Suggestions
Francis deSouza
CEO/investor/entrepreneur passionate about improving life through breakthrough technology.
Francis deSouza is a visionary entrepreneur known for his passion for leveraging cutting-edge technology to enhance human life.
He has an impressive track record of co-founding successful ventures such as SynthLabs, IMlogic, and Flash Communications, all of which were subsequently acquired by tech giants.
Francis deSouza has held high-profile positions including CEO of Illumina, President at Symantec, and Product Unit Manager at Microsoft, showcasing his extensive experience in leadership roles across top-tier organizations.
He pursued his education at Massachusetts Institute of Technology, where he studied Master of Science and Bachelor's of Science in Electrical Engineering and Computer Science, indicating a strong foundation in technical expertise.
With a diverse background, Francis deSouza has been actively involved in various organizations, serving as an investor, board member, advisor, and co-founder across companies like Moonhub, Deel, 4baseCare, Moonwalk Biosciences, and Essential AI.
His advisory and leadership roles also extend to notable entities such as The Walt Disney Company, Citrix, and SynthLabs, highlighting his broad spectrum of influence and expertise in the tech and business sectors.
Throughout his career, Francis deSouza has demonstrated a keen eye for innovation and strategic partnerships, contributing significantly to the advancements in AI alignment, IT security, real-time collaboration, and enterprise messaging management.
His journey from a researcher at MIT Media Lab to founding and leading multiple successful ventures attests to his entrepreneurial acumen and relentless drive for shaping the future of technology and business.
Highlights
How can an organization ensure that their AI large language models (LLMs) meet legal and brand standards? That’s hard. For example, instructing an LLM to avoid a topic eg "Pink Elephant" can increase the likelihood that it references it - the “Pink Elephant” problem.
This research paper by Louis Castricato, Nathan Lile, Suraj Anand, Hailey Schoelkopf, Siddharth Verma, and Stella Biderman, introduces 2, new, fundamental, ideas. First, it demonstrates a promising algorithm, Direct Principle Feedback (DPF), a novel simplified form of constitution AI, for generating synthetic data that does not rely on a ranking mechanism to generate effective reinforcement learning from AI feedback (RLAIF) data. Second, it highlights the importance of allowing users to decide their own "pink elephant" at inference time, giving them autonomy to decide what an aligned/safe model means for them.
Using these methodologies, a 13B parameter language model (a fine-tune of llama-2-13b) performs nearly on par with GPT-4 on a set of pink elephant related evaluation tasks.
Read the paper here: https://lnkd.in/g6eNZysE
#ai #artificialintelligence #rlhf #rlaif EleutherAI Brown University SynthLabs
Illumina’s groundbreaking PrimateAI-3D algorithm (trained on 233 primate species) scans ~70 million variants helping clinicians/researchers predict disease risk and drug targets, with high accuracy and without ancestry bias. Congrats Alex Aravanis MD PhD, Kyle Kai-How Farh MD PhD, and the whole team! #AI #genomics
We can’t wait to see how clinicians like Dr. Stephen Kingsmore harness this tool to unlock even more answers for patients.
#AI #genomics #deeplearning
