Cool perspective on statistical learning!

Step away from the keyboard, now!

There is more poverty in rural areas and fewer government services.

Interesting summary of the costs and benefits of rural residents. www.ers.usda.gov/amber-waves/...

International AI Safety Report

Bengio chairs the international AI safety report. internationalaisafetyreport.org. It is a good faith effort to assess the whole spectrum of AI risks. (I served as one of the external advisors.)

I do wonder what Congress would do if it actually passed restrictions on Trump’s war. Do they really want to share responsibility for this disaster?

In 2021, an Oregon State robot set the 5k legged robot record. It required a battery change. This new result is a huge accomplishment in battery capacity and locomotion efficiency

there is a lot of noise

but most of it is FEAR

and you cannot let let the fear of other people override the facts in front of you

no one knows the future and nothing is promised. But the better, the more skilled, the more knowledgeable you are about what is new

that earns self determination

Your car has wheels.

Doing it with a legged robot is a major advance, particularly from an energy-efficiency perspective. Just a few years ago, completing a 5k required swapping batteries.

Right; forests are only part of a broader strategy

One amazing water technology is forests. They store and release water like reservoirs, and they sequester carbon at the same time.

Wild! I’d never heard of this.

It is not easy to figure out all of the options. Thanks

This is absolutely terrible. Google, how far you have fallen! What's next? Fingerprints? DNA?

A lovely place!

Why? I figure any conversations moving off X is good. I'd prefer that they come here, of course...

When using LLMs for literature search, I don't think I'm falling for a scam or behaving immorally. In fact, as a computer scientist, I'm grateful that these LLMs have been trained on all of arXiv. They are certainly better than the Science Citation Index of yesteryear

I like the ability to give a complex natural language description of my search query to google/bing rather than having to figure out what keywords to use. It can often uncover terminology and references I would not have found otherwise. (But I do my own writing and thinking, of course.)

I had the same question. Surely they can collect training data for this. Is there a weakness in their architecture that prevents it from learning in these situations? This has been going on for months now (since 9/25)!

The percentage of emergency department visits in the US for Covid-19 has fallen to its lowest level since March 2020.

1/

This is wonderful! I hope the trend continues.

As an older male, I got my Covid-19 booster shot today. Do you know if there is any prospect of better-targeted flu vaccines (e.g., using mRNA technology)?

And certainly not install an app!!!

Newsrooms should not block the Internet Archive. Any public service publisher should not block the Internet Archive. If anything, we should all be directly supporting the Internet Archive.

The trend lines are down across party lines. I spent my career in higher education, and I think there is plenty of room for improvement.

Yes! I keep thinking they will slow down these releases and then it will be worth my time to figure out how to use them and what they are good for.

Better to have too many than to be dependent on only the big expensive ones. Indeed, I'm worried that we will be stuck in that situation in 2-3 years from now.

These systems are trained to imitate. Human emotions (especially when verbalized) presumably have high power to predict future words or at least to globally shift the vocabulary. The models learn this. Users can prompt this.

It always struck me as more negative; eg bird brains

A question of non-Americans. Does “hen’s night” have more derogatory overtones than bachelorette?

Yes, if the characters are not Americans, they shouldn’t talk like Americans. I recall reading Harry Potter in Spanish, and of course the translator needed to keep as many Britishisms as possible.