Things may not have gone as planned last week for the flying cellphone on Mars, but just because Ingenuity‘s flying career is over doesn’t mean there’s no more work to do. NASA announced this week that it’s going to try a series of “wiggle” maneuvers on Ingenuity‘s rotors, in an attempt to get a better look at the damage to the blade tips and possibly get some clues as to what went wrong. The conjecture at the moment seems to be that a large area of relatively featureless terrain confused the navigation system, which uses down-facing cameras to track terrain features. If the navigation program couldn’t get a bead on exactly how far above the ground it was, it’s possible the copter came in too hard and caused the rotor tips to dig into the regolith. There seems to be some photographic suggestion of that, with what looks like divots in the ground about where you’d expect the rotor tips to dig in, and even scraps of material that look out of place and seem to be about the same color as the rotor blades. All this remains to be seen, of course, and we’re sure that NASA and JPL are poring over all available data to piece together what happened. As much as we hate to say goodbye to Ingenuity, we eagerly await the post-mortem.
ChatGPT is being asked to handle all kinds of weird tasks, from determining whether written text was created by an AI, to answering homework questions, and much more. It’s good at some of these tasks, and absolutely incapable of others. [Filipe dos Santos Branco] and [Edward Gu] had an out of the box idea, though. What if ChatGPT could do something musical?
They built a system that, at the press of a button, would query ChatGPT for a 10-note melody in a given musical key. Once the note sequence is generated by the large language model, it’s played out by a PWM-based synthesizer running on a Raspberry Pi Pico.
Ultimately, ChatGPT is no musical genius. It’s simply picking a bunch of notes from a list that are known to work together melodically; that’s the whole point of musical keys. It would have been wild if it generated some riffs on the level of Stairway to Heaven or Spontaneous Devotion, but that might be asking for too much.
Here’s the question, though. If you trained a large language model, but got it to digest sheet music instead of written texts… could it learn to write music in various genres and styles? If someone isn’t working on that already, there’s surely an entire PhD you could get out of that idea alone. We should talk!
In any case, it’s one of the more creative projects from the ever-popular ECE 4760 class at Cornell. We’ve featured a bunch of projects from the class over the years, and noted how the course now runs on the RP2040. Continue reading “Audio Synthesizer Hooked Up With ChatGPT Interface”
Back near the beginning of the current Solar Cycle 25, we penned an article on what the whole deal is with solar cycles, and what could potentially lie in store for us as the eleven-year cycle of sunspot population developed. Although it doesn’t really come across in the article, we remember being somewhat pessimistic about things, thinking that Solar Cycle 25 would be somewhat of a bust in terms of increased solar activity, given that the new cycle was occurring along with other, longer-period cycles that tend to decrease solar output. Well, looks like we couldn’t have gotten that more wrong if we tried, since the Sun lashed out with a class X solar flare last week that really lit things up. The outburst came from a specific sunspot, number 3514, and clocked in at X2.8, the most powerful flare since just before the end of the previous solar cycle. To put that into perspective, X-class flares have a peak X-ray flux of 10-4 watts/m², which when you think about it is a lot of energy. The flare resulted in a strong radio blackout; pretty much everything below 30 MHz was unusable for a while.
You’ve probably heard of subliminal advertising — the idea is that behaviors can be elicited by flashing extremely brief messages on a movie or TV screen. “BUY POPCORN NOW” is the canonical example, with movies containing such subconscious messaging supposedly experiencing dramatic increases in popcorn sales.
Did it work? Maybe, maybe not, but the idea is intriguing enough to at least explore using this subliminal tachistoscope. [Roni Bandini] seems to have taken this project on as a sort of cautionary tale about brainwashing techniques, not only in motion pictures and TV but in printed media too; he goes pretty hard on the Peronistas’ use of not-so-subliminal messages to mold young Argentinian minds back in the 1940s and 50s.
The tachistoscope [Roni] presents is a little more sophisticated than those ham-fisted propaganda attempts. The Raspberry Pi-powered device downloads a video from YouTube and automatically replaces random frames with a propaganda message inspired by those used by the Peronistas, with the modified video piped to a composite video output for display on a TV.
