Category: AI

In 2020, a breakthrough in chip design emerged with the introduction of AlphaChip, a novel reinforcement learning (RL) method created to accelerate and optimize the chip layout process. Since then, AlphaChip has transformed the way computer chips are designed, generating layouts that are now used in hardware across the globe, from data centers to mobile devices. The method, developed by a team of AI researchers, was first introduced as a preprint and later published in Nature, receiving widespread recognition for its real-world engineering applications. Today, AlphaChip continues to advance chip design, with a Nature addendum detailing its impact, a release of pre-trained checkpoints, and an announcement of its formal name.

AlphaChip’s AI-driven approach has made significant contributions to the chip design industry, particularly in designing Google’s custom AI accelerator chips—Tensor Processing Units (TPUs)—which power a range of artificial intelligence (AI) systems. By applying reinforcement learning to chip floorplanning, AlphaChip is able to generate superhuman chip layouts in a matter of hours, a task that would otherwise take human designers weeks or even months.

The emergence of AI-generated art has raised numerous questions, particularly around its legality and artistic value. Can AI-generated works be copyrighted if they’re created using image generators trained on existing art? Does this process turn art into a free-for-all, allowing anyone to reproduce an AI-generated image for any purpose? And, perhaps most fundamentally, is AI-generated art truly “art,” or is it merely an imitation—a pastiche?

While the legal aspects of AI art are being debated in courts, the question of whether people actually appreciate and want to engage with AI-generated art will soon be tested in a more public arena. Los Angeles is set to become home to the world’s first permanent museum dedicated entirely to AI art. Dubbed Dataland, the museum is expected to open next year in a prime location, neighboring the Museum of Contemporary Art and The Broad Museum, placing it at the heart of the city’s thriving art scene.

Liquid AI, a startup co-founded by former MIT researchers from the Computer Science and Artificial Intelligence Laboratory (CSAIL), has introduced its first multimodal AI models, the “Liquid Foundation Models (LFMs).” These models represent a bold departure from the transformer architecture that has dominated AI development since the release of the 2017 paper “Attention Is All You Need.”

Unlike the current wave of generative AI models built on the transformer architecture, Liquid AI aims to develop foundation models from “first principles,” taking an engineering approach akin to building engines, cars, or airplanes. This fundamental shift has led to models that outperform transformer-based alternatives of similar size, such as Meta’s Llama 3.1-8B and Microsoft’s Phi-3.5 3.8B.

AI has long struggled with truth and correctness, and ironically, much of the problem stems from the way human thinking shapes these systems. But a new wave of artificial intelligence is emerging—one that sheds the constraints of human logic and ventures into the unknown, potentially advancing machine learning far beyond human capabilities.

DeepMind’s AlphaGo marked a turning point in AI development. Unlike its predecessors, which relied heavily on human instruction and knowledge, AlphaGo learned to play the game of Go with no human input, only using a process called self-play reinforcement learning. This meant it discovered the rules and strategies on its own by playing countless games and analyzing outcomes.

Researchers at MIT have developed a groundbreaking generative AI model that significantly simplifies the process of determining the structures of crystalline materials, such as metals, rocks, and ceramics. Traditionally, scientists have relied on X-ray crystallography to uncover these structures, but this new approach offers a more efficient and versatile alternative. The model has far-reaching implications for industries relying on materials like batteries, magnets, and superconductors.

“Understanding the structure of a material is crucial for virtually any application,” explained Danna Freedman, the Frederick George Keyes Professor of Chemistry at MIT. “It’s fundamental for superconductivity, magnets, photovoltaics—essentially anything that is materials-centric.”

Stockholm-based startup SweGreen is transforming the concept of “farm fresh” by bringing the farm directly into supermarkets. With installations across Sweden and Germany, SweGreen is pioneering a sustainable way to grow crops while simultaneously reducing food waste and carbon emissions.

In the U.S. alone, it’s estimated that 16 billion pounds of food are wasted annually in supermarkets. As much as 30% of the food that reaches store shelves is discarded, often because customers only select the most visually appealing fruits and vegetables. This level of waste, coupled with global challenges like climate change, shrinking agricultural land, and geopolitical tensions, raises concerns about future food security. By 2050, with the world population nearing 10 billion, finding innovative ways to ensure sustainable food production is more urgent than ever.

