Tessera
Folgen

Alex Irpan

@alexirpan.bsky.social

Research Scientist @ Google DeepMind. Formerly Robotics, now AI Safety. Has a blog. Views are my own.

21 Posts1384 Follower151 Folgt
Alex Irpan@alexirpan.bsky.social · yesterday

The GDM AGI safety team is hiring! If you have concerns about the implications of increasingly powerful AI, please apply. There are openings in both London and the US UK: www.google.com/about/career... US: www.google.com/about/career...

Research Engineer, AGI Safety and Alignment, DeepMind

The Artificial General Intelligence (AGI) Safety and Alignment Team (ASAT) aims to reduce existential and catastrophic risk from AGI and eventually Artificial Superintelligence (ASI). We research nove...

google.com

Alex Irpan@alexirpan.bsky.social · 3w ago

In retrospect "AI CEOs become celebrities and attract corresponding levels of desired and undesired attention" was a very obvious prediction that many people failed to account for (including me)

Alex Irpan@alexirpan.bsky.social · 4mo ago

1. Obviously terrible to have a Molotov thrown against your house, not appropriate response. 2. Of all analogies to make, "ring of power" is a choice, given the story's theme that the only way to stop the ring's destructive power is to destroy it. blog.samaltman.com/2279512

-

Here is a photo of my family. I love them more than anything. Images have power, I hope. Normally we try to be pretty private, but in this case I am sharing a photo in the...

blog.samaltman.com

Alex Irpan@alexirpan.bsky.social · 9mo ago

First paper since switching into AI safety team🎉 We look at problems that could be solved if the model behaved consistently over a set of prompts, and tried training that in output space and internal activations. Both were effective. See thread or paper for details.

Alex Turner@turntrout.bsky.social · 9mo ago

New Google DeepMind paper: "Consistency Training Helps Stop Sycophancy and Jailbreaks" by @alexirpan.bsky.social, me, Mark Kurzeja, David Elson, and Rohin Shah. (thread)

The abstract of the consistency training paper.
Alex Irpan@alexirpan.bsky.social · last yr.

"I don't play gacha games because they're a scam" vs "Let me do one more hyperparam sweep before giving up. One more prompt tuning run. I swear we'll beat baseline. I know it's gonna beat the baseline this time. It's gonna win. This time for sure."

Alex Irpan@alexirpan.bsky.social · 2y ago

The ship has sailed, but I wish the ML reporting default was % incorrect rather than % correct. It better matches loss curves and magnifies the capture of edge cases. 95% accuracy -> 97.5% accuracy = meh 5% error -> 2.5% error = omg we've halved the error rate