Microsoft’s latest Phi4 LLM has 14 billion parameters that require about 11 GB of storage. Can you run it on a Raspberry Pi? Get serious. However, the Phi4-mini-reasoning model is a cut-down version with “only” 3.8 billion parameters that requires 3.2 GB. That’s more realistic and, in a recent video, [Gary Explains] tells you how to add this LLM to your Raspberry Pi arsenal.
The version [Gary] uses has four-bit quantization and, as you might expect, the performance isn’t going to be stellar. If you are versed in all the LLM lingo, the quantization is the way weights are stored, and, in general, the more parameters a model uses, the more things it can figure out.
While most of us won’t ever play Wimbledon, we can play Pong. But it isn’t the same without the thrill of the sportscaster’s commentary during the game. Thanks to [Parth Parikh] and an LLM, you can now watch Pong matches with commentary during the game. You can see the very cool result in the video below — the game itself starts around the 2:50 mark. Sadly, you don’t get to play. It seems like it wouldn’t be that hard to wire yourself in with a little programming.
The game features multiple AI players and two announcers. There are 15 years of tournaments, including four majors, for a total of 60 events. In the 16th year, the two top players face off in the World Championship Final.
There are several interesting techniques here. For one, each action is logged as an event that generates metrics and is prioritized. If an important game event occurs, commentary pauses to announce that event and then picks back up where it left off.
We really want to see a one- or two-player human version of this. Please tell us if you take on that challenge. Even if you don’t write it, maybe the AI can write it for you.
The future of healthy indoor plants, courtesy of AI. (Credit: [Liam])Like so many of us, [Liam] has a big problem. Whether it’s the curse of Brown Thumbs or something else, those darn houseplants just keep dying despite guides always telling you how incredibly easy it is to keep them from wilting with a modicum of care each day, even without opting for succulents or cactuses. In a fit of despair [Liam] decided to pin his hopes on what we have come to accept as the Savior of Humankind, namely ‘AI’, which can stand for a lot of things, but it’s definitely really smart and can even generate pretty pictures, which is something that the average human can not. Hence it’s time to let an LLM do all the smart plant caring stuff with ‘PlantMom’.
Since LLMs so far don’t come with physical appendages by default, some hardware had to be plugged together to measure parameters like light, temperature and soil moisture. Add to this a grow light and a water pump and all that remained was to tell the LMM using an extensive prompt, containing Python code, what it should do (keep the the plant alive), and what Python methods are available. All that was left now was to let the Google’s Gemma 3 handle it.
To say that this resulted in a dramatic failure along with what reads like an emotional breakdown on the part of the LLM would be an understatement. The LLM insisted on turning the grow light on when it should be off and had the most erratic watering responses imaginable based on absolutely incorrect interpretations of the ADC data, flipping dry and wet. After this episode the poor chili plant’s soil was absolutely saturated and is still trying to dry out, while the ongoing LLM experiment, with an empty water tank, has the grow light blasting more often than a weed farm.
So far it seems like that the humble state machine’s job is still safe from being taken over by ‘AI’, and not even brown thumb folk can kill plants this efficiently.
While tools like CRISPR have blown the field of genome hacking wide open, being able to predict what will happen when you tinker with the code underlying the living things on our planet is still tricky. Researchers at Stanford hope their new Evo 2 DNA generative AI tool can help.
Trained on a dataset of over 100,000 organisms from bacteria to humans, the system can quickly determine what mutations contribute to certain diseases and what mutations are mostly harmless. An “area we are hopeful about is using Evo 2 for designing new genetic sequences with specific functions of interest.”
To that end, the system can also generate gene sequences from a starting prompt like any other LLM as well as cross-reference the results to see if the sequence already occurs in nature to aid in predicting what the sequence might do in real life. These synthetic sequences can then be made using CRISPR or similar techniques in the lab for testing. While the prospect of building our own Moya is exciting, we do wonder what possible negative consequences could come from this technology, despite the hand-wavy mention of not training the model on viruses to “to prevent Evo 2 from being used to create new or more dangerous diseases.”
Most people have wished for the ability to talk to other animals at some point, until they realized their cat would mostly insult them and ask for better service, but researchers are getting closer to a dolphin translator.
