BenoitEssiambre

8 ms·

61.
by BenoitEssiambre 2y ago
Python has open source implementations that act as defacto standard, as specification of the details.
62.
by BenoitEssiambre 2y ago
You don't think Mozilla adopting chromium or a fork of it would fix a lot of their problems?
63.
by BenoitEssiambre 2y ago
Imo, it's just too difficult to maintain compatibility through a natural language specification. Things like browsers need some kind of open source reference implementation to act as a spec to maintain interoperability. https:/&#
64.
by BenoitEssiambre 2y ago
Yeah projects like this that take into account more solid macroeconomic principles like Milton Friedman's k-percent rule ( https://en.wikipedia.org/wiki/Friedman%27s_k-percent_rule ) are the only ones that deserve
65.
by BenoitEssiambre 2y ago
The question then becomes: when do you break out code or not? Unfortunately (or fortunately if the art and craft or programming fascinates you), the answer is not easy. It seems to have to do with avoiding over-fitting or under-fitting the
66.
by BenoitEssiambre 3y ago
Low bit parameters is always talked about in terms of performance benefits but I wonder if allowing the LLM to combine parameters to represent values, means it can select the resolution of each value, that is use a kind of internal scientif
67.
by BenoitEssiambre 3y ago
Thanks for engaging with my wild ideas. I do think that's a good description yes. There are corollaries like Dependency Length Minimization I mentioned, which I think helps lower the MDL. Putting related things together allows using sm
68.
by BenoitEssiambre 3y ago
That's interesting. Knowing when to decompose systems into modules indeed seems to be key. This is a complex problem because, I think, the choice of the optimal model depends on the uncertainty you have about the reality behind the dat
69.
by BenoitEssiambre 3y ago
I always wondered if there would be a market for smaller scale apps. I'm thinking something sold through an app "farmer's market". This could be a Patreon/Etsy style platform where maybe app devs would do streamed l
70.
by BenoitEssiambre 3y ago
Yup. I'm convinced that part of Ticketmaster's business proposition is to absorb the hate for artists who want to maximize profits but don't want to look responsible for the ruthless capitalist approach because it would taint
71.
The Divine Coincidence of Abstraction with the Bayesian Occam's Razor(benoitessiambre.com)
3 points by BenoitEssiambre 3y ago|0 comments
72.
by BenoitEssiambre 3y ago
I can add an anecdotal data point to this. Intermittent fasting only started mildly working for me without making me feel weak and head-achy when I re-added electrolytes in the form of dill pickles, Lipton cup of soup or high enough dose su
73.
by BenoitEssiambre 3y ago
In Laplace's 1795 essay introducing what is now known as Bayesian probabilities, one of the first example given was of a kind of simple language model, potentially making the first word used in such models, "Constantinople".
74.
by BenoitEssiambre 3y ago
It occurs to me that after all these years of the world using Wiener et al.'s "cyber" terminology in sci-fi to refer to digital connectivity and over-hastily applying the term to internet and web things, the concepts it origi
75.
by BenoitEssiambre 4y ago
>could the observed effect be significant with a larger trial? Sure. But that's always true for a negative result. Sure, this is true, it's one of the reasons why results being significant or not is not very relevant. At some p
76.
by BenoitEssiambre 4y ago
> In a randomized controlled trial you either find a significant difference in your metrics, or you don't. There's no other option. This is a poor way of thinking about statistics. Whether you reject or not a sharp null hypothe
77.
by BenoitEssiambre 4y ago
That is incorrect. The title is completely wrong. "Study failed to find a reduction" _does not equal_ "There wasn't a reduction". From the abstract: "The risk of death from colorectal cancer was 0.28% in the in
78.
by BenoitEssiambre 4y ago
I sometimes use an even closer to vanilla option based on just the Lit templating engine: https://benoitessiambre.com/vanilla.html
79.
by BenoitEssiambre 4y ago
On top of these options, with modern browsers, it's quite easy to make reactive components in plain vanilla ( https://benoitessiambre.com/vanilla.html ). I find this approach very usable for smaller simpler projects wh
80.
by BenoitEssiambre 4y ago
Also upcoming insane electricity prices: https://twitter.com/JavierBlas/status/1556998118922665984
81.
Economic Misconceptions of the Crypto World(noahpinion.substack.com)
5 points by BenoitEssiambre 4y ago|0 comments
82.
by BenoitEssiambre 4y ago
In this case I ended up with two very simple trees which I train for each ATM (with a prior trained on the larger fleet). You could try a mix though. I didn't have time to try too many things. The following step which I didn't men
83.
by BenoitEssiambre 4y ago
To add to this. One of the best thing about this approach is how little computational power it requires compared to NNs. I was able to finish this project in little time without having to get extra hardware approved. One core of our DB is n
84.
by BenoitEssiambre 4y ago
I start with common sense priors or "intuition" but also sometimes write code to try different combinations and assess accuracy. For example, not too long ago I was trying to estimate and predict ATM usage and considered using day
85.
by BenoitEssiambre 4y ago
I like decision trees and this helps support my case for using them. I often go even further and don't use an algorithm to build the trees but build trees myself along intuitive causal lines and use the data to train its parameters. I
86.
by BenoitEssiambre 4y ago
Http requests work great with relational dbs. This is not UDP. If the TCP connection is broken, an operation will either have finished or stopped and rolledback atomically and unless you've placed unneeded queues in there, you should k
87.
by BenoitEssiambre 4y ago
To clarify, I think stateless microservices are good. It's when you have too many DBs (and sometimes too many queues) that you run into problems. A single instance of PostgreSQL is, in most situations, almost miraculously effective at
88.
by BenoitEssiambre 4y ago
I'm glad this is becoming conventional wisdom. I used to argue this in these pages a few years ago and would get downvoted below the posts telling people to split everything into microservices separated by queues (although I suppose it
89.
Sim Central Bank Afterthoughts and Something About AI and Empathy(benoitessiambre.com)
1 points by BenoitEssiambre 4y ago|0 comments
90.
by BenoitEssiambre 4y ago
I've started playing with AWS App Runner ( https://aws.amazon.com/apprunner/ ) which makes things fairly simple. You just point to a github repo, tell it the build command and the run command and it will deploy you

More ›