Purdue University engineers have found that autonomous vehicles (AVs) could soon be powered by advanced chatbots like ChatGPT, enabling them to better understand and respond to passengers’ commands. These chatbots are built on large language models (LLMs), a type of artificial intelligence algorithm designed to interpret natural language and continuously learn from vast amounts of data.

This breakthrough study, published on the preprint server arXiv, is set to be presented at the 27th IEEE International Conference on Intelligent Transportation Systems on September 25. It may be one of the first real-world experiments testing how well AVs can use LLMs to understand and act on passenger instructions.

In response to rising operational costs and labor shortages, Hestia Technology Limited, founded in Hong Kong in 2018 by chemistry professor Chen Guanhua, is transforming restaurant operations with its cutting-edge robotic kitchen systems. The company’s mission is to enhance efficiency and accessibility while delivering traditional Chinese and international cuisine through automation.

Hestia’s fifth-generation robots offer a complete automation solution for restaurant kitchens, handling everything from ingredient retrieval and preparation to serving and cleanup—entirely without human involvement. The system, which consists of a larder, conveyor belt, and multicooker, can replace up to three chefs and prepare more than 200 different dishes, including Chinese regional favorites like Hunan and Sichuan, as well as Italian cuisine.

A 64-year-old man with amyotrophic lateral sclerosis (ALS), a degenerative nerve disease, has successfully controlled Amazon’s Alexa digital assistant using only his thoughts, thanks to groundbreaking brain-computer interface technology. The company behind the innovation, Synchron, announced on Monday that the patient, named Mark, was able to stream shows, make video calls, control smart home devices, and more using just his mind.

The key to this technological feat is an implant placed in a blood vessel on the surface of Mark’s brain. This implant enables him to mentally “tap” icons on an Amazon Fire tablet, allowing him to command Alexa without using his hands or voice.

A study presented at the European Respiratory Society (ERS) Congress in Vienna revealed that the chatbot ChatGPT outperformed trainee doctors in assessing complex respiratory diseases, such as cystic fibrosis, asthma, and chest infections. The research also showed that Google’s chatbot Bard performed better than trainees in certain areas, while Microsoft’s Bing chatbot matched the trainees in overall performance.

The findings suggest that large language models (LLMs) like these could help trainee doctors, nurses, and general practitioners in triaging patients more efficiently, potentially easing the strain on healthcare systems. The study was led by Dr. Manjith Narayanan, a consultant in pediatric pulmonology at the Royal Hospital for Children and Young People, Edinburgh, and a senior lecturer at the University of Edinburgh.

While ChatGPT has propelled OpenAI to a $100 billion valuation with its 200 million weekly active users, the darker side of AI is quietly flourishing. Underground markets are exploiting large language models (LLMs) for illegal purposes, with some earning as much as $28,000 in just two months, according to a recent study published on arXiv by researchers from Indiana University Bloomington.

The study, which analyzed over 200 examples of malicious LLMs (dubbed “malas”) on underground marketplaces from April to October 2023, revealed two primary categories of these illicit models. Some are built as uncensored versions of open-source LLMs, while others are commercial models that have been “jailbroken” using specific prompts to bypass safety protocols.

The remarkable progress in artificial intelligence (AI) over recent years can largely be attributed to one factor: scale. Around the beginning of this decade, researchers noticed that as they increased the size of their algorithms and fed them more data, the capabilities of AI models skyrocketed. The latest AI models, boasting hundreds of billions to over a trillion network connections, are learning to write, code, and process information by ingesting vast amounts of data—essentially a large fraction of the internet.

This scaling, however, demands enormous computing power. According to nonprofit AI research organization Epoch AI, the computing power dedicated to AI training has been quadrupling every year. Should this trend continue, by 2030, AI models could be trained with up to 10,000 times more compute than today’s cutting-edge models, such as OpenAI’s GPT-4. This would represent a leap as significant as the difference between GPT-2’s rudimentary text generation in 2019 and GPT-4’s advanced problem-solving abilities in 2023.