barrkel

22 ms·

451.
by barrkel 2y ago
o1-mini is a small model (knows a lot less about the world) and is tuned for reasoning through symbolic problems (maths, programming, chemistry etc.). You're using a calculator as a search engine.
452.
by barrkel 2y ago
It's definitely blurry, because as the price escalates companies will want a human in the relationship for multiple reasons: price discrimination / market segmentation, onboarding (perhaps with professional service for configurati
453.
by barrkel 2y ago
Value added tax is a tax on the difference between the purchase price of the inputs and the sale price of the outputs; that is, it's a tax on the value that company specifically adds. The way it works in practice is VAT is added to the
454.
by barrkel 2y ago
SF has always been about the present time of the writer, and is usually most interesting when it perturbs some element of reality to expose something interesting about the present.
455.
by barrkel 2y ago
There's the raw talent, and then there's everything else: occupying powerful positions near the top of the hierarchy, the social network, the experience, the track record, the lessons learned from mistakes made, and so on. There&#
456.
by barrkel 2y ago
The point about data being identified by name and not location is quite strong I think. It pushes concerns that lots of applications would have under conditions of high latency and intermittent connectivity to the network itself, rather tha
457.
by barrkel 2y ago
Consider a society that has left Earth. If light takes minutes to travel between locations, is a chatty protocol what you want? Or do you need a whole host of different application-level solutions to this problem? There's a different l
458.
by barrkel 2y ago
If you have a lot of files, consider find piped to xargs with -P for parallelism and -n to limit the number of files per parallel invocation. Only a tiny bit more complex but often an order of magnitude faster with today's CPUs. Use -p
459.
by barrkel 2y ago
Knowing the specific multiplies and QKV and how attention works doesn't develop your intuition for how LLMs work. Knowing that the effective output is a list of tokens with associated probabilites is of marginal use. Knowing about rota
460.
by barrkel 2y ago
Ollama doesn't have a UI.
461.
by barrkel 2y ago
Why would someone expect interacting with a local LLM to teach anything about inference? Interacting with a local LLM develops one's intuitions about how LLMs work, what they're good for (appropriately scaled to model size) and ho
462.
by barrkel 2y ago
I don't think they're quite glue people in the way that Schmidt seems to have meant, but they're definitely people on the edge of teams that interface between different groups of people - PMs interface with sales, support, cu
463.
by barrkel 2y ago
I'm going to try out the new travel mode on a transatlantic flight tomorrow - mildly excited by the possibility of escaping the claustrophobic confines of 11 hours in an airplane.
464.
by barrkel 2y ago
I got a Quest 3 recently, and the necessity for side-loading apps onto it to make it useful is kind of puzzling. SideQuest is almost mandatory. You can't easily copy files on or off it to network locations without CX File Explorer, sid
465.
by barrkel 2y ago
Product managers, technical program managers. You know. Glue people.
466.
by barrkel 2y ago
The new hotness was C++. Object orientation was the AI of its day. TP jumped on it so fast with 5.5 it ended up throwing away that object model for Delphi a few years later. TP had little impact on Windows, I agree. TP was for a particular
467.
by barrkel 2y ago
> It did this using highly-optimized code, written in assembly language, by very skilled people. "People" here meaning Anders Hejlsberg almost single-handedly.
468.
by barrkel 2y ago
CORS - or specifically, not permitting cross-origin requests by default - is for preventing CSRF, cross-site request forgery. In particular if the user's credentials (cookies) are passed along with a request to a third party site, then
469.
by barrkel 2y ago
LLMs aren't good when you drift out of the training distribution. You want to be hitting the meat in the middle and leveraging the LLM to blast through it quickly. That means LLMs are great for scaffolding, prototypes, the v0.1 of new
470.
by barrkel 2y ago
Developer productivity doesn't map very directly to compensation. If one engineer is 10x as productive as another, they're lucky if they get 2x the compensation.
471.
by barrkel 2y ago
How do you think it makes people who write ("generate" cheapens it) the interesting responses feel? Answering people's questions has two main rewards to the writer: it feels good to help someone else, and it boosts your ego t
472.
by barrkel 2y ago
New account, banal observation, invitation to interact. It's a pattern I've seen recently on what are apparently LLM-powered engagement bots on Twitter.
473.
by barrkel 2y ago
The Pathé newsreel in the video embed at the bottom of the page explicitly states that milk is tasty "when you know how to drink it" while showing a man pouring almost certainly alcoholic liquor into the milk.
474.
by barrkel 2y ago
For sure, they are good for associative and analogy searches for well connected points in concept space, but leaf nodes are totally pulled out of the ether. E.g. you can get great translation of source code from one language to another, but
475.
by barrkel 2y ago
Search requires a huge and ongoing capital investment. Keeping an index online for fast retrieval isn't cheap. LLMs are not tools for search. They are not good at retrieving specific information. The desired outcome from training is no
476.
by barrkel 2y ago
What makes you think a website with "AI" is a big product? IMO AI is positioned to be a commodity, and that's how Meta is approaching it, and of course doing their best to make it happen. I don't think, on the basis of w
477.
by barrkel 2y ago
Rote means learning by repetition.
478.
by barrkel 2y ago
What do you call it when you remember things so you can repeat them but you can't generalize? E.g. if you learn a poem or phrase in a foreign language, but can't reuse the words in different contexts? Or being able to recite a rul
479.
by barrkel 2y ago
The Swiss vignette doesn't come close to covering the actual cost of the motorways. 10 million vignettes sold, 40 CHF each, ~400M CHF gross revenue. https://www.swissinfo.ch/eng/business/what-you-need-to-know-
480.
by barrkel 2y ago
The primary time-saver of a train is you can work on the train.

More ›