These findings validate the notion that making models reason in a much more representative latent space is better than making them talk to themselves via [Chain of Thought] Your Next ‘Large’ Language Model Might Not Be Large After All | Towards Data Science
Towards Data Science
Your Next ‘Large’ Language Model Might Not Be Large After All | Towards Data Science
A 27M-parameter model just outperformed giants like DeepSeek R1, o3-mini, and Claude 3.7 on reasoning tasks
Fascinating. Thanks for this. *Don’t get hypnotized by the headline. Study the architecture of the story. This is how modern propaganda works: → Emotion first → Evidence later (if ever) → Amplify through influencers → Polarize the population → Fracture alliances → Profit* View quoted note →
The 33-year-old artist told reporters in Mexico City that "in an age that seems not to be the age of faith or certainty or truth, there is more need than ever for a faith, or a certainty, or a truth" https://www.perplexity.ai/page/rosalia-s-spiritual-lux-earns-.DTT16zPRym0Df7oC6IQRw
*making bullshit is cheap. verifying what is true, is expensive* View quoted note →
You may not become rich through Bitcoin but it will connect you to people who oppose fiat money specifically and injustice generally. This connection is valuable.
*Magnification in chaotic systems is accomplished by focusing on small changes in the system (such as when Lorenz changed the number of decimal places in the data) which caused drastic reactions. When magnified, chaotic systems display erratic behaviour. Under further scrutiny, patterns can be seen in this chaos. Upon even further magnification, however, more erratic behaviour is uncovered, and beneath that more "fine geometrical structure* View quoted note →
*Knowledge production has become cheaper than knowledge verification. It's a real problem*
Stacker News
An MIT Student Awed Top Economists With His AI Study—Then It All Fell Apart \ stacker news ~econ
Apparently an MIT student fabricated data on an AI study that got a lot of traction and even made its way to being cited by Congress. The fabricati...
*Like many things in modern life, the edifice of academia is built on norms and cultural standards that no longer obtain. The body is dead but the corpse isn't entirely cold yet*
Stacker News
An MIT Student Awed Top Economists With His AI Study—Then It All Fell Apart \ stacker news ~econ
Apparently an MIT student fabricated data on an AI study that got a lot of traction and even made its way to being cited by Congress. The fabricati...
*as new coin production keeps shrinking in the coming yrs, [Bitcoin’s] barometer only gets sharper at measuring global liquidity* View quoted note →
*The “fiat influencer” …depends on an audience that remains docile and constantly overstimulated, and structurally incapable of long-form thought. Bitcoin, especially in its cypherpunk instantiation, demands intentionality. It potentially collapses the economics of distraction. So why would an influencer promote a tool that erodes the very psychological environment his livelihood depends on?* View quoted note →