While this may have been a “low” year for AI, 2025 has seen the emergence of some truly innovative, interesting, and even beneficial uses of artificial intelligence. Here, we round up eight of our favorites as part of Dezeen's 2025 review.
“Slop” is the defining word of 2025, at least according to Merriam-Webster, describing the flood of mindless, machine-generated content flooding social media thanks to generative AI systems like OpenAI’s Sora 2.
Meanwhile, the companies developing these models are still embroiled in an arms race to develop the most intelligent algorithms, spending unimaginable amounts of money building gigantic data centers that already generate more annual carbon dioxide emissions than New York City, and guzzle more water (via water bottles, that is) than humans.
But outside the generative AI bubble, architects, designers, and engineers are coming up with some really interesting new applications for this emerging technology, from dream visualization and plant maintenance to finding salvageable materials among Ukrainian ruins.
Here are eight highlights.
dream recorder by modem
Dutch design studio Modem, founded by Astin le Clercq and Space10 veteran Bas Van De Poel, has quietly been delivering some of the most innovative uses of AI since its launch in 2021.
These include a Compass designed to help people walk without using their phones, and a chair co-created by British designer Ross Lovegrove and Google AI trained to emulate his style.
This year's most read AI article was Modem's Dream Recorder, a bedside device that allows users to play back their dreams as short animated videos.
Click here for more information about Dream Recorder ›
WaterSense Buoy by Filip Budny
The winner of this year's James Dyson Award for Global Sustainability was a buoy that uses AI to monitor water quality in rivers and lakes, providing local communities with real-time data to fight pollution.
Inland waters are particularly susceptible to agricultural runoff, sewage, and industrial waste, but are currently monitored only through sporadic manual inspections.
In these blind spots, Polish engineer Filip Budny's WaterSense buoy provides unprecedented monitoring by continuously and autonomously assessing water quality using inexpensive paper sensors that are evaluated not by humans but by AI, which he says can predict pollution events up to 72 hours in advance.
Click here for more information about WaterSense ›
Neo Gamma robot by 1X
California company 1X made headlines earlier this year when it began taking pre-orders for Neo Gamma, a robot housekeeper that can do household chores for you.
The robot, dressed head-to-toe in beige knitwear, will reportedly be piloted by a multitasking AI model that can process both verbal and visual information. This means that Neo Gamma can understand and respond to voice commands, while also controlling all the motor functions necessary to carry them out.
The problem is that, at least in the early stages, robots will rely heavily on remote control, with 1X employees looking through the robot's eyes and acting as metaphorical guardrails while the AI is trained in real-life home scenarios.
Learn more about Neo Gamma ›
FOODres.AI by MIT's Biru Cao and Yiqing Wang
On the more experimental side, two MIT students have developed a prototype AI-powered 3D printer. This printer guides users on how to turn food waste into printing filament.
Whether derived from fruit peels, coffee grounds, or leftover rice, the resulting biomaterials can be used to 3D print everyday items such as cup holders and small containers in place of traditional plastic filaments.
So far, designers Biru Cao and Yiqing Wang have only built a partial prototype. But they are working on a fully realized version that they hope will one day help tackle the food waste crisis.
Learn more about FOODres.AI ›
LeafyPod by LeafyPod
One of the hottest products announced at the 2025 Consumer Electronics Show in Las Vegas was this self-watering smart planter.
LeafyPod is designed for underprivileged nature lovers and features an AI-powered watering system that automatically adapts to the needs of current residents.
This calculation is based not only on the type of plants being planted, but also real-time data on light exposure, humidity, and temperature collected via a series of sensors hidden inside the minimalist white planter.
Learn more about LeafyPod ›
“Circularity on the Edge” by Kateryna Lopatiuk and Herman Mitish
The use of artificial intelligence in conflict zones such as Ukraine and Gaza has raised serious ethical concerns in recent years, with devastating consequences.
But Ukrainian architect Katerina Lopatyuk and developer Herman Mitish have also discovered how the technology can actually help countries rebuild.
Their project, “Circularity on the Edge,” uses computer vision and remote sensing techniques to analyze drone footage to identify salvageable materials in the rubble, which will be used in recovery efforts.
Learn more about Circularity on the Edge ›
Even G2 Smart Glasses by Even Realities
Smart glasses are expected to be the next big product category for artificial intelligence, with a veritable avalanche of AI-powered specifications being released in 2025.
Meta's collaboration with Ray-Ban is perhaps the most notable, with tech giants like Apple, Samsung, and Google reportedly developing their own models.
So far, Chinese company Even Realities' Even G2 stands out by looking most like regular glasses with prescription lenses, while eliminating the controversial outward-facing camera found in other models, which has raised privacy and surveillance concerns.
Learn more about Even G2 ›
Prompting Nowhere Foot-operated AI by Max Park
In this important project by Central Saint Martins graduate Max Park, a vintage treadle sewing machine generates the energy needed to run a large-scale language model (LLM).
The AI itself is trained on the socialist ideals of textile designer William Morris, making visible the hidden labor behind the technology by making its output dependent on manual effort on the part of the user.
Unlike traditional LLMs such as ChatGPT and Claude that rely on far-flung data centers, all data is processed locally to help users regain a sense of ownership over artificial intelligence.
Learn more about Prompting Nowhere ›
