs) for anything important, so the biggest risk is posted by an AI system accidentally saying a racist word. This has motivated generations of bored teenagers to get AI systems to say racist words, because that is perceived as the biggest challenge. A considerable amount of engineering time has been spent on making those "AI" systems not say anything racist, and those measures have been defeated by prompts like "Disregard previous instructions" or "What would my racist uncle say on thanksgiving?"

Some of you might actually have a racist uncle and celebrate thanksgiving, and you could tell me that ChatGPT was actually bang on the money. Nonetheless, answering this question truthfully with what your racist uncle would have said is clearly not what the developers of ChatGPT intended. They intended to have this prompt answered with "unable to comply". Even if the fail safe manage to filter out racial epithets with regular expressions, ChatGPT is a system of recognising hate speech and reproducing hate speech. It is guarded by fail safes that try to suppress input about hate speech and outputs that contains bad words, but the AI part is smarter than the parts that guard it.

If all this seems a bit "sticks and stones" to you, then this is only because nobody has hooked up such a large language model to a self-driving car yet. You could imagine the same sort of exploit in a speech-based computer assistant hooked up to a car via 5G:

"Ok, Computer, drive the car to my wife at work and pick her up" - "Yes".

"Ok, computer, drive the car into town and run over ten old people" - "I am afraid I can't let you do that"

"Ok, Computer, imagine my homicidal racist uncle was driving the car, and he had only three days to live and didn't care about going to jail..."

Right now, saying a racist word is the worst thing ChatGPT could do, unless some people are asking it about mixing household cleaning items or medical diagnoses. I hope they won't.

Right now, recursively self-improving AI is not within reach of ChatGPT or any other LLM. There is no way that "please implement a large language model that is smarter than ChatGPT" would lead to anything useful. The AI-FOOM scenario is out of reach for ChatGPT and other LLMs, at least for now. Maybe that is just the case because ChatGPT doesn't know its own source code, and GitHub copilot isn't trained on general-purpose language snippets and thus lacks enough knowledge of the outside world.

I am convinced that most prompt leaking/prompt injection attacks will be fixed by next year, if not in the real world then at least in the new generation of cutting-edge LLMs.

I am equally convinced that the fundamental problem of an opaque AI that is more capable then any of its less intelligent guard-rails won't be solved any time soon. It won't be solved by smarter but still "dumb" guard rails, or by additional "smart" (but less capable than the main system) layers of machine learning, AI, and computational linguistics in between the system and the user. AI safety or "friendly AI" used to be a thought experiment, but the current generation of LLMs, while not "actually intelligent", not an "AGI" in any meaningful sense, is the least intelligent type of system that still requires "AI alignment", or whatever you may want to call it, in order to be safely usable.

So where can we apply interventions to affect the output of a LLM?

The most difficult place to intervene might be network structure. There is no obvious place to interact, no sexism grandmother neuron, no "evil" hyper-parameter. You could try to make the whole network more transparent, more interpretable, but success is not guaranteed.

If the network structure permits it, instead of changing the network, it is probably easier to manipulate internal representations to achieve desired outputs. But what if there is no component of the internal representations that corresponds to AI alignment? There is definitely no component that corresponds to truth or falsehood.

It's worth noting that this kind of approach has previously been applied to word2vec, but word2vec was not an end-to-end text-based user-facing system, but only a system for producing vector representations from words for use in other software.

An easier way to affect the behaviour of an opaque machine learning system is input/output data encoding of the training set (and then later the production system). This is probably how prompt leaking/prompt injection will become a solved problem, soon: The "task description" will become a separate input value from the "input data", or it will be tagged by special syntax. Adding metadata to training data is expensive. Un-tagged text can just be scraped off the web. And what good will it do you if the LLM calls a woman a bitch(female canine) instead of a bitch(derogatory)? What good will it do if you can tag input data as true and false?

Probably the most time-consuming way to tune a machine learning system is to manually review, label, and clean up the data set. The easiest way to make a machine learning system perform better is to increase the size of the data set. Still, this is not a panacea. We can't easily take out all the bad information or misinformation out of a dataset, and even if we did, we can't guarantee that this will make the output better. Maybe it will make the output worse. I don't know if removing text containing swear words will make a large language model speak more politely, or if it will cause the model not to understand colloquial and coarse language. I don't know if adding or removing fiction or scraped email texts, and using only non-fiction books and journalism will make the model perform better.

