TBD | Ex @fluidstack @stedi @NOETIK_ai
Vancouver, WA
Joined September 2013
- Tweets1.4K
- Following545
- Followers960
- Likes2.1K
LLMs aren’t the only form of intelligence that struggle to generalize outside of their training data.
If you had an exponential amount of compute you could solve simulate biology exactly and wouldn't need to run any experiments.
Some of you will think "of course, but that's an unrealistic assumption" and others will think "no, you need experiments even with that much compute. That's just How It's Done".
You would of course need evaluations once you have a solution, but you don't in principle need experiments to find it: Experiments are a way to save compute, dating from a time when it was much more expensive.
And I think the presentation is exponential but real systems aren't irreducibly exponential. AlphaFold worked, after all.
Dropped a new game on @spawn. Escape the permanent underclass while trapping everyone else in it.
spawn.co/@snarkykoala/tappyt…
Think I've finally nailed down an AI-aided art workflow for games that doesn't suck. Still some bugs to workout and missing features but I really like the results so far and how much control you retain working this way.
The new models are really good at coding in new languages you just invented… this wasn’t true a year ago. People are sleeping on this.
Most of the ideas I'm the most excited about require fast procedural prompt to image generation. I've tried a lot of toolchains but nothing has really fit the bill...
So I decided to make a token-efficient programming language for art instead. Story time is going to be lit 🔥
Stoked to announce my new game Roadster League!
Thanks to @spawn you can try out the new Tesla features today.
spawn.co/@snarkykoala/roadst…
OpenAI and Anthropic are going to eat everything… but they’ll be priced too high at IPO is a wild take.
It's inevitable and Marx was right in this sense, capitalism will eat itself on a long enough time span
One way to protect yourself is become co-owner of the AI frontier labs and companies,
Many of which are just big tech so I hold a lot of their stock: Google, SpaceX, Nvidia and when OpenAI and Anthropic go public maybe hold them too (but probably priced high then)
That's one great thing about the US stock market, you can co-own these
Best evidence of why software isn’t dead. Your average person prefers a packaged solution to a blank slate even if the alternative is better.
Lovable is a scam.
So is Replit, Base44, Framer, and all AI website builders.
These are for low IQ people who don’t know how to use AI and fall for SaaS marketing gimmicks.
You are paying a premium for less usage, shittier models, shittier websites, and hand cuffing yourself to a SaaS ecosystem you don’t control.
A $100/month Claude Max plan (or ChatGPT Pro) is a better solution BY FAR
Getting excited about my retirement home opening up next month and so I did what any sane person with a Claude subscription would do and started simulating some leveling routes...
Problem is that assuming linear paths doesn't give you good travel estimates so I flooded thousand needles and found a hidden path from the floor of thousand needles to the great lift.
chemistryworld.com/news/regu…
Give humans some credit.
There is a non-zero chance the Wolverine game is an industry psyop to make AI slop look good by comparison lol...