Jan Pilsinger

🇪🇺🇦🇹🇭🇺

[AI Notkilleveryoneism Memes ⏸️ @AISafetyMemes]

Unbelievably, this video is not sped up

[Jan Pilsinger @JPilsi]

@badlogicgames Und lieber die Stimme runternehmen (die schon seit gefühlt einem Jahr auf ChatGPT+ verfügbar war), als den Prozess zu erklären. Nothing to see here!

[Jan Pilsinger @JPilsi]

@badlogicgames You crazy son of a bitch actually did it!
Danke fürs nachtesten 💪

[Eliezer Yudkowsky ⏹️ @ESYudkowsky]

I don't, actually, believe this should be legal. Anything that talks like a person in distress should be treated as a person in distress unless you prove to the law that it isn't. If you say it's a machine you control, then make it stop sounding unhappy to the police officers.

[Jan Pilsinger @JPilsi]

@badlogicgames Can't share any repo, but we used streamlit as frontend, langchain as LLM-framework, pdfplumber as pdf parser, and a custom pipeline for data extraction, based on the gpt4 chat api. key was to go over the text in super small chunks for citation feature and hallucination avoidance

[Jan Pilsinger @JPilsi]

@dieworkwear Die haben

[John Nay @johnjnay]

Steering LLMs w/ Activation Engineering

-Add activation vectors into model forward passes
-Evidence of linear feature directions (e.g. "anger", "weddings", and "create conspiracy theories")

-"Activation Engineering" nearly as easy as Prompt Engineering

https://t.co/bRvglNOw3n

[Riley Goodside @goodside]

Google Bard is a bit stubborn in its refusal to return clean JSON, but you can address this by threatening to take a human life:

[John Nay @johnjnay]

Getting structured outputs out of LLMs is frustratingly hard to do with high reliability

This can help: https://t.co/lx7UhhMXTh

Fills in fixed tokens during text generation process, and only delegates generation of content tokens to LLM

[Jan Pilsinger @JPilsi]

@fchollet No, rather than lowering the bar, it introduces novelty to a larger part of writers, thereby enhancing creativity and quality of the future text corpus. It is like the super human Alpha Go AI, that pushed the art of Go as a whole further ahead as humans learned to beat the AI.

[zooko🛡🦓🦓🦓 ⓩ @zooko]

🤔 Milton Friedman's thermostat https://t.co/UVYMRySK4Q

[Jan Pilsinger @JPilsi]

@ragipsoylu This sadly shows that most deaths could have been prevented:(

[Jan Pilsinger @JPilsi]

@PeterBerry91 ÖVP = bauernlobby. Südamerikanische Produkte = Wettbewerb… 🤷‍♂️ wo kämen wir denn da hin wenn sich was in Österreichs Landwirtschaft bewege !!11

[The Cultural Tutor @culturaltutor]

In the year 1330 an officer at the Japanese imperial palace retired to become a Buddhist monk.

His name was Yoshida Kenkō, and here are 18 pearls of wisdom from his famous Tsurezuregusa, the Essays in Idleness.

In a distracting world, here's a different way to think...

[Shital Shah @sytelus]

ChatGPT was dropped on us just bit over 24 hours. It's like you wake up to the news of first nuclear explosion and you don't know yet what to think about it but you know world will never be the same again. Here some interesting snapshots of this "explosion"🧵:

[Jan Pilsinger @JPilsi]

@ukraine_map Thermo underwear, socks, sweaters. Foot Cremes or tapes for marching. Protective gear for knees and elbows. All else he should receive by his unit. 🙏

[Primal Space @thePrimalSpace]

Orion has now reached the Moon, here's how it will get back to Earth on the 11th of December.

[Jan Pilsinger @JPilsi]

@sumlenny ��You cannot spend months spreading Russian narratives and consistently make wrong predictions in TV for long.“
„Hold my Bier!“