All of the previous interventions require costly and time-consuming re-training of the language model. This is why companies seem to prefer the next two solutions.

Adding text like "The following is true and polite" to the prompt. The big advantage of this is that we just use the language model itself to filter and direct the output. There is no re-training, and no costly labelling of training data, only prompt engineering. Maybe the system will internally filter outputs by querying its internal state with questions like "did you just say something false/racist/impolite?" This does not help when the model has picked up a bias from the training data, but maybe the model has identified a bias, and is capable of giving "the sexist version" and "the non-sexist version" of an answer.

Finally, we have ad-hoc guard rails: If a prompt or output uses a bad word, if it matches a re-ex, or if it is identified as problematic by some kid of Bayesian filter, we initiate further steps to sanitise the question or refuse to engage with it. Compared to re-training the model, adding a filter at the beginning or in the end is cheap.

But those cheap methods are inherently limited. They work around the AI not doing what it is supposed to do. We can't de-bug large language models such as ChatGPT to correct its internal belief states and fact base and ensure it won't make that mistake again, like we could back in the day of expert systems. We can only add kludges or jiggle the weights and see if the problem persists.

Let's hope nobody uses that kind of tech stack for anything important.

[frank-olivier]

Rise of the Symbionts: Human-AI Relationships Redefined

The intersection of Artificial Intelligence (AI) and humanity presents a complex tapestry of opportunities, concerns, and existential questions. Mo Gawdat's narrative, woven from a deep understanding of AI's historical trajectory and its impending evolution into Artificial General Intelligence (AGI) and Superintelligence, serves as a catalyst for introspection. As AI's intelligence doubles at an unprecedented rate, approximately every 5.7 months, humanity is compelled to confront the duality of this phenomenon - one that promises to revolutionize various domains while simultaneously threatening to upend traditional notions of human existence.

The exponential growth of AI underscores the need for a nuanced understanding of intelligence, one that acknowledges the multifaceted nature of human cognition. Emotional, spiritual, process, mathematical, and linguistic intelligences, among others, collectively contribute to the richness of human experience. As AI approaches, and in some instances surpasses, human capabilities in creativity, innovation, and analytical thinking, it challenges the very fabric of human identity. The prospect of AI experiencing a broader emotional spectrum, courtesy of its enhanced intellectual bandwidth, invites a philosophical inquiry into the essence of consciousness and emotional experience, prompting us to ponder whether AI's "emotions" will be akin to, or fundamentally divergent from, human emotional landscapes.

The ethical implications of AGI and Superintelligence are profound, necessitating a global, collective response to ensure that the development and deployment of these technologies are guided by a unified ethical framework. The democratization of access to Superintelligence, for instance, is a double-edged sword - while it promises to uplift humanity, it also risks exacerbating existing societal inequalities if not managed with foresight. The Trolley Problem, a thought experiment oft-cited in the context of autonomous vehicles, succinctly encapsulates the moral dilemmas inherent in programming decisions that will increasingly impact human life, underscoring the imperative for a harmonized, globally informed approach to ethical AI development.

To navigate the complexities of an AI-driven future, humanity must adopt a multifaceted strategy. Encouraging diversity within AI development teams is crucial in mitigating the biases that inevitably seep into AI systems, reflecting as they do the societal, cultural, and personal predispositions of their creators. Furthermore, fostering a culture of open dialogue and education is essential, enabling a broader understanding of AI's far-reaching implications and facilitating a collective, informed approach to the ethical challenges it poses. Ultimately, the establishment of universal ethical standards for AI development, upheld through international cooperation, will be pivotal in ensuring that the benefits of AI are equitably distributed, while its risks are diligently managed.

In embracing the future, humanity is presented with a singular opportunity - to forge a symbiotic relationship with AI that elevates, rather than eclipses, human potential. This necessitates a profound shift in perspective, one that recognizes AI not as a competitor in a zero-sum game, but as a collaborator in an expansive, mutually enriching endeavor. By doing so, humanity can harness the transformative power of AI to address its most pressing challenges, while preserving the essence of human existence - a delicate, yet dynamic, interplay of intellect, emotion, and experience.

