pshc

12 ms·

91.
by pshc 3y ago
And yet the company where I work can't find enough Flutter talent.
92.
by pshc 3y ago
Even if we replaced all fossil fuels overnight we still wouldn't have "solved" climate change--the damage we have done is locked in for the long term. We need active mitigation to restore Arctic ice, de-acidify the oceans, sa
93.
by pshc 3y ago
That's true, and one can bias logits in llama.cpp and friends too, but those are global biases that affect the entire output rather than being specified per-token. Uploading a grammar or a wasm binary to the inference engine does seem
94.
by pshc 3y ago
I don't think this is correct; previously you could already control output by reading tokens one at a time from the LLM until you hit a stop character. My take from the grammar-based sampling PR is that you ask llama.cpp to constrain t
95.
by pshc 3y ago
The superior alternative is drawing a diagram on a piece of paper to exercise the mind and show to collaborators, and never setting foot anywhere close to the UML tarpit.
96.
by pshc 3y ago
Just dreaming over here, but if someone had the opportunity to rebuild a Factorio from the ground up, I bet they could design something massively scalable. Something based in cell automata, like how water flow works in Minecraft. Current ce
97.
by pshc 3y ago
When I hear "workstation" what comes to mind is a daily driver for business purposes, so rock-solid reliability is utmost. Redundant components, ECC ram, rated to run 24/7 with ample cooling. Performance per watt is also up t
98.
by pshc 3y ago
I think part of this is because GPT doesn’t have any “hidden variable” storage and doesn’t get any prep time up front to come up with something coherent. Just completes the next token based on the previous context.
99.
by pshc 3y ago
BARRET: Have y'all ever considered that none of this is real? That we're just part of some game, a simulation? TIFA: Barret, what are you talking about? AERITH: It's an interesting thought, though. Sometimes, everything does
100.
by pshc 3y ago
Here I was thinking it was merely a conduit to buying an O-1 visa.
101.
by pshc 3y ago
Seems like the energy bill would be more expensive in the long run compared to getting a cheap sound vide accessory.
102.
by pshc 3y ago
The image is quite bitcrushed, so the underscores are probably missing due to sampling error. Notice how on the same baseline there’s a missing period. But it’s a fishy image regardless.
103.
by pshc 4y ago
Looks smoother, transparency, data small enough to inline in the HTML or JSON payload, supports not just JPEGs but also PNGs, WebPs, GIFs. IMO I don't really care for a 75%-loaded progressive JPEG. Half the image being pixelated and ha
104.
by pshc 4y ago
Proof of stake involves lockup of capital, wasting the time value of money (despite it earning a small yield), and also taking on the risk of being slashed.
105.
by pshc 4y ago
Well, I wasn't worried until I saw your comment ;)
106.
by pshc 4y ago
It's an interesting idea for sure, but loss doesn't go down forever. I think this ends with a highly overfitted network that grinds to a halt as the loss function hits local minima. Even if you get past that, there's no conse
107.
by pshc 4y ago
I don't think it's that easy at all. The work function must be cheap and canonical to check, and the consensus algorithm has to be rigorous, or else it's too easy to attack the security of the network. DoS, Sybil, 51%, social
108.
by pshc 4y ago
Wouldn't untrusted weird input increase loss and be rejected?
109.
by pshc 4y ago
This writing style is more characteristic of LinkedIn to me.
110.
by pshc 4y ago
> The best way to get someone's password isn't phishing, it's threatening to hit them in the head with a wrench. Crypto has built much to fend off wrench attacks. Multi-sig vaults, timelocks. The first line from the arti
111.
by pshc 4y ago
I presume the stills are taken with two different cameras, film and digital.
112.
by pshc 4y ago
It isn’t equivalent. Optical fiber beats 5G on those criteria.
113.
by pshc 4y ago
Unison has potential to underpin a paradigm like that: https://www.unison-lang.org/learn/the-big-idea/
114.
by pshc 4y ago
"Only my pet solution will work (which is in political gridlock) so give up and don't try to innovate or change anything."
115.
by pshc 4y ago
At least on my machine, Retry never worked.
116.
by pshc 4y ago
The consensus algorithms used by cryptocurrency fit into group C. Smart contracts are programmable incentive mechanisms. I am hoping we figure out how to use them to coordinate productively, rather than for financial schemes.
117.
by pshc 4y ago
One problem with human governance is we can't seem to figure out how to coordinate without burning excessive amounts of energy signalling status and power at each other. What if we could do better?
118.
by pshc 4y ago
Must funnel ever increasing resources into wasteful and destructive conflict. How about using AI to figure out how to cooperate at scale?
119.
by pshc 4y ago
A great game that scratches the Dwarf Fortress itch. My issue with ONI is that I have a great time building a base for a while, and then invariably I fuck up containment somehow, leaking germs/pollution/CO2 everywhere, at which po
120.
by pshc 4y ago
Back in the early days of mobile games (Zynga et al) I worked at one of those shops helping them build a better Skinner Box.

More ›