| Hi there, Welcome to the 113th edition of Heartcore Insights, curated with 🖤 by the Heartcore Team. If you missed the past newsletters, you can catch up here. Now, let’s dive in!
The advent of LLMs is causing a fundamental shift in how data is processed and understood: “One huge implication of LLMs transforming data processing into semantic processing is the transformation of Extract-Transform-Load (ETL) data processes to Extract-Contextualize-Load (ECL) semantic processes.” “[...] Contextualization refers to the process of adding the right type and amount of information at the right point of time, so as to help structure a conversation in a specific direction, and at a specific depth”
Prior to LLMs, there’s been an industry and a range of jobs to be created around ETL and ELT processes that are not becoming any less important. The advent of LLMs are however creating a new set of requirements in data engineering: LLMs as the form function for data interaction, using natural language to interact with unstructured natural language documents Data is unstructured — we’re no longer working with well-defined data, LLMs need structure on unstructured prose to navigate words in a meaningful manner There is much more data to work with, especially now that unstructured data is in play, and this new under-tapped data in focus is qualitative and mostly textual.
Information retrieval is a key problem to solve in order to unlock enterprise adoption of LLMs and new use cases that may flow downstream from there and has created a need for new infrastructure to build on. The author believes the key for better retrieval is context: Improved Retrieval Accuracy: Enhancing the retrieval mechanism to more accurately identify and fetch relevant pieces of information from the knowledge base. Incorporation of Expert Knowledge as deterministic rules: Leveraging feedback from domain experts to define or refine rules and heuristics for more effective chunk and information retrieval and understanding. Memory & Personalization: Incorporating memory and personalization into contextual optimization involves leveraging user-specific data and interactions to refine and adapt information retrieval [...]
The past paradigm of ETL and ELT is a useful analogy to illustrate how ECL and ELC data processes will look. ETL becomes ECL and ELT becomes ELC, in essence: “An example of an ECL process would be the use of document hierarchies. In the document hierarchies process, data is extracted from the underlying documents and chunked, the data is contextualized into a hierarchical layer of metadata, and stored in a knowledge graph.” “The ELC process: [The] main use case here is in recursive retrieval, whereupon core concepts and ideas are fixed within a knowledge graph and information from various pages and documents are iteratively fed into the knowledge graph over time.
ETL (ECL) workflows: ELT (ELC) workflows: Letter to the shareholders of Berkshire Hathaway and a tribute to Charlie T. Munger, by Warren Buffett Charlie Munger died on November 28, just 33 days before his 100th birthday. “[...] he told me – correctly! – that I had made a dumb decision in buying control of Berkshire. But, he assured me, since I had already made the move, he would tell me how to correct my mistake.” “In what I next relate, bear in mind that Charlie and his family did not have a dime invested in the small investing partnership that I was then managing and whose money I had used for the Berkshire purchase” Charlie, in 1965, promptly advised me: “Warren, forget about ever buying another company like Berkshire. But now that you control Berkshire, add to it wonderful businesses purchased at fair prices and give up buying fair businesses at wonderful prices. In other words, abandon everything you learned from your hero, Ben Graham. It works but only when practiced at a small scale.” “Charlie never sought to take credit for his role as creator but instead let me take the bows and receive the accolades. In a way his relationship with me was part older brother, part loving father. Even when he knew he was right, he gave me the reins, and when I blundered he never – never –reminded me of my mistake.” I (Warren) have long been in charge of the construction crew; Charlie should forever be credited with being the architect.
Cost a Capital and Capital Allocation, Michael Mauboussin Reddit and Scaling Laws, Gad Allon Greenlea Lane on Conviction and Quality, Josh Tarasoff On investment theses as lagging indicators, David Peterson On Venture Research, Brett Bivens Investing at the edge of large markets under transformative pressure, Nick Grossman On VC returns and portfolio construction, OpenLP State of the Venture market, Redpoint Why ESG investing might never recover, WSJ A Modern take on the “acquihire”, Inflection’s announcement of the recent dealings with MSFT
Enapi, a Germany-based provider of a connectivity platform for collaboration in the electric vehicle (EV) charging industry, raised €2.5M with Project A/Seedcamp - link BLIXT, a Sweden-based startup developing semiconductor-based circuit breakers and household fuses raises €5M with USV/ERV - link Packfleet, a UK-based tech-enabled carbon-neutral courier, raises £8M with General Catalyst/Voyager - link Embat, a Spanish-based fintech software for real-time treasury management, raises $16M with Creandum/Samaipata/4Founders - link Photoroom, a France-based mobile AI photo-editing software, raises €40M with Balderton/Aglaé - link RobCo, a Germany-based startup offering affordable and connected robotics automation solutions for SMEs, raises €39M with Lightspeed/Sequoia - link
Quilt, a US-based startup building an AI assistant for solutions teams, raises $2.5M with Sequoia - link DraftWise, a NYC-based provider of an AI-powered contract and negotiation platform for lawyers, raised $20M with Index/Earlybird - link Cognition AI, a US-based startup building a software development assistant, raises $21M with Founders Fund/Elad Gil - link Viam, a US-based software platform to develop & monitor smart hardware, raises $45M with USV/Battery - link The Browser Company, a US-based maker of a new generation of internet browser, raises $50M with Pace Capital - link
Foundry, a US-based orchestration platform designed to simplify the utilization of AI computing resources., raises $80M with Sequoia/Lightspeed - link Together AI, a US-based platform designed for developing and deploying open-source generative AI and models, raises $106M with Salesforce/Coatue/Lux - link Mews, a Netherlands-based hotel property management system, raises $110M with Kinnevik/Revaia/Goldman Sachs - link Glean, a US-based AI-powered work assistant that connects and understands enterprise knowledge, raises $200M with Kleiner/Lightspeed/Citi Ventures - link Monzo, a UK-based digital-only bank platform, raises $430M with CapitalG/GV/HongShan Capital - link
Welcome to Laerke Rud Hansen joining the Heartcore team as an investment principal in Copenhagen! She’s joined in Copenhagen by Anne Klussendorf who’s strengthening our finance team We’ve been active with new investments (however none of them yet announced) and we’re excited to have partnered with founders in categories across our core investment themes: Reimagining Work Protecting the Planet Empowering People
Stay tuned as there’ll be much more to come also in terms of new portfolio additions and funding rounds!
Thank you for being a loyal subscriber of our monthly Insights - please feel free to share this newsletter with anyone you’d think would appreciate it! Share Heartcore Insights | |
No comments:
Post a Comment