bigyikes

10 ms·

91.
by bigyikes 2y ago
I still have nightmares about trying to set up SSL with nginx and my own self-managed certificates. I set the same thing up with Caddy in minutes and it’s been running flawlessly for years. Shoutout to Let’s Encrypt as well for making this
92.
by bigyikes 2y ago
that’s nothing compared to the average package-lock.json! it actually doesn’t seem like that long of a list if you only consider one version per library
93.
by bigyikes 2y ago
Will anybody catch up to SpaceX? As an outsider it seems they are impossibly far ahead of the competition. Blue Origin still has only achieved sub-orbital flight, a far cry away from reusable super heavy orbital flight. Meanwhile SpaceX is
94.
by bigyikes 2y ago
What do you suppose happens to all of the people renting from landlords with multiple properties?
95.
by bigyikes 2y ago
Not a good analogy. “Permission” and “persistence” have the same prefix but entirely different semantics. They also occur more commonly in everyday life. AuthN and AuthZ are similar in in spelling, appear in similar contexts, and are less c
96.
by bigyikes 2y ago
Not to be that guy, but this is the kind of task that LLMs are great at assisting with.
97.
by bigyikes 2y ago
“Story” mode is why I like to organize my code top-down, often with definitions appearing after their usages. It makes the code read more like a story from top to bottom. Some like to organize their code such that any term always appears be
98.
by bigyikes 2y ago
The trick is binary addition works the same as regular addition, but is so simple that even a silly little logic gate can do it. Think about adding two bits. There are only four possible ways to do it. 0+0=0 0+1=1 1+0=1 1+1=0 (carry 1) If y
99.
by bigyikes 2y ago
Is it stagnation in reasoning ability, or is OpenAI pulling their punches? It’s suspicious that despite being trained on audio tokens in addition to text and image tokens it performs almost exactly the same as GPT-4. GPT-4o could be a half-
100.
by bigyikes 2y ago
Yes, they can, and they should get better at this over time. There is a demo video where the presenter breathes heavily and asks the AI is able to notice it as such when prompted. It can’t just detect tone, it seems to also be able to use t
101.
by bigyikes 2y ago
How do you generalize that for all inputs though?
102.
by bigyikes 2y ago
It didn’t just demonstrate the ability to sing, but also the ability for two AIs to cooperate! I’m not sure which was more impressive
103.
by bigyikes 2y ago
The AI duet really starts to hint at what will make AI so powerful. It’s not just that they’re smart, it’s that they can be cloned. If your wallet is large enough, you can make 2 GPTs sing just as easily as you can make 100 GPTs sing. What
104.
by bigyikes 2y ago
If true, makes me wonder what kind of regression testing OpenAI does for these models. It can’t be easy to write a unit test for hallucinations.
105.
by bigyikes 2y ago
That old feature uses Whisper to transcribe your voice to text, and then feeds the text into the GPT which generates a text response, and then some other model synthesizes audio from that text. This new feature feeds your voice directly int
106.
by bigyikes 2y ago
Besides increasing the vocabulary size, one way to use “fewer tokens” for a given task is to adjust how the tokenizer is trained with respect to that task. If you increase the amount of non-English language representation in your data set,
107.
by bigyikes 2y ago
Sam has stated that he knows the month GPT-5 will be released. Given the amount of time and uncertainty involved in training and red-teaming these models, we can assume GPT-5 exists if we take Altman at his word.
108.
by bigyikes 2y ago
Plot model performance over the last 10 years and show me where the convergence is. The graph looks like an exponential and is still increasing. Every exponential is a sigmoid in disguise, but I don’t think there has been enough time to say
109.
by bigyikes 2y ago
It isn’t clear that we are running out of training data, and it is becoming increasingly clear that AI-generated training data actually works. For the skeptical, consider that humans can be trained on material created by less intelligent hu
110.
by bigyikes 2y ago
There is an increasing amount of evidence that using AI to train other AI is a viable path forward. E.g. using LLMs to generate training data or tune RL policies
111.
by bigyikes 2y ago
You can get a lot closer to the center of mass of the black hole, which should drastically increase acceleration since it falls off with distance squared.
112.
by bigyikes 2y ago
Someone in the video talks about Homer as though he were the author of the epics. My impression was that these epics came about via oral tradition, with centuries of recitation by bards. Maybe this idea was just put in my head by Julian Jay
113.
by bigyikes 2y ago
This article, like so many AI hit pieces, fails to make any credible claim against AI. This one spent far too long propping up a tenuous analogy to smoke and mirrors. Is current-gen AI valuable? Obviously. Look at how many people say as muc
114.
by bigyikes 2y ago
i hate javascript, typescript is ok i don’t understand the hate for TS. after the initial learning curve, it has only made my life easier. i can’t imagine making a large refactor on a JS code base - the fear of breaking some unknown contrac
115.
by bigyikes 2y ago
> All of this should be done by professionals and nothing is "retail" ready. Don’t worry, just train the LLM to always append “This is not financial advice.” to their responses. Boom, retail ready.
116.
by bigyikes 2y ago
I learned Rust, with great help from ChatGPT-4. If I can learn from AI-generated content, then I totally believe that AI can too.
117.
by bigyikes 2y ago
Ignoring the “spits out training data” bit which is at best misleading, it’s interesting that you use the word “abstract” here. I recently followed Karpathy’s GPT-from-scratch tutorial and was fascinated with how clearly you could see the m
118.
by bigyikes 2y ago
Wouldn’t you expect higher level jobs to pay less then “grunt” jobs then?
119.
by bigyikes 3y ago
I grew up in the south and was fortunate enough to move to the Bay a few years ago. I was expecting it to be some kind of utopia, with futuristic technology on every corner. In reality, it is roughly the same as anywhere else in America. A
120.
by bigyikes 3y ago
> If someone gets 20% on an exam I don't go "great, thats 20% of the way there!!!", instead I go "you clearly didnt attend, try again next time". Sure, if little Bobby gets a 20% I’ll whoop his ass, but if the in

More ›