Suggestions
Pedro Alves
Data Science Leader for Microsoft IDEAs - HIRING Machine Learning Scientists
Pedro Alves, who uses the LinkedIn username pedroalvesds, is an advisor at CannaGo and has a diverse background in data science and machine learning. Here are some key points about him:
-
Current Roles:
- Advisor at CannaGo since November 2020.
- CEO and Managing Partner at Moirai Capital since July 2024.
- Technical Advisor at Aktana since June 2015.
-
Past Roles:
- Partner Data Science Manager for Microsoft’s IDEAs group from October 2021 to December 2022.
- Founder and CEO of Ople.ai from January 2017 to September 2021.
- Director of Data Science at Sentient Technologies from April 2016 to January 2017.
- Chief Data Scientist at Banjo from May 2014 to April 2016.
-
Education:
- Bachelor of Science in Computer Science from the University of Notre Dame.
- Master of Science in Bioinformatics from Indiana University Bloomington.
- Ph.D. in Computational Biology (not completed) from Yale University.
-
Expertise: Pedro Alves is highly skilled in leveraging data science and machine learning to drive organizational success. His work spans various fields, including computational biology, gene networks, and AI-enabled platforms.12
Highlights
Many industries are still too far behind in the adoption of A.I.
Even if development of newer and better LLM's stopped we would still have at least 5 more years of massive advancements in A.I. via its adoption by various industries.
Here is the full article.
The Airplane Has Already Been Invented: Unlocking Years of Value from Today's LLM's https://t.co/Akm7ufMRxD
Large language models are genuinely impressive. They write coherent essays, answer complex questions, and power tools that millions of us use every day. They are already transforming how we work. Yet they remain fundamentally statistical predictors: they generate one token at a time, committing immediately to the most probable continuation. There is no internal stage where the model first reasons and only then speaks. Ask an LLM “engineer or doctor?” It often begins “You should become a … and is then gently pushed toward “doctor” because that phrase is vastly more common in training data than “an engineer.” The choice is shaped by corpus frequency and grammar, not by any actual weighing of trade-offs. Ask it to verify 17 × 24 (408 or 398). Many models will declare “After careful checking, it is definitely 408” before performing the multiplication, because confident-sounding text is more frequent in their training data than cautious accuracy. All current advances — chain-of-thought, retrieval, tool use, test-time compute, mixture-of-experts — are valuable scaffolding that make the models far more useful, but they do not give the architecture native reasoning ability. We are making the finest, most capable horse carriage the world has ever seen. We have not yet invented the automobile. Current LLMs already deliver enormous practical value and will continue to do so. True general intelligence, however, will require a new paradigm that can build genuine world models, run real counterfactuals, and detect its own uncertainty from the ground up. We are not one more scaling run away. We are one architectural revolution away. https://t.co/Tu3rzk5Hpm via @LinkedIn

