| Hi, I put together a newsletter called Interesting Things that curates 40+ interesting stories and links every week about tech with a splash of business and science. Looking through the past few editions, there has been an average of one AI/deep-learning/neural-network related article per week: Issue #001 Shedding light on fairness in AI with a new data set — Facebook publishes a dataset of of 45,186 human face-to-face videos with an emphasis on fairness and diversity across age, gender, skin tone, & ambient lighting. — https://ai.facebook.com/blog/shedding-light-on-fairness-in-a...) Issue #002 Geometric foundations of Deep Learning — An attempt to unify a broad class of ML problems from the perspectives of symmetry and invariance. — https://towardsdatascience.com/geometric-foundations-of-deep... Latest Neural Nets Solve World’s Hardest Equations Faster Than Ever Before — New approaches to solve entire families of partial differential equations and make it easier to model complicated systems. — https://www.quantamagazine.org/new-neural-networks-solve-har... Issue #004 Superhistory, Not Superintelligence — Simpler algorithms and more data do better than more complex algorithms with less data. What are the implications? — https://breakingsmart.substack.com/p/superhistory-not-superi... Reimagining Databases for Vector Embeddings — How is a vector database different from a conventional database? — https://www.pinecone.io/learn/vector-database/ Issue #005 New system cleans messy data tables automatically — A new machine learning system uses probabilistic programming to clean dirty datasets and tell you how confident its answers are. — https://news.mit.edu/2021/system-cleans-messy-data-tables-au... Language models like GPT-3 could herald a new type of search engine — A new idea from researchers could make web search more like talking to human experts. — https://www.technologyreview.com/2021/05/14/1024918/language... Issue #006 Hacker's guide to deep-learning side-channel attacks: the theory — Learn the concepts behind deep-learning side-channel attacks. Use it to recover AES cryptographic keys. — https://elie.net/blog/security/hacker-guide-to-deep-learning... Archaeologists train a neural network to sort pottery fragments for them — The network turned out to be as good at the job as human archaeologists. — https://arstechnica.com/science/2021/05/archaeologists-train...) (Issue #007 has a very interesting GPT-3 related article.) So ... while my newsletter is not a dedicated AI blog, you might still find relevant AI stuff in there (along with other interesting stuff — I wrote some more about why it matches the interest profile of HN very well at https://news.ycombinator.com/item?id=27343897.) If you're interested, please have a look at https://bengtan.com/interesting-things. |