At this point I feel like we understand pretty well what's going on with LLMs: - Outputs are roughly equivalent to kernel smoothing over positional embeddings (arxiv.org/pdf/1908.11775…) - The learned computation model is *probably* bounded by RASP-L (arxiv.org/pdf/2310.16028…) -

How to be as "smart" as Auto-Regressive LLMs: - memorize lots of problem statements together with recipes on how to solve them. - to solve a new problem, retrieve the recipe whose problem statement superficially matches the new problem. - apply the recipe blindly and declare

I’ve just been just told by the staff at Pret A Manger that there is no water in espresso 🙈

honestly, Word, we have enough CPU to keep the table of contents updating automatically

Yann LeCun says he is working to develop an entirely new generation of AI systems that he hopes will power machines with human-level intelligence. It could take up to 10 years to achieve, he tells the Financial Times in an interview on.ft.com/3KbShLF

I'm playing with G-Eval to test the LLM outputs using LLM. It roughly works until it doesn't. How am I supposed to reason with test result: "the actual output's prompt is in Polish which mismatches the language-prompt specified as Polish, aligning correctly" #llm #gpt #deepeval

Convolutional Neural Networks in action

Next time you estimate development think about leaky pipelines: hiandrewquinn.github.io/til-site/posts…

Intellectual obesity - I love the term: gurwinder.blog/p/the-intellec…

We live in the future; it's just not distributed evenly: hakaimagazine.com/videos-visuals…

Long awaited - standardised way to run web app in lambda github.com/awslabs/aws-la…

White-Collar Work Is Just Meetings Now - The Atlantic theatlantic.com/ideas/archive/…

As Apple Intelligence is rolling out to our beta users today, we are proud to present a technical report on our Foundation Language Models that power these features on devices and cloud: machinelearning.apple.com/research/apple…. 🧵

Swift.org - Announcing Swift Homomorphic Encryption swift.org/blog/announcin…

Ruby on Jupyter nbviewer.org/github/SciRuby…

Explains why I found myself forced to not just block Musk, but also mute the terms “Elon”, “Musk”, “Elonmusk” to get a Twitter experience where I wouldn’t have every second tweet of his on my timeline. Case study worthy on how you degrade a social network long-term

Nice, I might be eventually able to use letter “m” in passwords for some, ancient services. But then again if they are already ancient, will their CISO actually care about the new NIST guidance? mastodon.social/@LukaszOlejnik…

We all love short papers fermatslibrary.com/s/the-unsucces…

I loved gevent. It was bringing all the benefits of event loop I needed and leaving me with a straightforward API on monkey patched threads. I never could understand why it was treated as an ugly child

Bookmarking wikenigma.org.uk/start