Carlos E. Perez @IntuitMachine

968 Followers
1.1K Following
4.7K Posts
FOLLOWS YOU The Artificial Intuition, Fluency & Empathy guy. AGI via Peirce's triadic inference.
Bloghttp://medium.com/intuitionmachine
Twitterhttps://twitter.com/IntuitMachine
Bookhttp://gum.co/empathy
Bookhttp://deeplearningplaybook.com
An ontology for document parsing
Ontology for Inference-Time Algorithms for LLMs
Yup! Everyone, including Snoop Dogg, is freaking out!
---
RT @pkedrosky
And now, the collected wisdom of Snoop Dogg on AI and existential risk: "Like, what the f**k?"
https://twitter.com/pkedrosky/status/1653955254181068801
Paul Kedrosky on Twitter

“And now, the collected wisdom of Snoop Dogg on AI and existential risk: "Like, what the f**k?"”

Twitter
Both "the extended mind" and "the medium is the message" are tightly related ideas.
---
RT @Meaningness
https://metarationality.com/rationality
https://twitter.com/Meaningness/status/1654528168492158978
Part Three: Taking rationality seriously | Meta-rationality

A pragmatic understanding of how systematic rationality works in practice can help you level up your technical work.

Meta-rationality
Part Three: Taking rationality seriously | Meta-rationality

A pragmatic understanding of how systematic rationality works in practice can help you level up your technical work.

Meta-rationality
Excellent summary of alternative kinds of prompt engineering.
---
RT @cwolferesearch
Prompt engineering for language models usually involves tweaking the wording or structure of a prompt. But, recent research has explored automated prompt engineering via continuous updates (e.g., via SGD) to a prompt’s embedding. Here’s how these techniques work… 🧵 [1/8]
https://twitter.com/cwolferesearch/status/1654251652956712961
Cameron R. Wolfe on Twitter

“Prompt engineering for language models usually involves tweaking the wording or structure of a prompt. But, recent research has explored automated prompt engineering via continuous updates (e.g., via SGD) to a prompt’s embedding. Here’s how these techniques work… 🧵 [1/8]”

Twitter
RT @cwolferesearch
AutoPrompt combines the original prompt input with a set of shared “trigger token” embeddings that are selected/trained to improve model performance by using gradient descent. Trigger tokens are shared across all inputs provided to the language model. [3/8]
RT @divbyzero
Cool proof and cool example of using CSS to create mathematical animations on web pages! Here's the page with the math https://x.st/visual-sum-of-cubes/ and here's the page describing the CSS https://x.st/spinning-diagrams-with-css/.
Visual Sum of Cubes

This article discusses a ‘visual’ derivation of the formula for 1³+2³+…+n³.

exist

RT @heidi_moss
@garymarcus @WiringTheBrain @TonyZador @anne_churchland @Nancy_Kanwisher @PessoaBrain @neurograce Definitely debunked. Here’s a great image from Northcutt, 2002 that shows the evolutionary expansion of regions. The connectome is also good way to demonstrate the interconnectedness, debunking the idea of “the lizard brain made me do it”.

https://www.frontiersin.org/articles/10.3389/fninf.2012.00014/full

Mapping the Connectome: Multi-Level Analysis of Brain Connectivity

Background and scope The brain contains vast numbers of interconnected neurons that constitute anatomical and functional networks. Structural descriptions of neuronal network elements and connections make up the ‘connectome’ of the brain (Hagmann 2005; Sporns et al. 2005; Sporns 2011), and are important for understanding normal brain function and disease-related dysfunction. A long-standing ambition of the neuroscience community has been to achieve complete connectome maps for the human brain as well as the brains of non-human primates, rodents and other species (Bohland et al. 2009; Hagmann et al. 2010; Van Essen and Ugurbil 2012). A wide repertoire of experimental tools is currently available to map neural connectivity at multiple levels, from the tracing of mesoscopic axonal connections and the delineation of white matter tracts (Saleem et al. 2002; Van der Linden et al. 2002; Sporns et al. 2005; Schmahmann et al. 2007; Hagmann et al. 2010), the mapping of neurons organized into functional circuits (Geerling and Loewy 2006; Ohara et al. 2009; Thompson and Swanson 2010; Ugolini 2011), to the identification of cellular level connections and the molecular properties of individual synapses (Harris et al. 2003; Arellano et al. 2007; Staiger et al. 2009; Micheva et al. 2010; Wouterlood et al. 2011). But despite the numerous connectivity studies conducted through many decades we are still far from achieving comprehensive descriptions of the connectome across all these levels. ...

Frontiers
RT @WiringTheBrain
@garymarcus @TonyZador @anne_churchland @Nancy_Kanwisher @PessoaBrain @neurograce @ylecun Paul Cisek has a really thoughtful, comprehensive article on brain evolution through a control system lens: Resynthesizing behavior through phylogenetic refinement https://pubmed.ncbi.nlm.nih.gov/31161495/
Resynthesizing behavior through phylogenetic refinement - PubMed

This article proposes that biologically plausible theories of behavior can be constructed by following a method of "phylogenetic refinement," whereby they are progressively elaborated from simple to complex according to phylogenetic data on the sequence of changes that occurred over the course of ev …

PubMed