With ChatGPT, Game NPCs Get A Lot More Interesting

Not only is AI-driven natural language processing a thing now, but you can even select from a number of different offerings, each optimized for different tasks. It took very little time for [Bloc] to mod a computer game to allow the player to converse naturally with non-player characters (NPCs) by hooking it into ChatGPT, a large language model AI optimized for conversational communication.

If you can look past the painfully-long loading times, even buying grain (7:36) gains a new layer of interactivity.

[Bloc] modified the game Mount & Blade II: Bannerlord to reject traditional dialogue trees and instead accept free-form text inputs, using ChatGPT on the back end to create more natural dialogue interactions with NPCs. This is a refinement of an earlier mod [Bloc] made and shared, so what you see in the video below is quite a bit more than a proof of concept. The NPCs communicate as though they are aware of surrounding events and conditions in the game world, are generally less forthcoming when talking to strangers, and the new system can interact with game mechanics and elements such as money, quests, and hirelings.

Starting around 1:08 into the video, [Bloc] talks to a peasant about some bandits harassing the community, and from there demonstrates hiring some locals and haggling over prices before heading out to deal with the bandits.

The downside is that ChatGPT is currently amazingly popular. As a result, [Bloc]’s mod is stuck using an overloaded service which means some painfully-long load times between each exchange. But if you can look past that, it’s a pretty fascinating demonstration of what’s possible by gluing two systems together with a mod and some clever coding.

Take a few minutes to check out the video, embedded below. And if you’re more of a tabletop gamer? Let us remind you that it might be fun to try replacing your DM with ChatGPT.

Continue reading “With ChatGPT, Game NPCs Get A Lot More Interesting”

Illustrated Kristina with an IBM Model M keyboard floating between her hands.

Keebin’ With Kristina: The One With The Breadboard Macropad

For their first custom, hand-wired keyboard, [terryorchard] aka [70rch] didn’t want to mess with making a total split, and we don’t really blame them. However, as you can see, they ended up with a monoblock split, which aside from being our own personal preference, looks fantastic, and also happened to be what fit on the print bed.

What you’re looking at is a 40% remix of the Alice layout with a columnar stagger. It’s also a bit 6×3 Corne-inspired on the ergonomic front. Brain-wise, it’s got an exposed Elite Pi driving a matrix of Kailh Choc pinks and an EC11 encoder. The encoder scrolls by default, and then becomes a volume knob on the numbers and symbols layer. One super cool thing about this keyboard is the secret third layer, which is unlocked by pressing the rotary encoder. This leads to some home row mods and disables the outside columns, culminating in a test 3×5 with two layers.

Via KBD #112

Continue reading “Keebin’ With Kristina: The One With The Breadboard Macropad”

Developing Warp Drive Might Take Antifreeze

In Star Trek, dilithium crystals — whatever those are — are critical to the operation of a starship’s warp drives. But a Texas professor thinks he can make a baby step towards a warp drive using ethylene glycol, which is commonly found in antifreeze.

While superluminal travel has been regarded as impossible for many years, recent work has suggested ways we might be able to circumvent the light-speed barrier. Unfortunately, all of these mathematical theories require energy and types of matter that we don’t know how to create yet. But [Dr. Chance Glen] believes that by shaping the energy in a specific way through a dielectric, the math can work out so that there’s no exotic negative energy required.

The experiment involves shooting RF energy into an antifreeze container and using a laser interferometer to detect gravity waves.  Of course, that will involve some very sensitive measurements to account for other tiny perturbations that might give false readings. As we’ve seen in the past, that’s a task easier said than done.

Does this make sense? Beats us. Our physics and math are too out of date to make a good guess about how much of this is real and how much is hype. Of course, if he does detect gravity waves, that will get us as close to warp drive as the invention of the telegraph got us to cell phones. Then again, you have to start somewhere.

If you want to know more about the state of rocket engines, including the nascent possibility of warp drives, we’ve discussed that before. Incidentally, if you think the experiment sounds a bit like the the Laser Interferometer Gravitational-wave Observatory (LIGO), you aren’t wrong.

Model Train Delivers Fresh Coffee

Model trains are good fun, though few of them serve any purpose beyond amusement or authentic railway simulation. [ProjectAir] decided to put his model train to practical use by having it deliver fresh espresso, and faced plenty of difficult challenges along the way.

