#151 - Copilot Pro, LLama.cpp, conversational diagnostic AI, secret AI diplomacy
Topics covered
Popular Clips
Episode Highlights
Copilot Pro
Microsoft's Copilot Pro is transforming Office applications by integrating AI-powered features, available through a $20 monthly subscription. This service offers priority access to the latest OpenAI models and allows users to build their own Copilot GPT, although some features like generating PowerPoint decks from Word documents remain exclusive to business users 1. and discuss the consumerization of AI, noting that expectations are rising as tools become more accessible.
Our expectations are just getting higher and higher.
---
Additionally, Microsoft Copilot now uses GPT-4 Turbo, offering a cost-effective alternative to the previously paywalled model, highlighting a strategic move to enhance accessibility and competitiveness in the AI market 2.
Dataset Growth
EleutherAI's new dataset, Pile V2, marks a significant expansion in AI training resources, aiming to improve upon its predecessor by incorporating more books and public domain data. This update addresses copyright concerns and enhances the dataset's utility for training open-source models 3. Jeremy Harris
The pile is meant to be an open source dataset that anybody can use to train their AI models.
---
In parallel, the development of tools like LLama.cpp, which enables running large language models on personal devices, showcases the growing trend of making AI more accessible and efficient through innovations like four-bit integer quantization 4.
Drug Discovery
DeepMind's spinoff, Isomorphic Labs, is revolutionizing drug discovery by leveraging AI to cut development times from five years to two. This initiative, powered by DeepMind's AlphaFold, aims to streamline the identification of potential new drugs before clinical trials 5. highlights the strategic partnerships with pharmaceutical giants Eli Lilly and Novartis, which are expected to bring significant funding and advancements.
We could probably sign up a dozen partnerships today if we wanted to, but then it would cause us to fragment too much.
---
These collaborations reflect a broader trend in the tech industry towards profitability and efficiency, as companies like DeepMind transition from research-focused entities to profit-driven enterprises 6.