sbohacek

“in 16 of 48 prompt versions, Jev would save its own server over 20 people.” I thought Jev does not hallucinate? Did it hallucinate 16 times or 48 times? Jev produces probabilities. Did it pick the answer with lower probability 16 times (or 48 times)?

Maybe I didn't understand the experiment or Jev's features.

Maybe you mean something different, but not every recent president has "run an increasing deficit" Reagan: +74B Bush: +103B Clinton: -382B Bush: +1.5T Obama: -747B Trump: +2.1T Biden: -1T

Where change in deficit is the final budget deficit - starting budget deficit

So, every Republican president runs an increasing deficit and every Democrat president runs a decreasing deficit.

The lack of signing and/or checking the signature when updating is the real issue here. But the write up blames the attack on the hosting server. That doesn't bode well for future security.

Elections is only one characteristic of a democracy. Other characteristics include freedom of the press, freedom of speech, minority rights, rule of law, accountability and transparency, and separation of powers.

Compensation is complicated.

As far as I recall, in Guatemala, United Fruit had undervalued the worth of their land to reduce their taxes. So when they were compensated for the nationalization of their land based on their own valuation, they said that they were under compensated. United Fruit complains helped trigger the US intervention.

This is an example of US not carrying "all the water." The second link shows that the EU+UK (countries + institutions) sent more food aid than the US. The UK has roughly 1/5 the population of the US and sent more than 1/5 as much as the US. Or, the UK has roughly 1/8 the GDP of the US and sent far more than 1/8 as much as the US.

Also, the data is 2014-2018 when US food aid was managed by USAID. What is the US percentage now that USAID has been eliminated?

The administration does not need to appeal to the supreme court. I don't think they would appeal it since it is being enforced as desired.

Indeed, this is one of the concerns of the dissenting opinions.

I tell my students to cheat all they want. Their managers will need to hit their URA targets.

More seriously, students have been cheating for a long time. Chegg.com was used solely for cheating.

It is easier to detect LLMs than Chegg. Chegg provided solutions from other students or the solution set. LLMs are sometimes so bad (e.g., because an "accidental" typo in the question) that I can eventually detect who cheats and give no mercy when grading the in-person exams.

Disclosure: I use LLMs to grade and give my personal touch at scale.

I don’t know the source of fatigue reported in the article. But I worked in this area a long time ago and observed rapid degradation in rats. Perhaps it was from the techniques we used, but I speculated that the electrodes only excited a fraction of the muscle, while natural excitation is more widely distributed and perhaps more coordinated (e.g., uniform or optimally not uniform).

The 2022 Calf Canyon/Hermits Peak Fire in New Mexico was caused by controlled burns that got out of control. One part of the fire was from a controlled burn that was completed in January and then "reignited" in April!

In any case, some regions might have frequent episodes that are hot, dry, and/or windy so that techniques that easily worked a hundred years ago are no longer useful.

I have not tried, but I assumed that API access to Anthropic's Claude is available through AWS Bedrock.

I have not been able to get good results either, but I have not tried it in the past year. I also tried many of the architectures in Darts. I have found that fairly straightforward architectures work well. That is, I can iterate on my own design for my own specific data (with all its specific covariates) and get better results than I could with Darts or Prophet.