🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Between the laptop I have today and the one I had seven odd years ago Linpack performance has gone up by 5x but Conjugate Gradient performance has gone up only by 2x. Communication is holding back computation right now.

o3's hallucinations are at another level. Today I had put wrong columns in my SQL join and o3 instead invented a whole hitherto unknown limitation in DuckDb, complete with link to irrelevant source, to explain the error.

Of the 3 major processors on my new laptop 'CPU', I have an easily installable toolchain only for the x86 cores. Setting up the tooling for the GPU or AI processor is too cumbersome to contemplate, documentation hard to find. I bought that stuff, I should be able to program it!

If you think that authoritarianism happens because of wrong people being in power, you are missing the root cause. The problem always and everywhere is that ...

Having learnt Unix in the last century, sometimes I miss out on newer tools. Liking 'glances' by Nicolargo 🐧 as an alternative to 'top' as a system monitor. I especially like how it can also show GPU usage.

From a research team I trust -- Of ~20k US-based academics on Twitter in 2022, 21% have since deleted their accounts. This differs a lot by field: above 30% for English and History, closer to 10% for Computer Science, Engineering, and Economics. (1/3)

After rewriting my data prep pipeline for the third time while the prediction metrics stubbornly refuse to improve: जुस्तुजू जिस की थी उस को तो न पाया हम ने इस बहाने से मगर देख ली दुनिया हम ने

> be me, first year PhD student > get obsessed with polyphonic music transcription during pandemic > for some reason, advisor lets me work on music > brilliant idea: explicitly encode chord structure into the model > spend months implementing complicated chord-aware audio

For me it is -RMSE right now.

The rewrite worked. Training is no longer bottlenecked by data loading. On a GPU VM training is now 10x faster than on my laptop, as it should be. ज़िंदगी जब भी तिरी बज़्म में लाती है हमें ये ज़मीं चाँद से बेहतर नज़र आती है हमें

Is chemistry a science? Why? Discuss.

This screen is getting burnt into my retina. Life as a data miner really wears you down.

Goodbye Stanley Fischer. When some of us were cut off from world trends in macroeconomics, your book provided a bridge

I realise with horror that the neural network model I've spent dozens of hours training is seriously biased. It has positive average forecast errors.