Suggestions
Nicolas Machado
Cofounder @ Lume (YC W23) | Stanford AI | Forbes 30u30
Nicolas Machado is a tech entrepreneur and co-founder of Lume, a Y Combinator-backed startup (YC W23).56 He has a strong educational background, holding both a Bachelor's and Master's degree in Computer Science from Stanford University, with specializations in Artificial Intelligence and Human-Computer Interaction.1 Machado has been recognized in Forbes 30 Under 30.56
Professional Experience
Machado's work focuses on using AI to help companies automatically map data.1 At Lume, he has been involved in developing features such as user permissioning.5 He also has experience in investing in and building tech companies, as well as digitizing offline businesses.4
Education
- Stanford University: Bachelor's degree in Computer Science (3.9/4.0 GPA), with a specialization in Artificial Intelligence (2017-2022)1
- Stanford University: Master's degree in Computer Science (4.1/4.0 GPA), with a specialization in Human-Computer Interaction (2020-2022)1
- Sorbonne Université: Exchange program (2019), studying history and art history in French1
Skills and Interests
Machado is multilingual, with native or bilingual proficiency in English and Portuguese, professional working proficiency in Spanish, and limited working proficiency in French.1 His interests include startups, founder life, tech, and AI, as evidenced by his LinkedIn posts.267
Recent Activities
Machado regularly shares insights about the tech industry, startup life, and AI on his LinkedIn profile. He has discussed topics such as data mapping challenges for customers3, the rollout of new features for Lume5, and developments in generative AI.7
Highlights
OpenAI released a model with a different paradigm - it now spends more time thinking before it responds. What does this mean?
In short, it can produce a long internal chain of thought before responding. This surely means larger token requirements & higher latency. But, smarter responses.
Excited about this direction OpenAI is thinking to true rationale and thinking.
This stat was crazy:
OpenAI o1 "exceeds human PhD-level accuracy on a benchmark of physics, biology, and chemistry problems (GPQA)"
A few implications I'm thinking about
- Long term, are the most intelligent models going to be much more latent or expensive? Will there be a flattening of the curve of price drops at this expense?
- are we closer to AGI than we think?
- can it tell how many "s" letters are in the word strawberry?
If you don't know what the strawberry is, ask in the comments, it's super interesting.
#OpenAI #AI #LLM
🧠 New research claims LLMs generate more novel ideas than human researchers? Here's the findings:
Stanford researchers compared research ideas generated by large language models (LLMs) to those from human experts in NLP.
Key takeaways: 💡 AI-generated ideas were judged more novel and exciting than those from human experts. ✅ No significant difference in the feasibility of ideas from both groups. 👩🔬 AI ideas were reviewed by 100+ experts from 36 institutions, ensuring robust evaluation.
However, some limitations emerged: 🔄 Only 5% of AI ideas were truly unique, hinting at a lack of diversity. ⚖️ AI evaluators still lag behind human reviewers in accuracy.
Interesting to see papers coming out evaluating if LLM can truly be creative - this is a debate I hear around the dinner table with friends often - and will likely continue for a long time.