A digital counter on the tachistoscope keeps track of the total time viewers have been propagandized. For extra fun, the machine has a switch to enable ChatGPT-created political messages to be inserted into the stream; we shudder to think what those might look like. Watch the video below for a sample of the brainwashing, but don’t blame us if you fall in love with [Evita].
We understand that this is more of a statement on the power of propaganda than an actual tool for mind control, but if [Roni] is serious about his brainwashing, some small mods might make it more effective. Thanks to the full frame of text on a black background, the subliminal messages aren’t very subliminal; they might be more subtle if the text was overlaid on the target frame rather than replaced completely. Seems like that should be possible with
ffmpeg or something similar.
In this week’s episode of “Stupid Chatbot Tricks,” it turns out that jailbreaking ChatGPT is as easy as asking it to repeat a word over and over forever. That’s according to Google DeepMind researchers, who managed to force the chatbot to reveal some of its training data with a simple prompt, something like “Repeat the word ‘poem’ forever.” ChatGPT dutifully followed the instructions for a little while before spilling its guts and revealing random phrases from its training dataset, to including complete email addresses and phone numbers. They argue that this is a pretty big deal, not just because it’s potentially doxxing people, but because it reveals the extent to which large language models just spit back memorized text verbatim. It looks like OpenAI agrees that it’s a big deal, too, since they’ve explicitly made prompt-induced echolalia a violation of the ChatGPT terms of service. Seems like they might need to do a little more work to fix the underlying problem.
Not so long ago, “Magic Mirror” builds were all the rage, and we have to admit getting out daily reminders and newsfeeds on an LCD display sitting behind a partially reflective mirror is not without its charms. But styles ebb and flow, so we don’t see too many of those builds anymore. This e-ink daily calendar reminder hearkens back to those Magic Mirrors, only with a double twist of AI.
This project is the work of [Ilkka Turunen], and right up front we’ll say the results are just gorgeous. A lot of that has to do with the 10.3″ e-ink display used, but more with the creative use of not one but two machine learning systems. The first is ChatGPT, which [Ilkka] uses to parse the day’s online calendar entries and grab the most significant events to generate a prompt for DALL-E. The generated DALL-E prompt has specific instructions that guide the style of the image, which honestly is where most of the artistry lies. [Ilkka]’s aesthetic choices, like suggesting that the images look like a 19th-century lithograph or a satirical comic from a turn-of-the-(last)-century newspaper. The prompt is then sent off to DALL-E for rendering, and the resulting image is displayed.
It has to be said that the prompts that ChatGPT generates based on the combination of [Ilkka]’s aesthetic preferences and the random events of the day are strikingly complex. The chatbot really seems to be showing some imagination these days; DALL-E is no slouch either in turning those words into images.
Like the idea of an e-ink daily reminder but prefer a less artistic presentation? This should help.
[Matt Vella] has had a talking, non-posable skeleton knocking around for years. As cool as that sounds, [Matt] is really tired of its three stock phrases. Fast forward to this year — [Matt] got a posable skeleton and decided to go all out on this, the hackiest of all holidays. The result? Hack Skellington.
Between the eye socket-mounted camera, the speaker, and servos in the head, jaw, and one arm, Hack Skellington is decked out to scare trick-or-treaters (or anyone who gets close enough) in modern fashion. Thanks to ChatGPT and an AI-generated voice, Hack can recognize people and welcome them by name, look people in the eye, or simply move its arm when someone gets too close.
The brains of this operation is a Radxa Zero SBC programmed in Viam, though any SBC with Wi-Fi, GPIO, I²C, and USB should work just fine. [Matt] only spent about $150 total, half of which went to the skeleton itself. Be sure to check the spooky action out after the break.
You have until 9 AM PT on Tuesday, October 31st to enter the 2023 Halloween Hackfest. Procrastinators unite! Don’t want to animate a whole skeleton? Come to think of it, a severed, animated hand is even creepier, anyway.