The next generation of Alexa is here, Alexa+

Hello everyone! Here’s what you need to know about AI today:

👉 Amazon revealed Alexa+ with lots of Gen AI features

👉 A new powerful open-source Agentic AI framework is released by Stanford

👉 Microsoft released new high-performance compact models 

and many more!

📧 Did someone forward you this email? Subscribe here for free to get the latest AI news everyday!

Read time: 5 minutes

What’s going on: Amazon has revealed Alexa+, a new generation of its popular voice assistant. This iteration harnesses the power of generative AI to transform Alexa into a more conversational, intelligent, and capable companion. Alexa+ offers enhanced natural language processing, personalized interactions, and what Amazon calls "agentic abilities", the capacity to take proactive actions on behalf of users.

What does it mean: Amazon is trying to breath life into its aging personal assistant. The upgraded Alexa+ introduces a suite of sophisticated features aimed at making daily life easier and more engaging. It can analyze documents and emails, extracting key information from uploaded files like recipes or study materials to provide summaries or even generate calendar events. Beyond these practical tools, Alexa+ brings a playful and creative side, capable of composing songs or crafting stories on demand.

More details: 

• The rollout will begin in the United States in the coming weeks, initially targeting select Echo Show devices such as the Echo Show 8, 10, 15, and 21, with broader compatibility expected to follow.

• For families, the assistant offers new interactive experiences through Amazon Kids+, such as “Stories with Alexa” and “Explore with Alexa,” which encourage children to dive deeper into topics of interest.

• Fire TV users will enjoy AI-driven content recommendations tailored to their preferences, complete with descriptive titles to streamline the search for something to watch.

• Amazon is using powerful large language models from Amazon Bedrock, which enable Alexa+ to orchestrate actions across tens of thousands of services and devices.

• Alexa+ is available free to Amazon Prime members with a $19.99 monthly subscription for non-Prime users.

What’s going on: Stanford University researchers have unveiled OctoTools, an open-source framework designed to enhance the reasoning capabilities of large language models (LLMs) without requiring additional training. This platform tackles the persistent challenges LLMs face with complex, multi-step reasoning tasks that demand logical decomposition or specialized knowledge. OctoTools allows LLMs to break down intricate problems into manageable subunits, outsourcing specific tasks to tools like calculators, code interpreters, and search engines.

What does it mean: Unlike traditional LLMs, which often need extensive training or curated datasets to adapt to new tools, OctoTools leverages "tool cards" which are standardized wrappers for tools such as Python interpreters and web-search APIs. When a prompt is introduced, OctoTools employs a planner module to devise a high-level strategy, identifying sub-goals and relevant tools. This plan is then refined and executed through a series of coordinated steps, with results verified and summarized to ensure accuracy and coherence.

More details:

• OctoTools surpasses classic prompting techniques and other LLM frameworks like Microsoft’s AutoGen, LangChain, and OpenAI’s function-calling API, achieving accuracy gains of 10.6% over AutoGen, 7.5% over GPT-Functions, and 7.3% over LangChain on benchmarks spanning visual, mathematical, scientific, and medical reasoning tasks.

• The framework provides a diverse set of tools to handle different types of tasks such as Image Captioner, Text Detector, Wikipedia Search, Object Detector, Google Search, URL Extractor, PubMed/Arxiv Paper Search, Pathology Classifier, and many other tools.

• Interested? Check out their GitHub.

🗣 ElevenLabs, known for its text-to-speech AI technology, has launched Scribe, a standalone speech-to-text model supporting over 99 languages, to compete with other speech detection services.

🚀🌕 Lonestar and Phison have launched a data center infrastructure to the moon on a SpaceX rocket, marking the first lunar data center with plans to expand it to a petabyte of storage, offering customers off-Earth data storage immune to disasters and hacking.

🖥 Framework, the company that builds modular and repairable laptops, has launched its first desktop PC, a small form-factor computer designed for gaming and local AI inference, powered by AMD's Ryzen AI Max processors and up to 128GB of RAM.

🤖 Nvidia CEO stated that next-generation AI models will require 100 times more computational power than previous iterations due to advanced reasoning methodologies.

📈 Nvidia reported a record fourth-quarter revenue of $39.3 billion, up 78% year-over-year, driven by strong demand for its AI-powered Blackwell chips, with data center sales reaching $35.6 billion.

💬 Microsoft has released new Phi-4 AI models, including Phi-4-mini and Phi-4-multimodal, which offer significant performance with less parameters. Phi-4-mini, with 3.8 billion parameters, excels in text-based tasks like math and coding, while Phi-4-multimodal, with 5.6 billion parameters, processes text, images, audio, and video.

AI + Personal finance

SoFi - Senior Data Scientist, Lending

Cleerly - Staff/Senior ML Scientist

Roblox - Senior Data Scientist - Ecosystem and Learning Platform

Invisible Technologies - Senior Machine Learning Engineer

Thank you for staying with us like always! If you are not subscribed, subscribe here for free to get more of these emails in your inbox! Cheers!