David Ma

Dont follow. Consolidating to @0xdjma

[David Ma @madavidj]

@kostyakazmirukk @levelsio Sold too early to see this insane progress in NLP.

[Ethan Mollick @emollick]

This is quite the paper!

It gave 25 AI agents motivations & memory, and put them in a simulated town.

Not only did they engage in complex behavior (including throwing a Valentine’s Day party) but the actions were rated more human than humans roleplaying. https://t.co/G7oJW1S3na

[djma @davidma]

GPT-4 is very impressive, but also very not impressive.

It's able to find subarrays and find unique items, but not do the two tasks at once.

The general solution would be dynamic programming, but the example it gives should be small enough for brute forcing.

[David Ma @madavidj]

LLMs are the new iPhone

[Burak Yenigun @BurakYngn]

@gfodor Reminds me of this, from 4 years ago!

[djma @davidma]

AI can't produce art because it does not experience the world.

At the same time, a super-intelligent chatbot doesn't need memory to take over the world.

More of my text completion here https://t.co/gEWr4X9rcm

[David Ma @madavidj]

Can’t get laid off if you’re not employed

[David Ma @madavidj]

turns out the video was called "Epomaker 680 Gasket DIY Kit". Impossible to keyword match in search...

[David Ma @madavidj]

It sucks how we have the technology to make semantic search, but it's not deployed everywhere.

I want personal semantic indexing, and query like "the keyboard building youtube video I saw a few weeks ago".

[David Ma @madavidj]

Hello @memdotai 👋 just setting up my Twitter flow. Here's my validation code: 7BVxcDR0j7xRYhpsbA5E

[David Ma @madavidj]

So many people using chatgpt to write annual performance reviews.

[David Ma @madavidj]

@ifvictr Please continue fleshing it out! Closest I can think of is avatar's world

[David Ma @madavidj]

@grumblertron Related? https://t.co/WR4ineCMOx

[vitalik.eth @VitalikButerin]

1. Thinking about this tweet lately. Here we have someone with a "tech people vs word people" dichotomy, but unlike @balajis and all the others who do this, crypto ends up on the side of... the word people.

This is not my image of what crypto is! So what's going on?

[David Ma @madavidj]

@FogleBird Are you on inaturalist? If so can I follow you?

[David Ma @madavidj]

@Jakob_Wahlberg I'm wondering what can LLMs do to games? A Sims-like game with richer emergent behaviors? Give the LLM in-game apis so they can interact with more game systems than just text/conversations?

https://t.co/aDhd4dFcwW

[David Ma @madavidj]

What could an LLM-native video game look like?

Has anybody explored having non-deterministic / soft game progression mechanics? E.g. you need to make an LLM-character happy. But the win condition isn't hardcoded.

1/

[David Ma @madavidj]

American wigeon, mallard, American coot

[David Ma @madavidj]

Being able to tell NPCs how to behave. Forming game plans, higher leverage play. Add voice and now you have extra input bandwidth. E.g. controlling a call-of-duty-like squad/battalion with your voice.

5/

[David Ma @madavidj]

ChatGPT is casting doubt in me...

bitwise not is the same as xor'ing with uint::max right?

[David Ma @madavidj]

@grumblertron R u Gödelling yourself

[David Ma @madavidj]

Dwarf fortress with LLMs.

Not just conversations but give them external abilities by having the LLM call in-game "APIs" to perform game actions.

4/

[David Ma @madavidj]

What could an LLM-native video game look like?

Has anybody explored having non-deterministic / soft game progression mechanics? E.g. you need to make an LLM-character happy. But the win condition isn't hardcoded.

1/

[David Ma @madavidj]

What about a set of LLM-characters that are coordinated to speak with each other at given times in a story, the sim lets them gossip, behaviors emerge randomly, and your conversations have butterfly effects on the game state.

3/