MIT President Sally Kornbluth and Provost Cynthia Barnhart launched a call for papers last summer relating to generative AI, with the aim of collecting effective strategies, policy suggestions, and calls to action in this expansive field. The response was overwhelming, with a total submission of 75 proposals, out of which 27 were chosen for seed…
Overfitting is a prevalent problem when training large neural networks on limited data. It indicates a model's strong performance on the training data but its failure to perform comparably on unseen test data. This issue arises when the network’s feature detectors become overly specialized to the training data, building complex dependencies that do not apply…
The computer vision sector is currently dominated by large-scale models that offer remarkable performance but demand high computational resources, making them impractical for real-world applications. To address this, the Google Research Team has opted to reduce these models into smaller, more efficient architectures via model pruning and knowledge distillation. The team's focus is on knowledge…
Last year, a request from MIT President Sally Kornbluth and Provost Cynthia Barnhart for research proposals concerning generative AI initiatives resulted in 75 submissions. Out of these, 27 were selected for initial funding. Inspired by the impressive response, Kornbluth and Barnhart issued a second call for papers last fall. This resulted in 53 more proposals,…
The evolution of Large Language Models (LLMs) in artificial intelligence has spawned several sub-groups, including Multi-Modal LLMs, Open-Source LLMs, Domain-specific LLMs, LLM Agents, Smaller LLMs, and Non-Transformer LLMs.
Multi-Modal LLMs, such as OpenAI's Sora, Google's Gemini, and LLaVA, consolidate various types of input like images, videos, and text to perform more sophisticated tasks. OpenAI's Sora…
The creation and implementation of effective AI agents have become a vital point of interest in the Language Learning Model (LLM) field. AI company, Anthropic, recently spotlighted several successful design patterns being employed in practical applications. Discussed in relation to Claude's models, these patterns offer transferable insights for other LLMs. Five key design patterns examined…
As the use of AI, specifically linguistically-minded model (LLM) agents, increases in our world, companies are striving to create more efficient design patterns to optimize their AI resources. Recently, a company called Anthropic has introduced several patterns that are notably successful in practical applications. These patterns include Delegation, Parallelization, Specialization, Debate, and Tool Suite Experts,…
Generative AI jailbreaking is a technique that allows users to get artificial intelligence (AI) to create potentially harmful or unsafe content. Microsoft researchers recently discovered a new jailbreaking method they dubbed "Skeleton Key." This technique tricks AI into ignoring safety guidelines and Responsible AI (RAI) guardrails that help prevent it from producing offensive, illegal or…
Self-supervised learning (SSL) has broadened the application of speech technology by minimizing the requirement for labeled data. However, the current models only support approximately 100-150 of the over 7,000 languages in the world. This is primarily due to the lack of transcribed speech and the fact that only about half of these languages have formal…
Large language models (LLMs) are known for their ability to contain vast amounts of factual information, leading to their effective use in factual question-answering tasks. However, these models often create appropriate but incorrect responses due to issues related to retrieval and application of their stored knowledge. This undermines their dependability and hinders their wide adoption…
In recent months, the Massachusetts Institute of Technology (MIT) called for papers on the topic of artificial intelligence (AI), set to construct effective roadmaps, policy recommendations, and action strategies across generative AI. The response was overwhelmingly positive with 75 paper proposals submitted. Out of these, 27 were chosen for seed funding. Given the successful outcome…
The integration of automation and artificial intelligence (AI) in fungi-based bioprocesses is becoming instrumental in achieving sustainability through a circular economy model. These processes take advantage of the metabolic versatility of filamentous fungi, allowing for conversion of organic substances into bioproducts. Automation replaces manual procedures enhancing efficiency, while AI improves decision making and control based…