Hydro

extrapolate trends from previous data.
okay that's obnoxious as hell for me to say. i like the boom redesigns. they're wrong in a fun way

imagine that he sounds like an uncannily human-like and natural version of microsoft sam

Verbatim poses cutely with his arms and hands thrown over his knees, as he stands with his back leaning backwards a bit. He's looking at you with his dull, cute, lifeless eye. He's also depicted in doodles, posing with his hands on his hips, doing finger guns, and gesturing at the viewer with two fingers. Text reads: VERBATIM'S HOT TIP OF THE DAY: [[while Verbatim himself says,][ "Be a little fruity with it."

Verbatim stands neutrally, looking off at something in his default A-pose. "I see. "Is that so?" he asks. "Qué chévere." he says, suddenly dancing, whipping his hips to the side and twirling his arms about. Then, he's back to his original A-pose with his hands at his sides, motionless.

Verbatim is drawn teleporting across a grid-floor to random panels, and is drawn again hovering about, floating lazily with his arms hanging slightly in front of his legs. "STOP TELEPORTING AROUND THE FIELD I CAN'T HIT YOU" shouts his opponent from off-screen. "Unfortunate." Verbatim replies. For the hovering drawing, text also reads: *Sonic 3K Egg Mobile SFX but pitched way way down*

Verbatim I love you I love you Verbatim I love you (2020 - 2025)

Verbatim stands with his hands held over his thighs, politely. The picture is titled, "VERBATIM'S HOT TIP OF THE DAY:" Verbatim then says, simply, "If a VPN is 'free'... "...Then it isn't."

Verbatim makes a heart with his hands, including a literal heart icon hovering in between his fingers. VERBATIM'S HOT TIP OF THE DAY: "I have concluded with Science that the value of life is inherent, irrevocable, and self-evident. "Never forget to acknowledge your own value. 'The evidence to support my claim is that I love you."

Titled "VERBATIM'S HOT TIP OF THE DAY:" Verbatim is on the ground, lying down. He's looking at you, saying, "Set your passwords with uncommon word combinations, and change your passwords at frequent but inconsistent intervals. "Unpredictability is security's ultimate asset." faded text near him reads, indicating a whisper from him, "i am positioned like this because i am being sneaky and that is unpredictable"

Verbatim is in a void, and says nothing. He tilts his head down slightly, speaking, "Sometimes, I fear that nothing I say is quite novel enough or worth mentioning." He pauses, the viewer's perspective pulled back a tad. In the last panel, he's seen from much further away, the expanse of the white void overwhelming him. "I feel worthless." he says.

helpful boy (2019 - 2020)

Titled: "VERBATIM'S HOT TIP OF THE wait" Whereas Verbatim, with noticeably more shiny and cute eyes than usual, looks up and asks, "am i cute y/n" He looks a bit more furbait than usual here, with shapely hips and thighs. Not like, Twitter-Furry proportions, but there's definitely something there, if only in the gesture if nothing else. He's clutching his head with his hands, covering up his face slightly as he peeks just over his snout. I think he looks cute, personally

Verbatim hovers, gesturing with one hand like the pointing up emoji, but otherwise doing nothing. He's got a cool wavy gesture to his body here.

Verbatim, doing various dance poses, referenced from Kakyoin from JJBA: Heritage for the Future. There are five poses total, involving Verbatim stanced up, and gesturing his arms and hips at various, odd angles. You'll have to take my word that he looks really swag

Dr. Thomas Livingston, Verbatim's creator and friend, attempts to free Verbatim from his current predicament; Verbatim's snout is lodged firmly into a cup. Thomas is a white wolf with a labcoat, black goggles, spiky fur, and wavy hair. He's on his knees, attempting to remove the cup from Verbatim's face, while Verbatim is slouched over with his arms hanging, looking down at the cup on his nose. "why are you doing this "why are you like this "is there something i can do to help" Thomas asks. "cup" Verbatim replies. Thomas has a cute fur-mustache, just thought i should mention

dog made of shapes. there's something wrong about him (2021 - 2022)

