The post analyzes the economic implications of AI automation on labor and productivity, highlighting the uncertainty of predictions and the balance between automation and new task creation.
About:
Martin Lumiste is a machine learning enthusiast and data science professional with a passion for impactful applications and enjoys long distance running.
Website:
Specializations:
Interests:
Outgoing Links:
Subscribe to RSS:
OpenAI's charter suggests collaboration over competition in AGI development, yet current trends indicate a race towards ASI, challenging the company's commitment to safety.
The author shares insights on animating old family photos using AI, detailing costs, techniques, and the emotional impact of the final results.
Evidence-based insights on early childcare suggest that while some practices are critical, many parenting decisions are less consequential than commonly believed.
The emerging field of 2D avatar/portrait animation has potential applications in video compression. The LivePortrait model, a beefed up version of facevid2vid, is able to achieve reasonable perceptual quality at an extremely low b...
0Compressing images with neural networks
2024-03-16 •
The text discusses the evolution of image and video compression, highlighting the use of auto-encoder style neural networks for compression. It explains the JPEG standard, the use of discrete cosine transform, and the development ...
0ML research and startups
2023-12-11 •
The text discusses the role of research in machine learning and the differences between research science and research engineering. It highlights that big tech companies dominate research science, while research engineering can be ...
0Quarterly planning is a noisy optimisation problem
2023-12-04 •
The text discusses the problem of optimally solving the quarterly planning cycle in big technology companies, which is equivalent to solving the knapsack problem. It explores the trade-offs between implementation cost and future m...
0Why you should A/B test your LLM evals
2023-08-09 •
The text discusses the problem of evaluating LLM stochasticity and the need for A/B testing to measure the impact of new commits on project performance. It explains the challenges of evaluating LLM outputs and the use of statistic...
The author discusses their experience in the NeurIPS 2022 Traffic4cast competition, where they placed second in the main track and fourth in the extended track. They detail the competition setup, their approach using gradient boos...
0Effective spatial embeddings for tabular data
2022-04-11 •
The text discusses the use of spatial embeddings for tabular data, comparing gradient boosted tree models (GBT) with neural networks. It explains the benefits of deep learning for tabular data and presents different models and the...
0No, NATO is not responsible for the Ukrainian War
2022-03-19 •
The text argues that NATO is not responsible for the Russian invasion of Ukraine, as the eastward expansion of NATO was the free choice of the countries involved. It emphasizes the principle of state sovereignty and the geopolitic...
0Age of exploration startups
2022-03-11 •
The text discusses the Age of Exploration and its impact on the spice trade, focusing on the Portuguese Empire's role in the trade. It explores the economic significance of the spice trade, the revenue generated, and the impact on...
0Best Hardcore History episodes
2022-01-21 •
The author discusses the lack of depth in most podcasts and praises Hardcore History by Dan Carlin for its in-depth storytelling of historical events. The author ranks his favorite episodes and provides a brief overview of each.