6x Productivity Gap

The biggest question about ai productivity is not if it will work, it's what incentives will be in place.

Warren Buffett emphasizes the power of incentives with quotes like, "If you have a dumb incentive system, you get dumb outcomes," and, "Show me the incentive and I will show you the outcome."

If employers do not properly incentivise increased productivity, employees will take those productivity gains and run with it. Americans could very much use a better work life balance, but if innovation isn't incentivised, it will be crushed.

If increased productivity only leads to increased work load, why would a top performer spread their techniques?

Business owners, ask yourself what outcome you want, and figure out how to incentivise to get what you want: if you don't, someone else will take your best employees.

Source: https://venturebeat.com/ai/openai-report-reveals-a-6x-productivity-gap-between-ai-power-users-and

I’m the Best (I said so)

I probably wouldn't take Google's word that their model is the best. Since results are marginal, and the leading team created the benchmark...

This is a very interesting benchmark. As results get more accurate they will be increasingly useful. Now I just need to hope this benchmark will be relevant beyond flexing their top score. (I know their ai course wasn't).

Source: https://deepmind.google/blog/facts-benchmark-suite-systematically-evaluating-the-factuality-of-large-language-models/

Hard Problems

All this new ai buzz has brought up another classic: the hard problems paradox. In cryptography, there is a need for problems that are easy to state and compute, but are hard to reverse engineer or solve efficiently without the key.

The interesting part is that no one knows if these "hard problems" actually exist! Complexity theory has not been able to prove that any problems exist that match those criteria; therefore the paradox that the entire field of securing your data is based on an improved theory!

If that theory is wrong, the entire security industry will collapse overnight. Certain new technology such as quantum computing already makes that theory rather dubious, and many headlines about how quantum computers just solved problems that would take a supercomputer billions of years are cryptography problems!

TLDR: nothing is safe, it's just a matter of how much someone wants something and what resources they have access to.

Source: https://www.quantamagazine.org/cryptographers-show-that-ai-protections-will-always-have-holes-20251210/