parpfish

27 ms·

511.
by parpfish 1y ago
I’m really impressed about how well the sound ID works for birds doing mimicry. We had a brown thrasher stringing together a long series of mimicked calls that got labeled correctly as a brown thrasher. I don’t even know how you’d approach
512.
by parpfish 1y ago
I can envision a new variant of the game that replaces lapped/circuit racing with a cross-country rally race. A madcap free-for-all where you need to find your own route and scheme against opponents like the movie “rat race” could fit
513.
by parpfish 1y ago
Also: until you play the game, you won’t know if you’re a big enough fan to want to display some merch for it
514.
by parpfish 1y ago
If you removed fares, you’d turn the subway into a mobile homeless shelter and everyone would stop using it
515.
by parpfish 1y ago
What would happen if you made a lossy image compression format derived from tons of scraped, copyrighted images. There’s no generative ability, but anytime you compress/decompress your image the model uses weights and biases learned fr
516.
by parpfish 1y ago
I think you could argue that openAI isn’t in the business of making art — it’s their users that make AI art that are in the business of making art. There are plenty of reasons for a model to ingest the art that are not directly related to c
517.
by parpfish 1y ago
Something I keep thinking about is that nobody would think twice if some young musician gave an interview that said "{older artist} was a huge inspiration. their songs were fundamental in shaping how i think about the artform", no
518.
by parpfish 1y ago
When people say that consciousness is “an illusion”, I think they mean to say that the illusion is that consciousness has the ability to direct our actions and the consciousness is an epiphenomenon. Our body processes sensory inputs and con
519.
by parpfish 1y ago
Theres plenty of awkwardness if you approach a bunch of guys that are all friends with each other. Can I join in this conversation, or is it a “closed” private clique?
520.
by parpfish 1y ago
The problem is more that you never bother to have the convo where ask how things are going because your in your mind you think you’re all up-to-date.
521.
by parpfish 1y ago
counterpoint: back in the long-long ago when i was online dating, the biggest boon it provided was that it gave you a space where you knew that it was okay flirt and that your intentions would not be misunderstood. you never had to guess
522.
by parpfish 1y ago
> Everyone is so mobile and connected online, they never have to break the ice and talk to those around them one of the unexpected consequences of social media is that people have been conflating being informed with being connected .
523.
by parpfish 1y ago
but a model could write blog posts that describe changes to the website as a blog-style changelog (e.g., 'today i spent an hour playing with CSS to change padding' or 'i refactored the backend to do more async calls') a
524.
by parpfish 1y ago
half expecting this to veer into some analog horror
525.
by parpfish 1y ago
An actual LLM use case! A model that generates AI slop blog posts so you don’t need to write content and can just focus on the fun parts of making the website
526.
by parpfish 1y ago
s/sacred/profane/p
527.
by parpfish 1y ago
there's also the fact that you can't really show pride in work if you're being forced to follow a script or the demands of some piece of bureaucratic paperwork. i think a prerequisite for being proud of your work is that you
528.
by parpfish 1y ago
i've heard people talk about a future where everybody gets bespoke custom media made by AI. I can see how it would work in theory now , where a machine has a list of my previous media consumption patterns and I can give it some list o
529.
by parpfish 1y ago
grad school is definitely a prestige move. not a 'get rich move', but def a prestige move. prestige is not just money. for med or law school, there are very clear hierarchies about who's better than who and next steps in your
530.
by parpfish 1y ago
i think its interesting that for so many college kids, the post-graduation options that continue to provide tracks are also treated as the more "prestigious" options (go to grad school. work at big3/faang, etc). it's n
531.
by parpfish 1y ago
Also: the “find your people” advice would be far more helpful at the beginning of college so you can maximize the various high-leverage opportunities around you.
532.
by parpfish 1y ago
And to top it all off, that dang apple hit him on the head
533.
by parpfish 1y ago
mitochondria is the powerhouse of the fatigue research
534.
by parpfish 1y ago
It’s interesting how “turbo” has had so much semantic drift that people don’t even know it’s a specific component in an engine. They just think it means “fast”. Wouldn’t be surprised to eventually see a “turbo” trim levels for EVs someday.
535.
by parpfish 1y ago
This is also how I've often thought about deep learning -- focusing on the geometry of the data at each layer rather than the weights and biases is far more revealing. I've always been hopeful that some algebraic topology master
536.
by parpfish 1y ago
I think before you would deploy this to prod, you should wrap it with a few guardrails to make sure it’s not hallucinating. Pretty simple — just take the output from the llm and see if it agrees with a simple mod2 operation. Of it agrees, r
537.
by parpfish 1y ago
Use this to add AI to your product to appease management. Next week, “refactor” it out and brag to manager about cost savings and performance boosts, don’t mention “removing the AI”.
538.
by parpfish 1y ago
I made up a similar number, so it’s been replicated
539.
by parpfish 1y ago
Well, yeah. Types don’t remove the need for ALL tests, but it removes the need for the most boring 90% so your test suite can just focus on the most important behaviors
540.
by parpfish 1y ago
i like static code because a codebase that applies types well gives you the same benefits of writing lots of tests for free .

More ›