It sounds simple, but the practicalities of the task proved difficult. After all, even a slight wobble is enough to tip a coffee cup off a small train. Automating everything from the railway itself to the kitchen coffee machine was no mean feat either. Plus, the aim was to deliver coffee from a downstairs kitchen up to an upstairs office. This meant finding a way to get the train to climb a steep staircase and to carry the coffee over a 20-meter journey without losing the caffeinated beverage in the process. That required the construction of a fancy train elevator to do the job — an impressive accomplishment on its own.

The final system is a joy to watch. Having a train roll into the upstairs workshop with a fresh brew certainly beats having to go all the way downstairs for a cup. Just don’t think about the fact that moving the coffee machine upstairs might have been a quicker solution.

Continue reading “Model Train Delivers Fresh Coffee”

Connecting A Keyboard To A Vintage PC-XT, The Hard Way

We’re not sure if there’s any single characteristic that qualifies someone as a hacker. After all, we’re a pretty eclectic bunch, with skills that range all over the map, and what one person feels is trivial, others would look upon as black magic. But there’s one thing we’re sure of: if you find yourself reading the original POST code for the PC-XT motherboard just to get a keyboard working, you’re pretty much our kind of people.

That was the position [Anders Nielsen] found himself in as work progresses on his “PC-XT from Scratch” project, which seeks to build a working mid-80s vintage IBM Model 5160 using as many period-correct parts as possible. The first installment of the series featured the delicate process of bringing the motherboard up, lest the magic smoke was released. After seeing some life out of the old board, [Anders] needed a little IO, specifically video and keyboard. The video side of the equation was relatively trivial, with an early-90s VGA card from eBay — not exactly period correct, but good enough to get something to display. Continue reading “Connecting A Keyboard To A Vintage PC-XT, The Hard Way”

LED Displays May Get Vertical Integration

If you zoom into the screen you are reading this on, you’d see an extremely fine pattern of red, green, and blue emitters, probably LEDs of some kind. This somewhat limits the resolution you can obtain since you have to cram three LEDs into each screen pixel. Engineers at MIT, however, want to do it differently. By growing thin LED films and sandwiching them together, they can produce 4-micron-wide LEDs that produce the full range of color, with each color part of a vertical stack of LEDs.

To put things in perspective, a standard TV LED is at least 200 microns across. Mini LEDs measure upwards of 100 microns, and micro LEDs are the smallest of all. A key factor for displays is the pitch — the distance from the center of one pixel to the center of the next. For example, the 44mm version of the Apple Watch has a pitch of around 77 microns. A Samsung Galaxy 10 is just over 46 microns.  This is important because it sets the minimum size for a high-resolution screen, especially if you are building large screens (such as when you build custom video walls (see the video below for more about that).

Continue reading “LED Displays May Get Vertical Integration”

Old 3D CAD Mouse Gets New Lease Of Life

[Jacek Fedorynski] had an old Magellan/SpaceMouse 3D mouse with a serial interface which made it impossible for him to use with modern hardware and software. The problem he faced was two pronged – the absence of serial interfaces in the hardware and the lack of appropriate drivers for the operating system. So he built a low cost, simple adapter to use his RS-232 Magellan/SpaceMouse with modern software.

The hardware required to build the adapter was minimal. A Raspberry Pi Pico, a MAX3238 based RS-232 adapter, a null modem adapter and a DB9 gender changer. Of course, a combination null modem – gender changer would have made things even simpler. Four of the GPIO pins from the Pico are mapped to the serial RX, TX, RTS and CTS pins.

On the software side, the code emulates a 3DConnexion SpaceMouse Compact, so it can be used with software like Fusion 360, 3ds Max, SolidWorks, Inventor, Maya and many others. On the host computer, only the standard 3DxWare driver package is needed. On the host computer, the old Magellan/SpaceMouse 3D will appear like a modern SpaceMouse Compact connected over USB. The only downside to this is that the SpaceMouse Compact has just two programmable buttons, so only two of the many buttons on the old Magellan mouse can be mapped.

Flashing the code to the Pico is also straightforward using the BOOTSEL mode. Hold down the BOOTSEL button when plugging in the Pico and it appears as a drive onto which you can drag a new UF2 file. Just drag-n-drop [Jacek]’s magellan.uf2 firmware and you’re done.

If you’d rather build your own, modern 3D mouse, check out the DIY Cad Mouse You Can Actually Build.