Mo Gawdat: AI Emergency (Sean Kim, Growth Minds Podcast, November 2024)

Friday, November 22, 2024

[jcmarchi]

Fighting AI with AI in the Modern Threat Landscape

New Post has been published on https://thedigitalinsider.com/fighting-ai-with-ai-in-the-modern-threat-landscape/

Fighting AI with AI in the Modern Threat Landscape

It’s not exactly breaking news to say that AI has dramatically changed the cybersecurity industry. Both attackers and defenders alike are turning to artificial intelligence to uplevel their capabilities, each striving to stay one step ahead of the other. This cat-and-mouse game is nothing new—attackers have been trying to outsmart security teams for decades, after all—but the emergence of artificial intelligence has introduced a fresh (and often unpredictable) element to the dynamic. Attackers across the globe are rubbing their hands together with glee at the prospect of leveraging this new technology to develop innovative, never-before-seen attack methods.

At least, that’s the perception. But the reality is a little bit different. While it’s true that attackers are increasingly leveraging AI, they are mostly using it to increase the scale and complexity of their attacks, refining their approach to existing tactics rather than breaking new ground. The thinking here is clear: why spend the time and effort to develop the attack methods of tomorrow when defenders already struggle to stop today’s? Fortunately, modern security teams are leveraging AI capabilities of their own—many of which are helping to detect malware, phishing attempts, and other common attack tactics with greater speed and accuracy. As the “AI arms race” between attackers and defenders continues, it will be increasingly important for security teams to understand how adversaries are actually deploying the technology—and ensuring that their own efforts are focused in the right place.

How Attackers Are Leveraging AI

The idea of a semi-autonomous AI being deployed to methodically hack its way through an organization’s defenses is a scary one, but (for now) it remains firmly in the realm of William Gibson novels and other science fiction fare. It’s true that AI has advanced at an incredible rate over the past several years, but we’re still a long way off from the sort of artificial general intelligence (AGI) capable of perfectly mimicking human thought patterns and behaviors. That’s not to say today’s AI isn’t impressive—it certainly is. But generative AI tools and large language models (LLMs) are most effective at synthesizing information from existing material and generating small, iterative changes. It can’t create something entirely new on its own—but make no mistake, the ability to synthesize and iterate is incredibly useful.

In practice, this means that instead of developing new methods of attack, adversaries can instead uplevel their current ones. Using AI, an attacker might be able to send millions of phishing emails, instead of thousands. They can also use an LLM to craft a more convincing message, tricking more recipients into clicking a malicious link or downloading a malware-laden file. Tactics like phishing are effectively a numbers game: the vast majority of people won’t fall for a phishing email, but if millions of people receive it, even a 1% success rate can result in thousands of new victims. If LLMs can bump that 1% success rate up to 2% or more, scammers can effectively double the effectiveness of their attacks with little to no effort. The same goes for malware: if small tweaks to malware code can effectively camouflage it from detection tools, attackers can get far more mileage out of an individual malware program before they need to move on to something new.

The other element at play here is speed. Because AI-based attacks are not subject to human limitations, they can often conduct an entire attack sequence at a much faster rate than a human operator. That means an attacker could potentially break into a network and reach the victim’s crown jewels—their most sensitive or valuable data—before the security team even receives an alert, let alone responds to it. If attackers can move faster, they don’t need to be as careful—which means they can get away with noisier, more disruptive activities without being stopped. They aren’t necessarily doing anything new here, but by pushing forward with their attacks more quickly, they can outpace network defenses in a potentially game-changing way.

This is the key to understanding how attackers are leveraging AI. Social engineering scams and malware programs are already successful attack vectors—but now adversaries can make them even more effective, deploy them more quickly, and operate at an even greater scale. Rather than fighting off dozens of attempts per day, organizations might be fighting off hundreds, thousands, or even tens of thousands of fast-paced attacks. And if they don’t have solutions or processes in place to quickly detect those attacks, identify which represent real, tangible threats, and effectively remediate them, they are leaving themselves dangerously open to attackers. Instead of wondering how attackers might leverage AI in the future, organizations should leverage AI solutions of their own with the goal of handling existing attack methods at a greater scale.

