The SSDs in my storage cluster that I bought for a bit under $800 a while back are now over $1300, a >50% price increase.
I'm only at 60% ish usage so hopefully I can keep my rate of data growth in check until the bubble pops. Otherwise I'm either gonna have to pay scalper prices for more NVMe or add some spinning rust back in.
At the rate things are going I've probably got a year or two before things get tight.
Just finished "The Word for World is Forest" by Ursula K. Le Guin. Can't believe I didn't read this one earlier, and this strengthens my resolve to finish off the rest of her stuff I have yet to read sooner. I think it benefits somewhat from having read it after "Four Ways to Forgiveness" which gives more of the Hainish context. Certainly none of the blurbs I had read about it did it any measure of justice, which is one reason I hadn't prioritized it. More than being about colonization, it's about a solution to the paradox of tolerance, and both the price and imperfections of that solution. As usual with Le Guin's science fiction, it's a rich companion to anarchist thought.
I think the typical objection to seeing it as an answer to the warlord question would be that it serendipitously positions the indigenous population with more power and a less ruthless opponent than in the imagined scenario, and it uses the League of Worlds as a sort of deus ex machina to foreclose further retribution. Ultimately that's why I think it's more about the paradox of tolerance than anything else, but I also think in regards to the warlord problem that we are too quick to underestimate just how numerous and enthusiastic the opponents of a warlord might be, and to overestimate the strength of technological weapons wielded by frail (and psychologically unarmored) humans.
In any case, Le Guin gives this book's alien humans yet another fascinatingly credible capability, and getting to see the introduction of ansible technology with all its implications is pretty cool too. Maybe not
Imagine ChatGPT but instead of predicting text it just linked you to the to 3 documents most-influential on the probabilities that would have been used to predict that text.
Could even generate some info about which parts of each would have been combined how.
There would still be issues with how training data is sourced and filtered, but these could be solved by crawling normally respecting robots.txt and by paying filterers a fair wage with a more relaxed work schedule and mental health support.
The energy issues are mainly about wild future investment and wasteful query spam, not optimized present-day per-query usage.
Is this "just search?"
Yes, but it would have some advantages for a lot of use cases, mainly in synthesizing results across multiple documents and in leveraging a language model more fully to find relevant stuff.
When we talk about the harms of current corporate LLMs, the opportunity cost of NOT building things like this is part of that.
The equivalent for art would have been so amazing too! "Here are some artists that can do what you want, with examples pulled from their portfolios."
It would be a really cool coding assistant that I'd actually encourage my students to use (with some guidelines).
#AI #GenAI #LLMs
There are finally real USB3 (not Thunderbolt, not USB4 PCIe Tunneling) 10GbE NICs!
RTL8159 chipset-based, it uses the generic cdc_ncm driver on Linux.
Also seems to work on macOS and iOS.
TCP throughput between GPD Pocket 4 and MacBook Pro M1 seems a bit low at ~6.2 GBit/s, but that might also be due to the connected devices (both sides are using the same adapter).
I don't have another 10GBit/s device here right now, sorry.
Don't have a USB 3.2 Gen 2x2 de…