DolphinGemma is an upcoming LLM based on the recordings from the Wild Dolphin Project. Using the hours and hours of dolphin sounds recorded by researchers over the decades, the hope is that the LLM will allow us to communicate more effectively with the second most intelligent species on the planet.
The LLM is designed to run in the field on Google Pixel phones, due to it being based on Google’s in-house Gemini product, which is a bit less cumbersome than hauling a mainframe on a dive. The Wild Dolphin Project currently uses the Georgia Tech developed CHAT (Cetacean Hearing Augmentation Telemetry) device which has a Pixel 6 at its heart, but the newer system will be bumped up to a Pixel 9 to take advantage of all those shiny new AI processing advances. Hopefully, we’ll have a better chance of catching when they say, “So long and thanks for all the fish.”
Will a 486 run Crysis? No, of course not. Will it run a large language model (LLM)? Given the huge buildout of compute power to do just that, many people would scoff at the very notion. But [Yeo Kheng Meng] is not many people.
He has set up various DOS computers to run a stripped down version of the Llama 2 LLM, originally from Meta. More specifically, [Yeo Kheng Meng] is implementing [Andreq Karpathy]’s Llama2.c library, which we have seen here before, running on Windows 98.
Llama2.c is a wonderful bit of programming that lets one inference a trained Llama2 model in only seven hundred lines of C. It it is seven hundred lines of modern C, however, so porting to DOS 6.22 and the outdated i386 architecture took some doing. [Yeo Kheng Meng] documents that work, and benchmarks a few retrocomputers. As painful as it may be to say — yes, a 486 or a Pentium 1 can now be counted as “retro”.
The models are not large, of course, with TinyStories-trained 260 kB model churning out a blistering 2.08 tokens per second on a generic 486 box. Newer machines can run larger models faster, of course. Ironically a Pentium M Thinkpad T24 (was that really 21 years ago?) is able to run a larger 110 Mb model faster than [Yeo Kheng Meng]’s modern Ryzen 5 desktop. Not because the Pentium M is going blazing fast, mind you, but because a memory allocation error prevented that model from running on the modern CPU. Slow and steady finishes the race, it seems.
This port will run on any 32-bit i386 hardware, which leaves the 16-bit regime as the next challenge. If one of you can get an Llama 2 hosted locally on an 286 or a 68000-based machine, then we may have to stop asking “Does it run DOOM?” and start asking “Will it run an LLM?”
[Geoffrey Litt] shows that getting an effective digital assistant that’s tailored to one’s own needs just needs a little DIY, and thanks to the kinds of tools that are available today, it doesn’t even have to be particularly complex. Meet Stevens, the AI assistant who provides the family with useful daily briefs. The back end? Little more than one SQLite table and a few cron jobs.
A sample of Stevens’ notebook entries, both events and things to simply remember.
Every day, Stevens sends a daily brief via Telegram that includes calendar events, appointments, weather notes, reminders, and even a fun fact for the day. Stevens isn’t just send-only, either. Users can add new entries or ask questions about items through Telegram.
It’s rudimentary, but [Geoffrey] already finds it far more useful than Siri. This is unsurprising, as it has been astutely observed that big tech’s digital assistants are designed to serve their makers rather than their users. Besides, it’s also fun to have the freedom to give an assistant its own personality, something existing offerings sorely lack.
Architecture-wise, the assistant has a notebook (the single SQLite table) that gets populated with entries. These entries come from things like reading family members’ Google calendars, pulling data from a public weather API, processing delivery notices from the post office, and Telegram conversations. With a notebook of such entries (along with a date the entry is expected to be relevant), generating a daily brief is simple. After all, LLMs (Large Language Models) are amazingly good at handling and formatting natural language. That’s something even a locally-installed LLM can do with ease.
[Geoffrey] says that even this simple architecture is super useful, and it’s not even a particularly complex system. He encourages anyone who’s interested to check out his project, and see for themselves how useful even a minimally-informed assistant can be when it’s designed with ones’ own needs in mind.
![The future of healthy indoor plants, courtesy of AI. (Credit: [Liam])](https://hackaday.com/wp-content/uploads/2025/04/plantmom_test_setup_with_plant.jpg)
