Human Programmers – An unDirectory for the digital renaissance (humanprogrammers.com)

Just like they "know" English. "know" is quite an anthropomorphization. As long as an LLM will be able to describe what an evaluation is (why wouldn't it?) there's a reasonable expectation to distinguish/recognize/match patterns for evaluations. But to say they "know" is plenty of (unnecessary) steps ahead.

sidewndr46 · 2m ago

This was my thought as well when I read this. Using the word 'know' implies an LLM has cognition, which is a pretty huge claim just on its own.

noosphr · 19m ago

The anthropization of llms is getting off the charts.

They don't know they are being evaluated. The underlying distribution is skewed because of training data contamination.

zer00eyz · 5m ago

No, they do not. No LLM is ever going to be self aware.

It's a system that is trained, that only does what you build into. If you run an LLM for 10 years it's not going to "learn" anything new.

The whole industry needs to quit with the emergent thinking, reasoning, hallucination anthropomorphizing.

We have an amazing set of tools in LLM's, that have the potential to unlock another massive upswing in productivity, but the hype and snake oil are getting old.

khimaros · 25m ago

Rob Miles must be saying "I told you so"

Human Programmers – An unDirectory for the digital renaissance (humanprogrammers.com)

Show HN: Freaky Font (freakyfont.app)

Novo Nordisk's patent on Ozempic expired in Canada over unpaid $450 late fee (bsky.app)

The Travel Writer's Dilemma: Share, or Gatekeep? (nytimes.com)

Dice Identification 101: Mold Characteristics (2019) (dicemaniacsclub.wordpress.com)

The Cadence (2020) (medium.com)

New York passes a bill to prevent AI-fueled disasters (techcrunch.com)

Show HN: I created a multiple JSON queries tool (app.postpilot.dev)

Streamlit state management for React developers (outlore.dev)

How I monitor and control all my powered devices (Zigbee and HA) (jeffgeerling.com)

Bear Essentials Vlog (old.reddit.com)

Reclaiming the future: Māori voices leading in the age of AI and quantum tech (thespinoff.co.nz)

I have discovered a new number (with Mathematical Proof)

Show HN: I made an online Unicode Cuneiform digital clock (oisinmoran.com)

Somalia's construction boom in Mogadishu gives women high ambitions (bbc.com)

Making Software Like Lego: How Aspire Brings the Pieces Together (medium.com)

Some sunscreen ingredients may disrupt sperm cell function (2016) (medicalxpress.com)

Show HN: Jiffly - AI agent that finds, calls, & books the best local services (getjiffly.com)

Large Language Models Often Know When They Are Being Evaluated (arxiv.org)

Show HN: An LLM Running on a PS Vita (github.com)

The State of Rideshare and Autonomous Vehicles (platformaeronaut.com)

Adaline (adaline.ai)

Show HN: Tapmytab – an open-source, Kanban with rich text editor on Chrome tab (github.com)

Malware hiding in plain sight: Spying on North Korean Hackers (aikido.dev)

LLM Debugger – Visualize OpenAI API Conversations (github.com)

Wolfenstein 3D on 8-Bit Consoles [video] (youtube.com)

Kontext Image: AI image editing for fast personalized creation (kontextimg.com)

Festival rockers told to take off smartwatches, moshpits spark emergency alerts (theguardian.com)

N. Korea discharges uranium waste into waters flowing to S. Korea (dailynk.com)

Nucleus Co-Op – Play local co-op with only one PC and one game copy (nucleus-coop.github.io)

BBS: The Documentary [video] (youtube.com)

Public Intellectual Privilege (fromthenew.world)

Why You're Wrong About AI Data (fromthenew.world)

Startup Idea Slot Machine App (idea-gen-johnrushx.lovable.app)

Sperm are very different from all other cells (bbc.com)

International comparison of optical clocks connected via fiber and satellite (opg.optica.org)

The Scandal of Outdated Science (hollisrobbinsanecdotal.substack.com)

Chinese AI Companies Dodge U.S. by Flying Suitcases of Hard Drives Abroad (wsj.com)

A way to insert a value in an empty List in a specific index (stackoverflow.com)

Measuring AGI: Interactive Reasoning Benchmarks [video] (youtube.com)

Memory Safety Isn't Just Rust: A Serious Look at GC (gizvault.com)

Q-learning is not yet scalable (seohong.me)

Beware General Claims about "Generalizable Reasoning Capabilities" of AI Systems (lesswrong.com)

Show HN: I made an AI app to suggest diffs to docs & notes (resilientdocs.com)

New York State Updates Warn Notices to Identify Layoffs Tied to AI (bloomberg.com)

In Munich, early signs of a European hyperscaler revolt (raconteur.net)

Reunion (genius.com)

How and Why We Fulfill In-House (usetrmnl.com)

'Ghost' AI students being enrolled in college as a financial aid scam (fortune.com)

Terence Tao: Hardest Problems in Mathematics, Physics and the Future of AI (youtube.com)

Large Language Models Often Know When They Are Being Evaluated

Comments (5)