Erçağ Pinçe

AI Engineer | Freethinker
Former Senior Scientist/PENN Data Science Fellow
MSE in AI @PennEngineers.

"Why do we fall?"

[Erçağ Pinçe @ercagpince]

@turkanlhanoglu Yanında taş gibi güzel bir kadınla mutlu ex'inizi görünce görürüm ben sizin "içi geçmiş sevginizi", "aniden defolup gitme dürtünüzü". Boş laflar.

[Mitchell Hashimoto @mitchellh]

I've got an agent in a loop optimizing a renderer with the goal to minimize frame times (and tests to measure). It got times down from 88ms to 2ms and allocations down from ~150K to 500. Sounds good, right? Wrong. This is exactly why agent psychosis is a big fucking problem.

As an experiment, I rewrote the Ghostty core render state in Go, with access to identically laid out data structures as Ghostty and the exact same validation tests. I made a purposely naive renderer (simple, correct, but slow). 88ms per frame with 150,000 allocations (horrendous, lol)!

I then kickstarted a Ralph loop to bring the frame times down. I told it it can't modify input data structures or the public API or tests (they're correct), but it can do anything else it wants. It got to work.

It has worked for about 4 hours. I've spent around $350 on this experiment so far. The results?

88ms => 1.5ms
150K allocs => ~500 allocs

Incredible right? Nope.

My hand-written renderer I ported has frame times (same benchmark) of ~20us (0.020ms) and 0 allocations in the update path.

This is the problem with psychosis and lacking systems understanding. If you don't understand the system, you're going to accept that this is an incredible result. If you understand the system, you'll see better solutions immediately and can do roughly 75x better on throughput.

The people who blindly trust agent output are in the former camp. They're sheeple, overdrinking from a fountain of mediocrity.

Standard disclaimer: I use AI all the time. I like AI. The point I'm making is to not blindly accept results. Think. Analyze. Learn.

[Erçağ Pinçe @ercagpince]

@ajansfenercom Music: Hotline Miami

[Erçağ Pinçe @ercagpince]

@mervenoyann This metric doesn't make any sense.

What's the control here? And, how can you establish a control when it's not guaranteed to reset to initial conditions?

If it's repeated many times, I am pretty sure system would be thermalized in long-term (with no net winner whatsoever).

[Melanie Mitchell @MelMitchell1]

New paper: "Large Language Models & Emergence: A Complex Systems Perspective" (D. Krakauer, J. Krakauer, M. Mitchell).

We look at claims of "emergent capabilities" & "emergent intelligence" in LLMs from perspective of what emergence means in complexity science.

⬇️

[Erçağ Pinçe @ercagpince]

@thinking_panda A typical day around Harvard Square. I am pretty sure it's not AI-generated.

[Erçağ Pinçe @ercagpince]

@beehaber Survivorship bias.

[Erçağ Pinçe @ercagpince]

@yetkingdom Mevzu bahis Black Mirror bolumu:
https://t.co/0BCEBDUuZL

[Erçağ Pinçe @ercagpince]

@yetkingdom Bu kadin bu isten iyi para kazaniyor olmali. 1M+ takipcisi var.

Ve her gun "bilime butun inancim kayboldu", "su scam, bu zirva" seklinde videolar paylasiyor. Yaygaraciligi paraya cevirme konusunda bayagi basarili.

Black Mirror bolumu gibi oldu dunya.

[Erçağ Pinçe @ercagpince]

Air grievances, make money.

[Sabine Hossenfelder @skdh]

In which I explain why I worry about scientific progress

https://t.co/KxAOWULCax

[goosewin @Goosewin]

talk is cheap

[Erçağ Pinçe @ercagpince]

@fchollet No Jupyter notebooks then(?)

[Erçağ Pinçe @ercagpince]

@cemilemreyavas @BaserYrp Tebrikler. Test datasini ve modeli bir gormek lazim.

[Erçağ Pinçe @ercagpince]

@_1s2_2s2_2p6 @canitti Nature degil. Scientific reports.

[Erçağ Pinçe @ercagpince]

@kayabay @canitti Nature degil. Scientific reports.

[Jack Morris @jxmnop]

BREAKING: The Nobel Prize in Physics has been awarded to ptrblock for “fundamental contributions to physics”

[Erçağ Pinçe @ercagpince]

A perfect idea!

[ₕₐₘₚₜₒₙ @hamptonism]

Students at Stanford have built alphaXiv, an open discussion forum for arXiv papers.

[さかき@「極楽に至る忌門」「明治恋々洋菓子帖」連載中 @sakakir]

トルコのおじさんあまりにかっこよくてスケッチしてしまった　描いてみるとあらためてわかるんですけど姿勢の美しさがすごいわ