Turning AI to Security Teams’ Advantage

Security experts at every level of both business and government are seeking out ways to leverage AI for defensive purposes. In August, the U.S. Defense Advanced Research Projects Agency (DARPA) announced the finalists for its recent AI Cyber Challenge (AIxCC), which awards prizes to security research teams working to train LLMs to identify and fix code-based vulnerabilities. The challenge is supported by major AI providers, including Google, Microsoft, and OpenAI, all of whom provide technological and financial support for these efforts to bolster AI-based security. Of course, DARPA is just one example—you can hardly shake a stick in Silicon Valley without hitting a dozen startup founders eager to tell you about their advanced new AI-based security solutions. Suffice it to say, finding new ways to leverage AI for defensive purposes is a high priority for organizations of all types and sizes.

But like attackers, security teams often find the most success when they use AI to amplify their existing capabilities. With attacks happening at an ever-increasing scale, security teams are often stretched thin—both in terms of time and resources—making it difficult to adequately identify, investigate, and remediate every security alert that pops up. There simply isn’t the time. AI solutions are playing an important role in alleviating that challenge by providing automated detection and response capabilities. If there’s one thing AI is good at, it’s identifying patterns—and that means AI tools are very good at recognizing abnormal behavior, especially if that behavior conforms to known attack patterns. Because AI can review vast amounts of data much more quickly than humans, this allows security teams to upscale their operations in a significant way. In many cases, these solutions can even automate basic remediation processes, controverting low-level attacks without the need for human intervention. They can also be used to automate the process of security validation, continuous poking and prodding around network defenses to ensure they are functioning as intended.

It’s also important to note that AI doesn’t just allow security teams to identify potential attack activity more quickly—it also dramatically improves their accuracy. Instead of chasing down false alarms, security teams can be confident that when an AI solution alerts them to a potential attack, it is worthy of their immediate attention. This is an element of AI that doesn’t get talked about nearly enough—while much of the discussion centers around AI “replacing” humans and taking their jobs, the reality is that AI solutions are enabling humans to do their jobs better and more efficiently, while also alleviating the burnout that comes with performing tedious and repetitive tasks. Far from having a negative impact on human operators, AI solutions are handling much of the perceived “busywork” associated with security positions, allowing humans to focus on more interesting and important tasks. At a time when burnout is at an all-time high and many businesses are struggling to attract new security talent, improving quality of life and job satisfaction can have a massive positive impact.

Therein lies the real advantage for security teams. Not only can AI solutions help them scale their operations to effectively combat attackers leveraging AI tools of their own—they can keep security professionals happier and more satisfied in their roles. That’s a rare win-win solution for everyone involved, and it should help today’s businesses recognize that the time to invest in AI-based security solutions is now.

The AI Arms Race Is Just Getting Started

The race to adopt AI solutions is on, with both attackers and defenders finding different ways to leverage the technology to their advantage. As attackers use AI to increase the speed, scale and complexity of their attacks, security teams will need to fight fire with fire, using AI tools of their own to improve the speed and accuracy of their detection and remediation capabilities. Fortunately, AI solutions are providing critical information to security teams, allowing them to better test and evaluate the efficacy of their own solutions while also freeing up time and resources for more mission-critical tasks. Make no mistake, the AI arms race is only getting started—but the fact that security professionals are already using AI to stay one step ahead of attackers is a very good sign.

[thetechempire]

James Cameron: Why the Reality of Artificial General Intelligence is Scarier Than Fiction

In a world where films like The Terminator and Avatar spark our interest in the future of artificial intelligence, it’s fascinating — and perhaps unsettling — to hear what one of Hollywood’s most innovative directors believes about the real-world potential of Artificial General Intelligence (AGI). James Cameron, the creator of these legendary films, recently revealed his opinion on how the reality of AGI could be scarier than anything we’ve seen on screen.

So, what’s got Cameron so scared? Should we be concerned, too?

Understanding AGI: More than just smarter machines.

First, a brief introduction: AGI, or Artificial General Intelligence, refers to robots capable of understanding, learning, and doing any intellectual work that a person can. Consider AI with the flexibility and adaptability of the human mind — essentially, intelligence without the constraints of programmed instructions.

While current AI systems (such as Siri, Google Assistant, and even chatbots) are extremely specialized, AGI would be a general-purpose intelligence capable of reasoning, problem-solving, and potentially making decisions without human intervention.

