Modders can change many things inside their favorite games, but dialogue from professionally voiced characters hasn’t been one of those things—at least until recently. AI voice generation could open up new modding avenues for some games, as it has already done with one Fallout 4 mod package.

[…]

They’re not just new labels on existing dialogue, either. RED, created by NexusMods user ProfMajowski (and first seen by us at PCGamesN), says it used ElevenLabs voice AI to generate its more in-character lines. The results can sometimes “sound a little ’emotionless,'” the creator writes, but “otherwise they basically sound like the real thing.” Nothing your character can newly say now will change the game’s mechanics or reactions, but it should sound a bit more in character.

I’m not down on “artificial intelligence” as a matter of principle – quite the opposite. This story right here is a great example of how AI can be used in productive, interesting ways that truly make something possible that either wasn’t possible before, or was simply entirely unrealistic.

Spoken dialog is hard to record for a whole slew of reasons, from cost to finding enough quality voice actors to the time it takes, and it’s usually only the biggest studios that have the ability to add it to their games. Even then it’s often a struggle, from bad voice acting overall to large role-playing games where e.g. the main quest is beautifully voiced, but side quests are either entirely unvoiced or clearly rushed by some cheap interns.

Thanks to technology like this, even small indie studios or mere mod developers can add something meaningful to their work that up until recently simply wasn’t realistic. It will make games meaningfully better – especially once technology improves a bit more and developers become proficient with it – without the need for hype.

Butler’s goal is to implement a virtual operating system with a strong focus on concurrency. Butler is not a computer hardware operating system. Instead, Butler is an application environment that runs on top of an existing system.

Not exactly an operating system in the classical sense, but it still seems like an interesting project.

Today, we’re introducing a major set of upgrades to the Framework Laptop spanning two new models – the Framework Laptop 13 (13th Gen Intel® Core™) and the Framework Laptop 13 (AMD Ryzen™ 7040 Series). We’ve not only scaled up performance and enabled an AMD-powered version for the first time, but we’ve also delivered refinements to the day-to-day user experience with a higher capacity battery, matte display, louder speakers, and more ridgid hinges.

And Framework kept their promise: these new mainboards can be ordered separately and fit into the existing Framework 13″ laptop.

The company also showed off their next product – a 16″ laptop that not only comes with an upgradeable GPU, but also a completely configurable input deck, so you can configure the keyboard and trackpad area in any configuration you like.

I’m so happy Framework is doing well, as it shows that glued shut, non-repeairable, and non-upgradeable laptops are not some sort of universal inevitable truth.

Google Bard is out—sort of. Google says you can now join the waitlist to try the company’s generative AI chatbot at the newly launched bard.google.com site. The company is going with “Bard” and not the “Google Assistant” chatbot branding it was previously using. Other than a sign-up link and an FAQ, there isn’t much there right now.

Google’s blog post calls Bard “an early experiment,” and the project is covered in warning labels. The Bard site has a bright blue “Experiment” label right on the logo, and the blog post warns, “Large language models will not always get it right. Feedback from a wide range of experts and users will help Bard improve.” A disclaimer below the demo input box warns, “Bard may display inaccurate or offensive information that doesn’t represent Google’s views.”

Google’s Android keyboard and spell checker still can’t get “it’s” vs. “its” right 80% of the time, so I’m not holding my breath.

Car companies have been increasingly using digital screens and soft-touch buttons in modern cars to save costs while looking ‘hi-tech’ – but Hyundai has committed to fight this trend for as long as possible.

Speaking at the launch of the new-generation Hyundai Kona, Sang Yup Lee, Head of Hyundai Design, said the new model deliberately uses physical buttons and dials for many of the controls, specifically air-conditioning and the sound system. Lee said this is because the move to digital screens is often more dangerous, as it often requires multiple steps and means drivers have to take their eyes off the road to see where they need to press.

Slowly but surely, it seems car makers are starting to see the light. A clean, button-less dashboard means nothing once it’s folded around your crushed skull because you couldn’t find the seat heating button without taking your eyes off the road and wrapped yourself around a tree in the process.

Just another reason to get an Ioniq 5 if we had the funds.

In fact, the broken bar barely even exists anymore. In the days of DOS, the character used for the pipe symbol (on the DOS command line) or for logical OR (in C/C++, for example) used ASCII code 7Ch (124 decimal), which was rendered as a broken vertical bar by the fonts used at least by the IBM MDA, CGA, EGA, and VGA cards. But nowadays that is no longer the case. The same ASCII codepoint is rendered as a solid vertical bar in Windows 10 or Linux, and also shown as a solid vertical bar on contemporary keyboards. What happened?

Who doesn’t love some great character and ASCII archeology?

We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning. GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various professional and academic benchmarks. For example, it passes a simulated bar exam with a score around the top 10% of test takers; in contrast, GPT-3.5’s score was around the bottom 10%. We’ve spent 6 months iteratively aligning GPT-4 using lessons from our adversarial testing program as well as ChatGPT, resulting in our best-ever results (though far from perfect) on factuality, steerability, and refusing to go outside of guardrails.

“Artificial intelligence” companies are iterating quickly now. I’m definitely looking forward to the new memes based on GPT-4.

Coming up with a title that explains the full story here was difficult, so I’m going to try to explain quickly.

Yesterday, Docker sent an email to all Docker Hub users explaining that anyone who has created an “organisation” will have their account deleted including all images, if they do not upgrade to a paid team plan. The email contained a link to a tersely written PDF (since, silently edited) which was missing many important details which caused significant anxiety and additional work for open source maintainers.

