wyager

7 ms·

61.
by wyager 1y ago
I see no issue with the satellite backhaul itself being unencrypted; anyone using the satellite provider should assume they're hostile and encrypt+authenticate everything they send anyway. I don't trust my ISP's fiber to be s
62.
by wyager 1y ago
This sounds worse than cash in almost every dimension.
63.
by wyager 1y ago
You don't actually need to have a "responsible person"; you can just have an AI do stuff. It might make a mistake; the only difference between that and an employee is that you can't punish an AI. If you're any good
64.
by wyager 1y ago
> with lessons learned A very small subset of possible lessons that could have been learned...
65.
by wyager 1y ago
This seems less than ideal to me. 1. Different languages have totally different allocation requirements, and only the compiler knows what type of allocator works best (e.g. generational bump allocator for functional languages, classic mallo
66.
by wyager 1y ago
> making the UK one of the world leaders in this technology Are they also planning on completely overhauling their economy and tax system to attract the engineers required to make this happen?
67.
by wyager 1y ago
> A good test is one that evaluates candidates on the tasks that they will do at the workplace Systematizing this is not feasible. The next best thing (in terms of predictive power for future job success) is direct IQ tests, which are il
68.
by wyager 1y ago
Leetcode with no prep is a pretty decent coding skill test The problem is that it is too amenable to prep You can move your score like 2stddev with practice, which makes the test almost useless in many cases On good tests, your score doesn&
69.
by wyager 1y ago
It's very minimal-boilerplate. It's done an exceptional job of eliminating procedural, tedious work, and it's done it in a way that doesn't even require macros! "Template Haskell" is Haskell's macro system
70.
by wyager 1y ago
> Why hasn't there been a minimal-boilerplate language and framework and programming environment? Haskell mostly solves boilerplate in a typed way and Lisp mostly solves it in an untyped way (I know, I know, roughly speaking). To pu
71.
by wyager 1y ago
I can construct even more compelling arguments for the uselessness or active harmfulness of most tech jobs. The only "actually useful" tech jobs if we're going to consistently apply a bar that excludes finance would be stuff
72.
by wyager 1y ago
I used to work there, so with the appropriate deference to that Upton Sinclair quote about paychecks: Market makers like JS vastly increase market liquidity across all sectors, which is required for modern high-efficiency economies to work
73.
by wyager 1y ago
Most likely they got slapped in india because a domestic market maker bribed the right regulators and JS did not
74.
by wyager 1y ago
You're giving way too much credit to the emergent intelligence of the CA legislature
75.
by wyager 1y ago
> or 56 x 1 Gbit servers 100% saturated Presumably a caching server would be 10GbE, 40GbE, or 100GbE 56Gbit/sec of pre-generated data is definitely something that you can handle from 1 or 2 decent servers, assuming each request does
76.
by wyager 1y ago
> I don't see how super duper auto complete will lead to AGI Autocomplete is the training algorithm, not what the model "actually does". Autocomplete was chosen because it has an obvious training procedure and it generaliz
77.
by wyager 1y ago
> a stochastic model for predicting text It's fascinating to me that so many people seem totally unable to separate the training environment from the final product
78.
Training Neural Audio Networks(yager.io)
2 points by wyager 1y ago|0 comments
79.
by wyager 1y ago
> god forbid you're researching a complex and possibly controversial subject and you want it to find reputable sources If you're really researching something complex/controversial, there may not be any
80.
by wyager 1y ago
> One of these seems like the solution to the other. Humans are not fungible cogs
81.
by wyager 1y ago
Well yeah, AI power consumption doesn't match the solar production curve.
82.
by wyager 1y ago
I remember when homebrew first came out it was pretty snappy and didn't do a bunch of extraneous nonsense. A remarkably precipitous decline in software quality
83.
by wyager 1y ago
Many people were extremely skeptical of the internet in the early 90s. You can find old clips of news shows basically mocking the idea.
84.
by wyager 1y ago
Is this "full" ECC, or just the baseline improved ECC that all DDR5 has? Either way, on my most recent NAS build, I didn't bother with a server-grade motherboard, figuring that the standard consumer DDR5 ECC was probably good
85.
by wyager 1y ago
> Large language models are impressive statistical text predictors — genuinely useful tools that excel at pattern matching and interpolation. Slightly OT: It's interesting how many (smart!) people in tech like the author of this art
86.
by wyager 1y ago
It sounds like the "scaling function" mentioned in the article may be intended to account for the nonlinear interaction of the noise.
87.
by wyager 1y ago
"100% foolproof" is not a realistic goal for any engineered system; what you are looking for is an acceptably low failure rate, not a zero failure rate. "100% foolproof" is reserved for, at best and only in a limited sen
88.
by wyager 1y ago
> you should never be using prompting to add guardrails like this in the first place This "should", whether or not it is good advice, is certainly divorced from the reality of how people are using AIs > you need actual restr
89.
by wyager 1y ago
Note that the world-model "everything Chinese is spying on you" actually produced a substantially more accurate prediction of reality than the world-model you are advocating here. As far as being "very welcoming", that&#
90.
by wyager 1y ago
> The solution here shouldn't be technical; it should be legal. The technical solution seems strictly preferable Legal "protections" only protect you up the moment a warrant is issued, if that

More ›