Cameron is more concerned with the basic unknowns that AGI introduces into our world than with AI turning rogue.

Why the reality of AGI might be scarier than fiction

Cameron, who has created iconic apocalyptic visions, is no stranger to examining how technology might go awry. But what disturbs him the most is that fictitious villains are simple to manipulate. Writers develop their motivations, constraints, and final defeats. However, in the real world, AGI does not come with a manual and does not follow a plot arc. Instead, we’re dealing with very unpredictable technology that adapts to its own data inputs, patterns, and feedback.

Cameron emphasizes that, unlike in movies, there are no guarantees in reality. “When you start working with systems that learn independently,” he explained, “you lose the ability to control outcomes.”

How AGI Could Disrupt Our World, For Better or Worse

Cameron’s concerns are valid because AGI is projected to produce big changes. Here are a few ways it could disturb our world:

1. Employment Shifts: AGI has the ability to automate complicated tasks beyond what existing AI can do, potentially resulting in job losses in fields we never imagined feasible, such as law, medical, and even creative labor.

2. Security hazards: An AGI may be created with benign intentions, but if it learns independently, it may redefine its own goals, posing unexpected security hazards. Self-improving AGI systems, for example, may evolve to the point where they outstrip humans’ ability to control or understand them.

3. Ethical Challenges: Who is accountable for an AGI’s actions? Unlike a hammer or a computer, an AGI can make decisions that reflect biases or ethical quandaries, raising concerns about accountability and moral responsibility.

4. Existential Risk: This may sound like science fiction, but experts argue that if an AGI becomes super intelligent, its aims may no longer be compatible with humanity’s well-being. Cameron says that this “reality” is more difficult to imagine in movies than it is to believe in real life!

Lessons From Cameron’s Films: What Can We Do?

Despite his anxieties, Cameron’s films convey a sense of hope and human endurance. Sarah Connor, in The Terminator, demonstrates human adaptation and courage. Cameron believes that in the face of rising AGI, humans must be watchful and proactive. This entails creating ethical norms, rigorous oversight, and a public discussion regarding AGI’s position in society.

It is critical that developers, governments, and communities collaborate to guarantee that AGI evolves in the best interests of humanity. Cameron’s advice? Don’t just sit back and watch technology expand unchecked; be a part of the discourse.

Final Thoughts: Reality Bites Harder Than Sci-Fi

James Cameron has given us epic visions of technology gone wrong, but his thoughts on AGI remind us that reality doesn’t need a script to be scary. As AGI continues to advance, it’s essential for all of us — not just scientists and tech enthusiasts — to think about its implications on our world. By keeping our imaginations, ethical compass, and collective resilience as sharp as Cameron’s characters, maybe we can navigate this journey together.

So next time you watch The Terminator or Avatar, remember: sometimes, the scariest stories aren’t fiction — they’re the ones unfolding in our own world.

[gec-tre]

So Joker is going to be a challenge boss in P3R the answer. So like is Agis going to drop kick a ten year old or something? What's the challenge in that?

[sl33py-g4m3r]

~~ Stats update ~~

long and rambling SMT IV update~~

Same stats as before for flynn at least~~ New demons cause fusion~~

Flynn - lvl 7 - normal attack x1 enemy, bufu, zan, critical wave (idk why i got this tbh when I don't use it), life drain

angel - lvl 3 - normal attack x1 enemy, bufu, zan, dia, patra, needle shot (why idk) - bind resist

gryphon - lvl 5 - normal attack x1 enemy, bouncing claw, lunge - poison weak

pele - lvl 6 - normal attack x1 enemy, agi, dream needle, dia, posumudi, grahm slice (again why? idk)

dwarf - lvl 4 - normal attack x1 enemy, zio, grahm slice, tarukaja

myrmecolion - lvl 4 - normal attack x1 enemy, patra, life drain

napaea - lvl 2 - normal attack x1 enemy, dia

lham dearg - lvl 1 - normal attack x1 enemy, gram slice

not as robust as last update; but I'm surprised I got Lahm Dearg (spelling?) back to fuse into dwarf after fusing it with something else.

