In this text, the author explains the concept of an "AI agent" and how it faces the same issue as the character in the movie "Groundhog Day," where it must re-read its entire story before performing an action. This leads to high costs due to repeated inputs. The author then introduces "prompt caching," a solution where previously processed inputs are saved and reused, reducing costs and maintai...
Bayer Crop Science uses machine learning models to analyze data from genomics, field trials, imagery, and environmental conditions for plant breeding. To manage and make repeatable and governable these ML model configurations, Bayer developed the Connected Decision Engine (CDE). As the usage increased, understanding the results became a challenge, leading to the creation of the Agentic Decision...
NVIDIA founder and CEO Jensen Huang recently visited Astana, Kazakhstan, where he discussed the country's push to build a strong ecosystem for artificial intelligence (AI). Kazakhstan's goal is to create AI that reflects its language and culture, utilizing its abundant energy resources to support homegrown AI and transforming energy into intelligence for export. The conversation also touched on...
In the early development of artificial intelligence (AI), engineers like those at MIT believed introducing randomness and irrationality were necessary for machines to mimic human-like creativity, according to MIT assistant professor Benjamin Lindquist. Tracing back to mid-20th century ideas, engineers saw creativity as stemming from unexpected leaps of logic or even a dream-like state. To emula...
Amazon Quick Sight's geospatial maps allow businesses to visualize patterns connected to location through point maps, filled maps, and now, layer maps. With the new point and LineString geometry types from GeoJSON files, users can mark specific locations and trace routes. This post covers dataset and GeoJSON structure, prerequisites, and advanced techniques for creating layer maps in Amazon Qui...
When updating Amazon SageMaker AI real-time endpoint configuration, changing the instance type, instance count, or environment variables through AWS CloudFormation or AWS CDK can result in deployment failure. This is due to SageMaker AI requiring the scalable target to be deregistered before updating the endpoint. However, deregistering the scalable target causes CloudFormation to lose track of...
Multi-Turn Reinforcement Learning (MTRL) is a training method used to teach models to complete multi-step tasks with feedback only after the entire sequence. In contrast to fine-tuning, MTRL doesn't rely on a single correct sequence or label. Agents interact with an environment, making decisions, and reporting rewards after each sequence ends. With MTRL, a model can be made proficient in a spec...
MIT's School of Humanities, Arts, and Social Sciences was celebrated at its 75th Anniversary Symposium for its value and contributions to teaching, research, and society. The event, held at the Linde Music Building, drew scholars, students, staff, and alumni from across the school. MIT leadership, including President Sally Kornbluth and Provost Anantha Chandrakasan, emphasized the importance of...
In this post, the author discusses the challenges of scaling a self-hosted text-to-speech (TTS) endpoint on Amazon SageMaker, particularly when dealing with unexpected traffic spikes. The author notes that the auto scaling policy, which is supposed to prevent overloading, relies on guesses about the target value. The post introduces Kokoro-82M, an 82-million-parameter TTS model, which is deploy...
Rescale's AI-driven platform accelerates innovation in engineering by integrating HPC, modeling, simulation, and AI. They used AI-DLC on AWS to build an intelligent triage agent, reducing operational overhead and speeding up deployment.
Amazon SageMaker AI introduces training session APIs, allowing users to drive GPU-backed sessions from their own Python process. Users can call operations like forwardbackward() and optimstep() to customize their training loops without managing infrastructure, enabling quick iteration and debugging.
NVIDIA's Cosmos platform, a leading open models platform, has been expanded with new training tools, simulation checkpoints, and Data Factory blueprints for building world models. Cosmos Data Factory brings together data curation, processing, and synthetic data generation tools, enabling multimodal and multisensory data curation at scale. This includes new features in Cosmos Curator for curatin...
In this text, the author describes their experience using random forest regression, specifically with Extra Trees learners, compared to the standard architecture. The goal is to predict a single numeric value using various regression techniques, with tree-based methods being one category. Random forest regression uses multiple decision trees with different subsets of the data, averaging the res...
IskayPet, a leading pet care company in Iberia with operations in Spain, Portugal, and France, faced a challenge in efficiently answering operational questions across retail, veterinary care, animal care, procurement, and people and onboarding, spread across hundreds of Microsoft SharePoint documents in four languages. The time-consuming process of manually searching and verifying information l...
This article discusses a solution to improve the response time of AI voice agents by closing the gap between generation and synthesis of text-to-speech (TTS) responses. Instead of waiting for the full response to be generated before starting synthesis, each sentence generated by an agent is immediately streamed to a hosted TTS model on Amazon SageMaker AI, allowing audio playback to begin befor...