Peter Boshard Olson @peterbolson Clear explanation from @LangChainAI’s @hwchase17: There are 4 main ways to give an AI app background info: 1. Instruction prompting 2. Few-shot examples 3. RAG 4. Fine-tuning It hadn’t occurred to me that these are all ways of accomplishing the same broad goal. 🧵(1/2) https://pic.twitter.com/9QBuH3xQOR Replying to @peterbolson @LangChainAI and @hwchase17
Read More »

Yi Ma @YiMaTweets  Our white-box transformer paper is over 100 pages. Just believe some truly important idea is better conveyed through wholesale than retail. 8-page conference papers are prone to encourage incremental progresses or fragmented ideas. One needs to see the big picture once a while. Replying to @YiMaTweets Just process your 100 pages and
Read More »

kamalikac @kamalikac  Do representation learning models memorize their training data? To understand this, we propose a new method called Deja vu to measure memorization in these models. #NeuRIPS2023 Replying to @kamalikac Full access to lossless complete source data is critical to accepted and trustworthy AIs. I criticize all AI groups now that refuse to cite
Read More »

Q-CTRL @qctrlHQ  Large-scale #quantum computers will require some form of error correction – a distant prospect. In the meantime, the best way to tame unruly, near-term quantum processors is “error suppression”. We’re proud to bring this technology to @IBM Quantum. https://buff.ly/3TgemhQ Replying to @qctrlHQ and @IBM Groups finally realize that quantum components are just components
Read More »

Lorena Barba @labarba@fosstodon.org @LorenaABarba Quoted in @Nature article “Is AI leading to a reproducibility crisis in science?” by @philipcball, I may sound a bit harsh, but it’s the truth… #SciML #reproducibility https://pic.twitter.com/T0HjaCuMK3 Replying to @LorenaABarba @Nature and @philipcball Not “leading to”. The sciences and much online already use closed methods and “nearly impossible to trace
Read More »

Sasi @freest_man  Naive Bayes Classification assumes each feature is independent of others. Naive Bayes classifiers are probabilistic classifiers that predict based on the probability of an object. They’re based on Bayes’ Theorem and assume that every pair of features being classified is… https://pic.twitter.com/KYIUhRKbmQ Replying to @freest_man Which is why GPTs grab as many nearby tokens
Read More »

Digital EU 🇪🇺 @DigitalEU  We have a message for you, and rest assured, it is not #AI-generated. Welcome to the world’s first-ever rules on #ArtificialIntelligence! ✅ Today’s political agreement on the 🇪🇺 #AIAct will ensure AI develops in a human-centric, transparent & responsible way. Read more ↓ Replying to @DigitalEU “will ensure” –> “will hopefully
Read More »