What a shitshow. We really have to start worrying about the future of Github, too, since I find it highly unlikely Microsoft isn’t planning similar moves in the future. If you’re hosting code at Github, I’d suggest looking at alternatives sooner rather than later, so you don’t end up like the people affected by something like this.

If you’re new to the Arm ecosystem, consider this a quick primer on terms you likely have seen before but might have questions about.

Well, exactly what it says.

If you’ve been reading my blog for a while, you might remember back in 2012 when I changed the startup sound on my Power Mac G3 (Blue and White). That was a fun introduction to the Forth programming language. I had to reverse-engineer just enough of Apple’s firmware update script to understand what was going on.

Recently, Aidan Halpin, a reader of this site, asked me if I could do the same kind of startup sound customization on his iMac. This particular iMac is officially known as the “iMac (Slot Loading)” and has a model identifier of PowerMac2,1. As you can guess from the name, it has a slot-loading CD-ROM drive unlike the original iMac that had a laptop-style tray-loading drive. By the way, Aidan’s iMac is special because it has a PowerPC G4 processor soldered onto the logic board instead of the original G3.

He sent me Apple’s last firmware update for this model: iMac Firmware Update 4.1.9. I went to work looking at the update contents to see if I could figure out how to modify the chime the same way I did with my Power Mac G3. I thought it would be fun to take everyone along for a ride and show exactly what was involved in changing the sound. And of course, this post wouldn’t be complete without also sharing the code for the utility I created to inject the new chime into the firmware update file.

This is an insane amount of work for something that doesn’t really matter in the end. I love it.

Hyperion Entertainment has released another update for AmigaOS 3.2 for classic Amigas, coming with a number of improvements and bug fixes. I’m not entirely sure what to make of all this, though, since the drama around the ownership of the Amiga operating system, the trademarks, and more, as well as continuous accusations of Hyperion not paying any of its developers, have reached a fever pitch, as documented in this elaborate piece. As much as I would want to dive into all this and properly vet every single source in that article, for the sake of my sanity, I am just not going to.

The soap opera around the Amiga has been going on for so long, and has jumped the shark so many times, I just don’t know where to start. I’ll leave you all with the detailed piece and its sources, and let you decide for yourself what to make of it all.

I ain’t got the patience for this.

Cobalt Networks were one of the early pioneers in network appliance hardware and produced some of the first turn-key webserver boxes you could buy, founded in 1996 as Cobalt Microserver. Cobalt boxes are immediately identifiable from their distinctive deep blue plastic bezels starting with the 1998 Cobalt Qube 2700. The Qube used a 150MHz QED RM5230; these CPUs are part of QED’s R5000 family and we’ll talk about their architecture a bit later. They came with 2.1GB hard disks with later larger options, 10Mbit Ethernet, 16MB of RAM standard with up to 256MB supported, and a “console” consisting of a backlit rear-mounted 2-line LCD and control buttons (on later machines, but not the original 2700, a serial port provided an actual console if you held down a button during startup). A fair number of typical configuration tasks such as setting its IP address could be done directly from the panel and the rest were intended to be done through its Perl-based web console. They were designed to run Linux from the ground up and shipped with Red Hat using a 2.0.x kernel.

Cobalt’s network appliances were so exotic back in the day, and once they started hitting the used market, I almost pulled the trigger quite a few times. These days, they’re harder to come by, and their use is, of course, inherently limited now, but that doesn’t make them any less eye-catching.

Tech-makers assuming their reality accurately represents the world create many different kinds of problems. The training data for ChatGPT is believed to include most or all of Wikipedia, pages linked from Reddit, a billion words grabbed off the internet. (It can’t include, say, e-book copies of everything in the Stanford library, as books are protected by copyright law.) The humans who wrote all those words online overrepresent white people. They overrepresent men. They overrepresent wealth. What’s more, we all know what’s out there on the internet: vast swamps of racism, sexism, homophobia, Islamophobia, neo-Nazism.

Tech companies do put some effort into cleaning up their models, often by filtering out chunks of speech that include any of the 400 or so words on “Our List of Dirty, Naughty, Obscene, and Otherwise Bad Words,” a list that was originally compiled by Shutterstock developers and uploaded to GitHub to automate the concern, “What wouldn’t we want to suggest that people look at?” OpenAI also contracted out what’s known as ghost labor: gig workers, including some in Kenya (a former British Empire state, where people speak Empire English) who make $2 an hour to read and tag the worst stuff imaginable — pedophilia, bestiality, you name it — so it can be weeded out. The filtering leads to its own issues. If you remove content with words about sex, you lose content of in-groups talking with one another about those things.

These things are not AI. Repeat after me: these things are not AI. All they do is statistically predict the best next sequence of words based on a corpus of texts. That’s it. I’m not worried about these things leading to SkyNet – I’m much more worried about smart people falling for the hype.

The JH7110 isn’t amazing. But it’s not bad, either.

I still wouldn’t recommend most people buy this board, unless you already know a lot about Linux and SBCs in general. That may change a year from now, but right now, this board isn’t targeted at the same market as a Raspberry Pi.

At around $100, and not being quite production-ready, I’m only recommending this board to people interested in exploring RISC-V for now.

This seems like an expected experience for a relatively new architecture that still has rather limited hardware and software support. When the first Raspberry Pi came out, the situation wasn’t much better either, so give it a few years and RISC-V will be in a better place in the market for sub-€100 single-board computers.