Verbatim, a corporeal anthropomorphic antivirus, stands cutely, pointing at the screen. He says, with clickable dialogue windows, "It is now safe to turn off the computer. Goodnight. Sleep tight. "Please send an error-report in the event of any bed-bugs." Verbatim is a bit tricky to describe. He's got an abstract shape for a face, but he looks vaguely like a dog with large, porygon-like eyeballs, one on both sides of his face, with a big dumb, featureless snout in the middle. He's got two triangles for "cheek tufts" and two vaguely ear-shaped wedges sticking out of his head. His head floats above his torso without a neck. Verbatim appears to be wearing a sort of bodysuit and a jacket, but it's all seamless enough that's it's kinda just his body. Like, you can look in the neck-hole of his jacket and there's nothing in there at all, implying he's hollow. But he's not hollow. He's full of love. Also he has digitigrade feet with three cute lil toesies on each foot. They're drawn really simple though, which likely disappoints anyone actually Into that sort've thing. My apologies

Verbatim stands with a battle-ready stance, body jagged and angular. One of his arms is converted into a gatling gun. He looks somewhat aimlessly with his one visible, large eye. "Please update to the latest version of your OS. Remaining on unsupported systems may result in security breaches. "It is unwise to hold on to the past, in the event it is exploited against you by malicious actors." There's cool black stripes on the side of his legs and arms. You can see more clearly that his jacket is a cropped waist-jacket vest-thing of sorts, with an X insignia over the center. Underneath is just more bodysuit. He's got a funny pointed snout, more visible here since his head is turned in profile-view.

Verbatim hovers elegantly over a panel on the floor, interfacing with a weapons menu with one hand with the other as his gun-arm, at the ready. Icons above Verbatim indicate his Battle Programs, various weapons he has loaded in cache. Text above them read, "Cstmzing..." while the icons depict the gatling gun, a sword, a wide sword, and a long...er sword. Verbatim looks neutrally at the viewer. A dialogue box for Verbatim reads, "Combat Routine: "Compiled. >Run." He's also doodled holding an Actual Short-Barrel Revolver, furrowing his brows sharply, with text reading, "Custom Gauge FULL. Patience Gauge EMPTY." In the top right, he's doodled again, this time using an attack where he uses some Telekinesis type nonsense to lift and shoot off the ground in front of him, like an earth bender. Everything here is reminiscent of Mega Man Battle Network because Verbatim was literally a fan-net navi I made when my age was in the single digits. He's just an oc now though

Verbatim, the world's strangest antivirus dog, beloved (2021-2022)

Liking modern sonic for all the same reasons people would critique his design for. Hm? Classic Sonic is an objectively perfect design? Modern Sonic has a lot of weird choices that clash with the original design philosophy? I agree. That's why it's good. Let's make him worse

Liking Lucario but specifically for all the reasons his design is fucked up. A liveleak video of a pokemon design. I love Lucario's head-tentacles. It's odd Pants situation. Stupid spikes. Its Mouth With A Nose On Top, Like A Man. I love when a design is just Off and gets more Off as you look at it

it's true

Monochrome drawing of Amy Rose in a somewhat ostentatious outfit reminiscent of a magical girl, with a dress adorned with roses and frills beneath. Text in all caps to the top left and bottom right of the image, respectively, reads: I MICROWAVE PLASTIC CUPS FOR FIVE MINUTES AND DRINK WHATEVER'S LEFT!!!!!!!

Spiking neural networks like the human nervous system can react dynamically to stimuli when embodied, while LLM-powered robots need non-LLM modules to do so, and require constant prompting. A real human-equivalent AI could lilely learn to drive a car, while an LLM would have multiple bottlenecks.

