FUTURE HOUSE INC

STARTING IN NOVEMBER 2023, FUTUREHOUSE EMBARKED ON A MISSION TO ACCELERATE INNOVATION IN BIOLOGY AND OTHER COMPLEX SCIENCES BY DEVELOPING AI SYSTEMS CAPABLE OF PERFORMING SOPHISTICATED RESEARCH AUTONOMOUSLY. FUTUREHOUSE BELIEVES THAT THE FIRST STEP TOWARD ADVANCING SCIENCE IS THROUGH THE CAREFUL CURATION AND SYNTHESIS OF EXISTING KNOWLEDGE AND EXPERIMENTAL RESULTS. HOWEVER, THE SHEER VOLUME OF SCIENTIFIC LITERATURE MAKES SYNTHESIS A SIGNIFICANT BOTTLENECK. LARGE LANGUAGE MODELS (LLMS) OFFER A PROMISING SOLUTION WITH THEIR CAPACITY TO COMPREHEND AND SUMMARIZE VAST AMOUNTS OF INFORMATION, POTENTIALLY TRANSFORMING SCIENCE BY AUTOMATING KNOWLEDGE SYNTHESIS ON A LARGE SCALE. DESPITE THIS POTENTIAL, CURRENT LLMS FACE LIMITATIONS, SUCH AS GENERATING HALLUCINATIONS, LACKING ACCESS TO THE LATEST DATA, AND PROVIDING UNRELIABLE REFERENCES. TO ADDRESS THESE CHALLENGES, FUTUREHOUSE HAS DEVELOPED TWO POWERFUL AI-DRIVEN TOOLS, WIKICROW AND PAPERQA, WHICH ENHANCE SCIENTIFIC RESEARCH AND KNOWLEDGE SYNTHESIS. WIKICROW AUTOMATES THE CREATION OF WIKIPEDIA-STYLE ARTICLES FOR TECHNICAL SCIENTIFIC TOPICS, ENABLING EFFICIENT DISSEMINATION OF KNOWLEDGE. BUILT ON FUTUREHOUSE'S INTERNAL LLM PLATFORM, PAPERQA, WIKICROW ACHIEVES STATE-OF-THE-ART PERFORMANCE ON TASKS LIKE PUBMEDQA AND LITQAAN INTERNAL RETRIEVAL-FOCUSED BENCHMARK THAT EVALUATES THE ABILITY TO RETRIEVE FULL-TEXT PDFS ACROSS THE SCIENTIFIC LITERATURE. DEMONSTRATING AI'S TRANSFORMATIVE POTENTIAL, WIKICROW HAS GENERATED DRAFT ARTICLES FOR OVER 15,600 HUMAN PROTEIN-CODING GENES LACKING COMPREHENSIVE WIKIPEDIA ENTRIES, COMPLETING THE TASK IN JUST 8 MINUTES PER ARTICLE. WIKICROW'S ACCURACY IN LITERATURE RETRIEVAL SURPASSES THAT OF PHD AND POSTDOC-LEVEL BIOLOGY RESEARCHERS, AS VALIDATED BY LITQA2. THIS TOOL WILL SERVE AS A CORNERSTONE FOR THE AI SCIENTISTS FUTUREHOUSE AIMS TO BUILD, DEMOCRATIZING ACCESS TO SCIENTIFIC KNOWLEDGE. THESE ADVANCEMENTS POSITION FUTUREHOUSE AT THE FOREFRONT OF AI-DRIVEN SCIENTIFIC DISCOVERY, PROVIDING TOOLS THAT SIMPLIFY AND EXPAND THE SYNTHESIS OF COMPLEX INFORMATION. ALIGNED WITH THIS MISSION, FUTUREHOUSE HAS FOCUSED ON BUILDING INTEGRATED TEAMS IN MACHINE LEARNING AND BIOLOGY. TO ATTRACT TOP TALENT, THE ORGANIZATION HOSTS INTERACTIVE WORKSHOPS THAT GIVE CANDIDATES AN AUTHENTIC GLIMPSE INTO FUTUREHOUSE'S INNOVATIVE CULTURE. THESE WORKSHOPS ENGAGE SKILLED, MOTIVATED CANDIDATES, ALLOWING THEM TO SHOWCASE THEIR ABILITIES AND CONNECT MEANINGFULLY WITH THE TEAM. BY FOSTERING REAL-WORLD INTERACTIONS, FUTUREHOUSE IDENTIFIES STRONG CANDIDATES, BUILDS LASTING RELATIONSHIPS, AND INSPIRES FUTURE HIRES WHO ALIGN WITH THE ORGANIZATION'S VALUES AND MISSION. THROUGH THESE INITIATIVES, FUTUREHOUSE IS ADVANCING THE FRONTIER OF AI IN SCIENTIFIC RESEARCH WHILE BUILDING A DEDICATED TEAM READY TO RESHAPE THE LANDSCAPE OF SCIENTIFIC DISCOVERY.