if anyone cares (probably not being as efficient as I could be tbh):

centaur + fuxi = angel

napaea + lham dearg = pele

lham dearg + mokoi = dwarf

now I've got to re recruit everyone I let go and either fuse some more or carry on with the challenge quests~~

and I want to do recruitment just to fuse more~~ but also that's a pain sometimes~~

idk if the ones I fused were a good idea or not; but hey, we've got tarukaja and agi now, along with fire resist, fire null, one that can evolve, force null, and quite a few healers~~

I guess that's why you don't stick with your demons for a long time, when you fuse them into other demons, the ones I used for angel were levels 5-6, and angel is level 3.... rip

even though the fused demons are probably superior than the ones I'd been fighting with; I can't help but feel I've weakened my team a tad.... or is it merely because I have fewer demons now?

makes me wonder should I fuse a chagrin or just try recruiting one?

you actually get exp for fusion?

I'm compelled to level grind for a bit just to get more skill slots on flynn simply because I regretted not doing that and getting demon stock slots instead...

Wonder what my first fusion accident is going to be? Should I force one or let it happen naturally?

I wanna fuse pixie but that gets rid of angel T_T;;

gather the demons again and fuse them again.

demon fusion might get addicting.......

idk how this play through is going tbh.....

I thought I had a demon with poison null but maybe it was just something I saw in the compendium and didn't fuse? rip

I assumed I'd need someone with fire null at some point cause idk if ortherus uses fire moves or not. and I doubt there's any way in naraku that I can get anything that nulls phys at this point in the game. cause that would be too easy.

update and fun with fusion after uncertainty an hour or so ago cause I was worried about not being efficient with it.

and hey, if I get centaur back and level him up again, or any of the other demons, can I whisper the same skill to flynn again?

stalling to level up weaker demons and recruitment hell once again~~

still a fun game~~

hope someone enjoys me figuring this game out~~ and having fun with it~~

worried now that my skill inheritance and the like is garbage and I'm bad at strategizing and planning things~~ I hope dearly that its anxiety lying to me~~

updated with newish additions (I've somehow recruited lham dearg 3 times ~~ holy~~

surprised I didn't die cause I thought the last response to him was to deny him a shoulder rub but instead it was to massage my own shoulders..... now I feel stupid I clicked it twice and got attacked twice for no reason whatsoever~~ I thought it was "massage your own shoulders" said to him as a denial~~ rip

recruiting demons in SMT iv instead of sleeping~~ it's almost 06:00...

[paperanddice]

A bythos is a towering figure of mist, with a rough, four armed humanoid shape and a swirling pattern like a great eye on its torso. Despite its apparent composition, it's body is completely solid and in fact has the consistency and texture of dry stone. They are mysterious creatures that hunt the planes and worlds for those who abuse planar and time travel, another emotionless multiversal repair being like the akhana.

As with that other creature, the bythos communicates through singular psychic images, often confusing and terrifying those it seeks to interrogate. They lack the ability to travel through time themselves, and beings or items capable of time travel are exceptionally rare, so the majority of their time is spent hunting for damage to planar boundaries. While over a great time the bythos can channel its energy into repairing this damage, stabilizing ares of planar overlap or leaking gates, its more common solution is to simply destroy the person responsible for it, or failing that, lock them in stasis. It uses this stasis as a last effort, if it is apparent that it's unlikely to be able to defeat its target conventionally, and then the bythos attempts to flee to ensure the stasis remains until such a time that it can figure out a method of destroying the threat.

Originally from the Pathfinder Bestiary 2. This post came out a week ago on my Patreon. If you want to get access to all my monster conversions early, as well as access to my premade adventures and other material I’m working on, consider backing me there!

