Mathematician who made sense of the universe's randomness wins Abel Prize • Hacker News Board

Re: Mathematician who made sense of the universe's randomness wins Abel Prize

#81

Post by sobriquet9 » Mon, Apr 01, 2024, 11:15 PM UTC

lukan wrote: post #76

sobriquet9 wrote: post #58

Earlier quoted context omitted.

If history was about making the connections, those connections could be used to make predictions that could be tested.

You can, but the world is still an chaotic place, so making a forcast on human geopolitics, is like making a weather forecast. Likely valid for the next days, but increasingly bad for longer timeframes. But one can make for example the prediction that the war in Ukraine will be going on for a little bit longer.

Weather forecasts can easily be compared to actuals, and distribution of residuals can be calculated. That’s how weather forecast models are quantitatively evaluated. A seven day weather forecast has 95% confidence interval of around plus-minus 10 degrees Fahrenheit, way better than “it’ll be like today”.

Your prediction about the war in Ukraine is not falsifiable.

Re: Mathematician who made sense of the universe's randomness wins Abel Prize

#82

Post by lukan » Tue, Apr 02, 2024, 8:07 AM UTC

sobriquet9 wrote: post #81

lukan wrote: post #76

Earlier quoted context omitted.

You can, but the world is still an chaotic place, so making a forcast on human geopolitics, is like making a weather forecast. Likely valid for the next days, but increasingly bad for longer timeframes. But one can make for example the prediction that the war in Ukraine will be going on for a little bit longer.

Weather forecasts can easily be compared to actuals, and distribution of residuals can be calculated. That’s how weather forecast models are quantitatively evaluated. A seven day weather forecast has 95% confidence interval of around plus-minus 10 degrees Fahrenheit, way better than “it’ll be like today”. Your prediction about the war in Ukraine is not falsifiable.

Look, I am not going to make a detailed geopolitical analysis here, but there are lots of sites and institutes who do, you can check them out. The difference with weather forcast is only, that human analysis of geopolitical events often hope to influence the outcome. But the weather will come whether we think it will be sun or rain.

And the prediction about the war in Ukraine would not be "it'll be like today". The current prediction is, that russia will continue to slowly make ground and move the front lines to the west. There are actually predictions how much and where exactly. I think that gets also falsified in some institutes, but is probably classified.

Re: Mathematician who made sense of the universe's randomness wins Abel Prize

#83

Post by sobriquet9 » Tue, Apr 02, 2024, 9:57 AM UTC

lukan wrote: post #82

sobriquet9 wrote: post #81

Earlier quoted context omitted.

Weather forecasts can easily be compared to actuals, and distribution of residuals can be calculated. That’s how weather forecast models are quantitatively evaluated. A seven day weather forecast has 95% confidence interval of around plus-minus 10 degrees Fahrenheit, way better than “it’ll be like today”. Your prediction about the war in Ukraine is not falsifiable.

Look, I am not going to make a detailed geopolitical analysis here, but there are lots of sites and institutes who do, you can check them out. The difference with weather forcast is only, that human analysis of geopolitical events often hope to influence the outcome. But the weather will come whether we think it will be sun or rain. And the prediction about the war in Ukraine would not be "it'll be like today". The c…

Are you saying there is some classified science also called history that makes quantitative, testable, falsifiable predictions?

Maybe so, but I have no clearance and have not seen it. And it would be completely different from history I studied in school. There wasn’t a single prediction, let alone quantitative prediction with time horizon and confidence interval in any of the history books I’ve read.

Re: Mathematician who made sense of the universe's randomness wins Abel Prize

#84

Post by neon5077 » Tue, Apr 02, 2024, 5:20 PM UTC

haolez wrote: post #65

neon5077 wrote: post #59

Earlier quoted context omitted.

I'm like this, and yes, very much so. The (US) academic system is not set up to accommodate anyone, really. It's designed to get someone just below average through their standardized testing and not much more. If you don't fit that mold you're shit outta luck. You either suffer through it like "normal" people, or you find it utterly intolerable and fail or drop out. Personally, I learn best backwards from everyone el…

That's one area where I think LLMs will shine outside of the hype zone: they could boost a student's ability to make progress without supervision. And I understand that there are great teachers out there and an LLM cannot replace that, but at the same time there are a lot of bad teachers wasting everybody's time.

Yup, I've actually gotten pretty decent results out of an LLM for subjects I'm already pretty familiar with. If we ever figure out the hallucination problem, LLMs could revolutionize education overnight.

I think that almost everyone would benefit enormously from having a focused and dedicated one-on-one tutor. Just imagine if you could call up the leading world expert in any field at any time to ask any question you could possibly have. We as a species would get so much more done.

At least that's what I want AI to be in the next decade or so. A tool to push humans to a much higher potential where we can solve our own problems more effectively. I'm afraid we'll skip that step, though and go straight to worshiping the AI that makes the most paperclips.

Anyway, I'm still experimenting off and on with local LLMs to get me closer to where I want to be. I'm not sure it's much faster to use an LLM and continually verify its output, but it does at least provide structure and guidance for my own self-teaching.