itsadok
Joined 16 years ago
Karma
185
Submissions
46
Member since
May 13
Recent submissions
Comment · 1 year ago
Comment · 2 years ago
In your assess_output_quality function, you ask the LLM to give a score first, then an explanation. I haven't been following the latest research on LLMs, but I thought you usually want the explanation first, to get the model to "think out loud" before committing to the final answer. Otherwise, it might commit semi-radndomly to some score, and proceed to write whatever explanation it can come up with to justify that score.
Comment · 5 years ago
This is kind of like how http sites look totally fine in most browsers, but an https site with a self-signed certificate will cause a "DANGER! ENTER AT YOUR OWN PERIL" screen to be shown.
Comment · 6 years ago
These were literally the first 3 questions on my review queue:
* https://stackoverflow.com/q/59344615 - A question by an absolute beginner, trying to do something that they have no clue how to start with. Already closed.
* https://stackoverflow.com/q/59341242 - A question about parsing a JSON response with jQuery. Two votes to close. The asker clearly does not know the word "parse".
* https://stackoverflow.com/q/57969318 - Someone trying to figure out an error message they're getting with kubernetes. This is exactly the kind of thing you get at the top of your google results when you hit the same error, and with one more vote to close, it will be forever locked with no useful information. Some asshole even downvoted the one answer that is there without adding any comments.
None of these questions are good, but they could be made better, and they all represent people with real problems that deserve help. Getting mad at people for "being lazy" (because if I, the expert, could easily find the answer to this, then why didn't you?!), is not productive.
Here's what I don't understand about all these SO deletionists: how is closing the question helpful in any way? If you don't find the question answerable, then don't answer it! But why block other people from trying to help? It's not like you're somehow "teaching" these people how to ask by blocking them. The user from question 59344615 (which got closed) did not post another question with better details. They just left the site, one more developer that doesn't have anywhere they can ask newbie questions. It sucks.
Comment · 8 years ago
The last time I evaluated GitLab, you needed the paid version to use features that I considered pretty basic, like merge request approvals and multiple code reviewers[1]. I'm wondering now if the GNOME people consider these unnecessary, or if I misunderstood what was possible with self hosting.
Comment · 9 years ago
Are there any downsides to multiple-licensing? What happens if I say "This software is released under every license listed in https://opensource.org/licenses/alphabetical"?
Just curious.
Comment · 9 years ago
Not OP, but some cool features I haven't seen in Linux:
- Quadruple-click smart selection
- "Selection respects soft boundaries" for selecting within a tmux pane
- "Rum coprocess" on keyboard shortcut, which allows doing stuff like http://brettterpstra.com/2014/11/14/safer-command-line-paste...
I just checked, and gnome-terminal doesn't even have split-pane. What do you guys use, anyway?
Comment · 10 years ago
While people still drive themselves, there's a huge incentive to come to work early or late, to avoid traffic. I don't see that incentive increasing with autonomous cars. If anything, I predict that more people will join rush hour traffic if they don't need to do any actual driving. However, this might be mitigated by the better driving of autonomous cars.
Comment · 10 years ago
This is similar to the "Contamination" effect, of how completely false statements can affect a person's judgement, even if they are told the statement are false.
http://lesswrong.com/lw/k3/priming_and_contamination/ http://lesswrong.com/lw/k4/do_we_believe_everything_were_tol...
Comment · 11 years ago
I'm pretty sure everything you used is already supported: http://buildbot.pypy.org/numpy-status/latest.html
Comment · 11 years ago
A common warm up question in a coding interview is to write (on the whiteboard) a program that prints the Fibonacci numbers in sequence. Arguably, you can write this fraction and claim this answers the question, though I don't know any programming language where this would actually work without a bunch of more code around it.
Edit: other comments have examples in Python, Clojure and bc.
Comment · 12 years ago
Obligatory: http://www.paulgraham.com/hundred.html
Comment · 12 years ago
A slightly easier to read version here: http://le-cretin-transnational.ch/wp-content/uploads/2014/04...
I would be interested in an AI-only product that would help me learn to passably immitate various English accents, like Australian, Irish and so forth, for fun. I know that ChatGPT Voice can do accents pretty well, I've been wondering if it would also be able to help me with mine, but I haven't tried it seriously.
[1] https://www.boldvoice.com/frequently-asked-questions