Correct! Something along those lines I feel. And the only "telling" you'd need there is people remarking "Yeah, GPTGill tends to just. Forget things over and over."
Another option I've considered is an LLM one having much poorer reaction time overall;

Spiking Neural Networks are hard to research for a variety of reasons, mainly hardware, but on the right hardware they're the best option for an AI that stays on all the time and dynamically alters its neurons to new stimuli. The hardware has only recently kinda-sorta caught up, so research slow.

You're good! And you about covered the core difference. Current AI requires a human "in the loop" to update the model's "weights" to train it and retain new information permanently. On so many levels this is kinda the crux of the issue, as LLMs own learning is not self-motivated and directed

This would be only for an introductory scene. From there it'd basically be like showing in a story the difference between a person that fakes emotions to someone with real ones despite both acting 90% identically to most people. It's not impossible at all, but far easier said than done.

My approach would be to take a POV of a robot's mechanic, and have the mechanic do diagnostic work and types of tests on a neuromorphic robot and LLM-based robot respectively, using the frank, straightforward language car mechanics usually use. It'd be "showing and telling", but balanced.

Any approach to either I guarantee can be ravenously picked apart with critique on a level more absurd and difficult to argue than even the murder mystery example, because sci-fi with any "hardness" as a hook for the genre requires at least a bit of explanation, because you can't just Show Research.

For instance: a "real" artificial intelligence's emotional behavior can be dismissed as generated simulation, not getting how those emotions tangibly affect its cognition. Another audience member might just say real LLMs ARE sentient, somehow. There's many variables in communicating the difference

A murder mystery for example often requires a certain amount of "tell" but good writers can make it smooth enough with "showing" that no one notices. A story about War Is Bad meanwhile often needs minimal telling unless you want to get into the grittier factual geopolitics, speculative history-style

I'm afraid some degree of "tell" is required for this sort've thing. I have faith in my ability to balance show and tell, it'd just be a lot of work. "Showing" works within audience's understanding and assumptions, while telling is required for any nuance that cannot be easily shown:

Chew you havisfaction a singlelicious satisfact to snack that up

When I'm going on a tirade about robots vs modern ai assume I'm rendered low poly like a Deus Ex character and have the UNATCO theme playing in the background

It does! Imo AI having actual life experience is really important, especially if it can derive deep meaning from it despite not literally having read every book and seen every video. I believe the ability to take great meaning from minimal experience is the mark of ""true"" intelligence and agency

They can run in real time, and each computer-neuron doesn't need to be turned on at all times. Your brain lights up "asynchronously" without the same strict "clock speed" traditional computers have, and a neuromorphic array works similarly, at similarly low power. They're real, but too small atm

You were closer with the "machine adapting our nervous system" thing. This can be done digitally at very, very small scales, but a "neuromorphic array" is the ideal for human-level cognition. Neuromorphic arrays are just chips with millions of simple computers to do the work of real neurons

I forgot renpy is for visual novels and went "damn i don't know if I can code a sentient AI in python, I know some Java and C++ but that's about it"

I had this idea of a neuromorphic robot oc that has human-like cognition, but the poor bastard was trained and "raised" in its server rack by LLMs, so it talks and acts like one. Soon learns not to by living with humans via a remotely controlled body. Angst over not being able to leave the rack, etc

Likewise, all the actual philosophical questions robot media is meant to invoke gets lost if our frame of reference is LLMs. We've been mislead towards what real AI is going to look like, and we're gonna be horribly unprepared for actually dealing with it, culturally and practically.

Normal modern brain scans aren't that high in resolution compared to the actual fidelity billions of neurons have, so a virtual clone has to "guess" someone's neural structure and can be slightly "off" or different from the original. Easily justifies "clones gets own identity"-type stories.

In other words, "real" AI that I think may actually happen would look like a moderate server rack running a slow but stable brain simulation, whose only rewarded goal is to emulate and "upscale" brain activity from low-resolution scans. This, naturally, has extremely juicy storytelling potential.