Maciej Glowacki

Postdoctoral researcher @CERN | Physics & Applied Machine Learning | Ideas through writing

[Maciej Glowacki @MMGlowacki]

The way we explore the universe is changing.

From:

- Writing a theory
- Designing an experiment
- Analysing data against theory

To:
- Large-scale pattern recognition

[Maciej Glowacki @MMGlowacki]

@Dorialexander all this precious data from European industry and science is currently being freely shipped out to foreign hyper-scalers

[Maciej Glowacki @MMGlowacki]

@linasbeliunas X bought them because they gathered a proprietary dataset of high quality code for training. It was 60B for a dataset

[Maciej Glowacki @MMGlowacki]

@CryptoTice_ this is the only way to redistribute wealth after covid money printing and stabilise the economy again, every country should follow

[Maciej Glowacki @MMGlowacki]

@CharlesMullins2 where's the research?

[Maciej Glowacki @MMGlowacki]

@TheVixhal The question is always does the substrate of the computation matter?

[Maciej Glowacki @MMGlowacki]

@championswimmer math, code, film, music are all just a way of decoding the brain latent. The brain latent is where you create your understanding, which can be agnostic to the final projection

[Maciej Glowacki @MMGlowacki]

@justalexoki I don’t think it’s that big of a leap, matter is made from particles which are themselves representations of symmetries, symmetries are absolutely fundamental. Why? Because symmetries are the most efficient way of compressing data; everything is information follows

[Maciej Glowacki @MMGlowacki]

@leecronin Creativity is synthesis from pulling in disparate sources of information. It’s interpolation rather than extrapolation

[Maciej Glowacki @MMGlowacki]

@Plinz Or forming abstractions that generalise OOD

[Maciej Glowacki @MMGlowacki]

@Gadget440 HS is just daft, he’s talking about the matrix while his whole life is controlled by the TikTok algorithm, but this guy there’s no hope for

[Maciej Glowacki @MMGlowacki]

@VibeMarketer_ Is this platonic reps confirmed?

[Maciej Glowacki @MMGlowacki]

@BeijingDai They took out the leadership in a few hours, wdym decline? The Warefare has just moved on, it’s now data driven intelligence

[Maciej Glowacki @MMGlowacki]

@JuliaEMcCoy Humans also need some pre-training, can’t go straight to OOD generalisation from nothing. But I agree from a certain age things need to change

[Maciej Glowacki @MMGlowacki]

@karankendre Maybe data centres in space was not such a bad idea

[Maciej Glowacki @MMGlowacki]

@chamath maybe the tests are outdated?

[Maciej Glowacki @MMGlowacki]

@WomanDefiner It’s technology, vast streams of data sources processed in real time by AI, it’s now very precise

[Maciej Glowacki @MMGlowacki]

From a purely technical view point, this exhibition of 5th Gen Warefare by the US is seriously impressive. I also see how AI fits into this.. synthesising actionable strategies from disparate data sources in close to real time

[Maciej Glowacki @MMGlowacki]

@levelsio Europe has a certain rhythm that places like the US/Asia don’t. The pattern is Europe invests, the US/Asia commercialises. Eg breakthroughs in Physics, the internet and AI. It’s the natural equilibrium point of the world, if everyone hustles there would be no true breakthroughs

[Maciej Glowacki @MMGlowacki]

@RnaudBertrand Europe has a certain rhythm that places like the US/Asia don’t. The pattern is Europe invests, the US/Asia commercialises. Eg breakthroughs in Physics, the internet and AI. It’s the natural equilibrium point of the world, if everyone hustles there would be no true breakthroughs

[Maciej Glowacki @MMGlowacki]

@fchollet Any task with a verifiable reward will be eaten by in-context RL so whether it’s well represented in th training set becomes less relevant as long as the underlying model has generalisation capability (which it does via pretraining)