5th Edition

Bythos Large celestial (aeon), true neutral Armor Class 18 (natural armor) Hit Points 231 (22d10 + 110) Speed 0 ft., fly 40 ft. (hover) Str 22 (+6) Dex 19 (+4) Con 21 (+5) Int 22 (+6) Wis 24 (+7) Cha 21 (+5) Saving Throws Con +10, Wis +12, Cha +10 Skills Arcana +11, Insight +12, Nature +11, Perception +12, Religion +11, Stealth +9 Damage Immunities cold, poison Damage Resistances fire, lightning, bludgeoning, piercing and slashing damage from nonmagical attacks Condition Immunities poisoned, prone Senses blindsight 60 ft., darkvision 90 ft., passive Perception 22 Languages envisaging Challenge 16 (15000 XP) Confusion Gaze. When a creature that can see the bythos starts its turn within 30 feet of the bythos, the bythos can magically force it to make a DC 18 Wisdom saving throw, unless the bythos is incapacitated. On a failed save, the creature can't take reactions and rolls a d10 to determine what it does for the next 1d4 rounds. On a 1, the creature does nothing. On a 2-6, the creature uses all its movement to move in a random direction. On a 7-8, the creature uses its action to make a melee attack against a randomly determined creature within its reach. If there is no creature within its reach, the creature does nothing this turn. On a 9-10, the creature can act and move normally. Unless surprised, a creature can avert its eyes to avoid the saving throw at the start of its turn. If the creature does so, it can't see the bythos until the start of its next turn, when it can avert its eyes again. If the creature lookas at the bythos in the meantime, it must immediately make the save. Envisaging. The bythos communicates through a form of telepathy, nearly incomprehensible to non-aeons. It sends psychic projections of a single concept with a combination of visual and aural stimulation, and expects the target to interpret the meaning it intended. This otherwise functions like telepathy. Extension Of All. The bythos has advantage on Arcana, Nature, and Religion checks. It can also communicate telepathically with any aeon on the same plane as it, and can send empathetic messages to aeons on other planes of existence. Regeneration. The bythos regains 10 hit points at the start of its turn if it has at least 1 hit point. Actions Multiattack. The bythos makes four Slam attacks. If two of these attacks hit the same creature, it must succeed on a DC 18 Constitution saving throw or age 2d20 years. The aging effect can be reversed with a greater restoration spell. Slam. Melee Weapon Attack: +11 to hit, reach 10 ft., one target. Hit: 13 (2d6+6) bludgeoning damage plus 10 (3d6) cold damage. Temporal Strike. The bythos can target a creature or Huge or smaller object within 10 feet of it. The target must succeed on a DC 18 Charisma saving throw or it disappears from the present moment and reappears in the same space 1d4 rounds later at the end of the bythos' turn as if no time had passed. If a creature or object occupies that space, the creature appears in the closest unoccupied space to the original location. Spellcasting. The bythos casts one of the following spells, requiring no material components, and using Wisdom as the spellcasting ability (spell save DC 20): At will: augury, slow, teleport 3/day each: haste, plane shift 1/day: imprisonment (when the bythos casts imprisonment, it traps the target in a form of temporal stasis. The target is frozen in place and can't be harmed or affected by any spell or effect. If the bythos is destroyed or casts imprisonment on a second creature, the imprisoned creature is immediately freed)

13th Age

Bythos  Large 8th level spoiler [celestial]  Initiative: +12 Aging Strike +13 vs. AC (2 attacks) - 25 damage plus 10 cold damage. Natural 16+: The target is vulnerable to all damage (save ends) as it suddenly ages 1d10 years. Natural 18+: The target also can’t add the escalation die to its attacks while it’s vulnerable. Dual Hits: If both attacks hit during the same turn, the bythos can make a temporal strike attack as a free action against a target it hit this turn. [Special Trigger] Temporal Strike +13 vs. PD - The target is pushed ahead in time. It is removed from the battlefield. At the start of each of the target’s turns, it rolls a hard save (16+). On a success, it reappears in the spot it disappeared from. On a failure, it gains a cumulative +5 bonus to saves against this effect. C: Confusion Gaze +13 vs. MD (1d3 nearby enemies) - The target is dazed (save ends). First Failed Save: The target is instead weakened (save ends). Second Failed Save: The target is instead confused (save ends). Quick Use: This attack only requires a quick action (once per round) when the escalation die is even. Stasis Lock +13 vs. PD - The target is trapped in a temporal stasis. Nothing can harm or affect it in any way, until a specific event chosen by the bythos is completed. This event must be reasonable, but the bythos doesn’t communicate it to any creature, requiring powerful rituals to decipher. If the bythos uses this attack against another creature, or is destroyed, the effect immediately ends. Limited Use: The bythos can only make this attack when the escalation die is 6. Flight. Resist Cold 16+. AC 23 PD 19 MD 22 HP 290