Digital Platform Policy Highlights - Digest 25
Q4 2023 and Q1 2024: This post delineates the impact of digital policies on the enhancement of AI systems that have generative capabilities. ICYMI, this is Digest no 25! Woot.
This post is part five of a series documenting policy changes and feature improvements introduced by platforms in Q4 2023 and Q1 2024.
TL; DR→ Policy changes to utilize increased computational capabilities as well as advances in AI include:
Google's launching Next-Gen AI Model: Gemini
Amazon launching a LLM-Powered Voice Assistant
Microsoft and Nvidia Introducing Windows AI Studio
Mercedes-Benz Integrating LLM-Powered Voice Assistant
Spotify Partnering with Google Cloud AI
Google's launches Next-Gen AI Model: Gemini
Google CEO Sundar Pichai introduced various versions of Gemini: Nano for Android devices, Pro for AI services, and Ultra for advanced applications. Launching with Bard and Pixel 8 Pro, Gemini Pro will be available through Google Cloud. Compared to OpenAI's GPT-4, Gemini boasts superior performance, especially in multimedia comprehension. Google envisions Gemini's evolution towards enhanced sensory awareness. The model emphasizes efficiency, is trained on Tensor Processing Units, and is designed for faster, cost-effective operation. Google is optimistic that Gemini will beat GPT-4. (link)
Amazon's LLM-Powered Voice Assistant Is Coming
Amazon introduced a new Alexa voice assistant powered by a large language model (LLM) that promises it will enhance conversational abilities and contextual understanding. The update aims to invigorate voice assistant technology and elevate smart home experiences. Some features include improved conversational skills, contextual comprehension, and handling multiple requests simultaneously. Advanced tools like Dynamic Controller and Action Controller allow seamless integration of third-party products. With the implementation of LLM along with 200 plus Home APIs within it, Alexa could easily manage home and give personalized, more humanized conversations with the owner. (link)
Microsoft and Nvidia Introduce Windows AI Studio
Microsoft announced Windows AI Studio, a platform for developers to access and customize AI models on their Windows PCs. It integrates tools from Azure AI Studio and services like Hugging Face, offering a guided setup for model configuration, particularly for small language models (SLMs) such as Phi, Llama 2, and Mistral. Nvidia also introduced updates to TensorRT-LLM, enabling efficient LLM processing on PCs with GeForce RTX 30 and 40 Series GPUs. This aligns with Microsoft's hybrid development pattern, which facilitates AI development across cloud and local devices. (link)
Mercedes-Benz Integrates LLM-Powered Voice Assistant
Mercedes-Benz is enhancing its in-car voice assistant with generative AI. It is implementing LLMs to enable more natural interactions and personalized experiences for drivers—the upgraded MBUX Virtual Assistant, running on the next-generation MB.OS operating system will create a human-like interface. It offers predictive features, such as playing morning news upon entry or joining meetings if the calendar indicates lateness. While Mercedes remains tight-lipped about its AI partner (Google it :) ) the assistant's traits include natural, predictive, personal, and empathetic elements. With the implementation of AI, Mercedes emphasizes safety amid rising in-car distractions, while is catching up with other car companies that have already implemented ChatGPT. (link)
Spotify Partners with Google Cloud AI
Spotify collaborates with Google Cloud's AI tools to enhance content discovery and personalized recommendations for its vast library of podcasts and audiobooks. Leveraging Google Cloud's LLMs (large language models), Spotify aims to analyze metadata and improve search results by augmenting details like titles, hosts, and show notes. The main benefit of implementing LLM models is that it increases the probability of identifying harmful content, which solidifies Spotify’s assault against sensitive and dangerous content. Lastly, Google Cloud's AI tools will refine Spotify's recommendation algorithms and enhance user experience by better understanding patterns behind favorite spoken content.(link)
OpenAI launches platform for custom ChatGPTs
OpenAI is launching a no-code platform for creating custom versions of ChatGPT, catering to specific use cases. This service, available exclusively to ChatGPT Plus subscribers, will be accessible through the GPT Store. It emphasizes utility over human-like personas, unlike competitors like Character.AI and Meta. Custom GPTs will feature integrations like web browsing and DALL-E, with a focus on maintaining privacy and monitoring for inappropriate content. The launch is significant: not only does it increase the value of a ChatGPT Plus subscription, but the platform could encourage developers to use OpenAI’s products by making it easier for their chatbots to be discovered.(link)
Research help from Jennifer Xie, Angelina Wang, John Mai, Marshall Singer, Anantesh Mohapatra and Anna Li (Thanks a ton, folks!)
Thank you for reading Platform Policy Research. If you know someone who like this stuff, please share it with them :)