23.12.22

Surge in number of people in hospital with nutrient deficiencies, NHS figures show

Exclusive: number of people in England and Wales admitted to hospital with deficiencies triples in 10 years

https://www.theguardian.com/uk-news/2023/dec/21/surge-in-number-of-people-in-hospital-with-nutrient-deficiencies-nhs-figures-show

Desalination system could produce freshwater that is cheaper than tap water

MIT engineers and collaborators developed a solar-powered device that avoids salt-clogging issues of other designs.

Sénat - Accueil

Le Sénat est l’une des deux chambres qui forment le Parlement. Il vote la loi, contrôle le Gouvernement et représente les collectivités territoriales.

https://www.senat.fr/

4.

◼️

ARTISTMaggi Hambling born 1945

MEDIUMOil paint on canvas

DIMENSIONSSupport: 2135 × 1827 mm

COLLECTIONTate

ACQUISITIONPresented by the artist in honour of Maria Balshaw 2018

REFERENCET15037

 

Meta-Learning: Why it’s a big deal, it’s future for foundation models, and how to improve it

How Meta Learning might be the future for creating better foundation models

https://machine-learning-made-simple.medium.com/meta-learning-why-its-a-big-deal-it-s-future-for-foundation-models-and-how-to-improve-it-c70b8be2931b

• Source

Sam Altman on AGI: Scaling large language models is not enough

Large language models alone are unlikely to lead to AGI, says Sam Altman. Finding the solution is the next challenge for OpenAI.

https://the-decoder.com/sam-altman-on-agi-scaling-large-language-models-is-not-enough/

• Sam Altman, CEO of OpenAI, believes that large language models alone are unlikely to be enough to create artificial general intelligence (AI) and that further breakthroughs are needed.

• Altman emphasized the importance of AI safety and the need to align AI with human values by using testing, external experts, safety monitoring systems, and best practices to ensure safety.

• He also discussed the need for global cooperation in the regulation of AI, suggesting that a global regulatory body similar to that for nuclear energy may be needed in the future…

Population-Based Meta Learning (PBML)

Population-Based Evolution Optimizes a Meta-Learning Objective

Meta-learning models, or models that learn to learn, have been a long-desired target for their ability to quickly solve new tasks. Traditional meta-learning methods can require expensive inner and...

https://arxiv.org/abs/2103.06435

https://arxiv.org/pdf/2103.06435.pdf

Algorithm 1: Population-Based Meta Learning

Define an initial set of genomes G. Each genome contains parameters, as well as a population count. 
Define a fitness function F(), which evaluates the fitness of a genome.; 
Define a stochastic mutation function M(), which creates offspring by mutating the parameters of a genome.; 
Define decay ratio D; the ratio of population that dies every generation.; 
Define offspring count C, the total number of offspring produced every generation.; 
Define ranking function Rank(), which outputs a genome’s floating-point rank between 0 and 1 based on fitness.; 
for timestep=0...N do 
    for g in G do 
        g.fitness = F(g.parameters); 
        g.rank = Rank(g.fitness); 
        g.population *= (1-D); 
        g.population *= g.rank; 
        g.population /= total_population; 
        for child in int(g.population * C) do 
            G.append(M(g)); 
        end 
    end 
end

类notion开源项目

penx

penx-labs • Updated Jan 8, 2025

 

3D-GPT: 3D MODELING WITH LARGE LANGUAGE MODELS

https://chuny1.github.io/3DGPT/3dgpt.html