pvillano

7 ms·

121.
by pvillano 2y ago
Someone once said the most fruitful research in AI is making models scale to larger compute/data. I think the same could become true for compilers, and I think equality saturation is the key. AI + equality saturation could scale the op
122.
by pvillano 2y ago
There was a leaked memo essentially instructing to form a committee when you make an illegal decision so that one person cannot be sent to jail. Does anyone remember this? I've had a hard time finding it
123.
by pvillano 2y ago
That's gotta be for a sandwich shop, so we can finally `sudo make sandwich`
124.
by pvillano 2y ago
You're right. In order to be a big problem, the error needs to be non-cancelling and inseparable. That makes me wonder: if you label good data, and generate data with the good label, how much benefit do you get from also training on ok
125.
by pvillano 2y ago
Is it worth paying for Kagi when it's future is uncertain? Are there any real competitors for paid search?
126.
by pvillano 2y ago
Image generation has its own problems with non-cancelling noise. For example, images are often generated with jpeg artifacts in regions but not globally. Watermarks are also reproduced. Some generated images have artifacts from CCD cameras
127.
by pvillano 2y ago
A blessing in disguise I guess
128.
by pvillano 2y ago
The circuit is a n-rollover keyboard with the appropriate keys held down
129.
by pvillano 2y ago
You'll get the same tingle at a greater distance. In fact, if the current is finite and current density decreases with distance, it is bounded by the inverse square law. 240v will give you more distance to react between the tingle radi
130.
by pvillano 2y ago
I won't say all investors are entitled and overconfident, inspired by grifters, emboldened by survivorship bias, and motivated by greed. That would be rude
131.
by pvillano 2y ago
Why don't they just rent the thing entirely, six months at a time? Recurring revenue, tiered pricing, no secondhand market, and less waste from planned obsolescence.
132.
by pvillano 2y ago
I wonder how it "reacts" to optical illusions? The ones we're familiar with are optimized for probing the limits of the human visual system, but there might be some overlap
133.
by pvillano 2y ago
What's the alternative?
134.
by pvillano 2y ago
"Bought" is just a fancy word for "granted a license for usage, subject to terms and conditions, which may be revoked at any time, for any reason, without any warning"
135.
by pvillano 2y ago
Cloudflare sells bulletproof vests
136.
by pvillano 2y ago
imo anti-cheat should mostly be server-side behavior based
137.
by pvillano 2y ago
I pushed a one-character fix and broke it a second time
138.
by pvillano 2y ago
Pulling intent out of a invalid statement is difficult but I think the intent is something like > Could the P versus NP question itself be [mathy-difficult in a way] that makes it impossible [for humanity to know the answer]? Which becom
139.
by pvillano 2y ago
Decidability, computability, existence, fruit, all have different meanings in an academic context and in a everyday context, and trying to use intuitions from the everyday meanings in an academic context leads to these "stupid question
140.
by pvillano 2y ago
I dream of a fines system that has "all the money you made doing this" as a minimum.
141.
by pvillano 2y ago
Everyone does this, all the time, without even trying. https://en.wikipedia.org/wiki/Memory#Construction_for_genera...
142.
by pvillano 2y ago
The best/worst part is that there are useful branches of mathematics which assume the Axiom of Choice is true AND there are useful branches of mathematics which assume the Axiom of Choice is false. That's one reason I take the vie
143.
by pvillano 2y ago
It would be interesting to see the Pareto front for Cost, Cycles and Area for individual puzzles. A badge for Pareto optimality would give everyone a chance of being a winner in some small way.
144.
by pvillano 2y ago
Hallucination implies a failure of an otherwise sound mind. What current LLMs do is better described as bullshitting. As the bullshitting improves, it happens to be correct a greater and greater percentage of the time
145.
by pvillano 2y ago
If text generation was done with an adversarial AI, it would be impossible to detect with AI by definition, but still not necessarily at a human level of quality. In that sense only a human is able to detect writing that's not at human
146.
by pvillano 2y ago
it turns out you don't need anything more than minimax for the general case Here's my solution https://github.com/pvillano/probabalistic-tic-tac-toe
147.
by pvillano 2y ago
You can still strategize when the probability of failure and success are equal. For example, O should choose the lower right because it gives them a greater than 50% chance of winning, whereas choosing another spot gives them a greater than
148.
by pvillano 2y ago
You're right. I'll work on the simpler problem of :) / :( first. I think that can be done with just minimax And then maybe win chance for each possible state of a purely random game
149.
by pvillano 2y ago
WRT to computing an exact solution, something something markov chains, transition matrices, eigenvalues. I think it is tractable
150.
by pvillano 2y ago
ChatGPT, rewrite this as a named function Sure, ``` UUID4 = uuid.uuid4() def default_id(): return UUID4 ```

More ›