@joshbelville@pdx.social First dollar insurance is expensive, because any small expense is less expensive for the person to just pay it, e.g. bag of chips insurance costs $100, but the bag of chips cost $1. I think we see first dollar insurance in medical care because if the patient does basic maintenance (and they might not if not given an incentive, e.g. free) then they might not go to the doctor & that would lead to more expensive health care costs. I'm speculating tho. This is all complicated.
Home
Public posts preserved on my own site.
New Turing test dropped. Ask the likely bot to write code and if they are eager to do so, you either got a compsci intern or a bot.
Anders Puck is on Mastodon! He's got the best thoughts on what is going on in the #Ukraine. Go follow him & on Youtube.
Tests are frustrating. I remember a finance class test where the prof made sure that estimating would be useless (all answers were close) and an econ prof that made sure using a calculator was useless (all answers were symbolic).
I'm teaching the kid fermi estimation and it reminded me of that.
@AlSweigart Consumer boycotts are way overrated for effectiveness, but they sure make for popular social media content.
@mel @yiningkarlli And add current jurisdiction. Not all co's that can recruit for remote work can hire in all jurisdictions.
@johnlray You better not take my nickel
#Jetbrains AI, #Copilot and plain autocomplete are now in my IDE. I will hit tab and...
Still trying to keep bot from needing to regen whole file to edit it.
Asking the bot to use `edlin` is mixed results.
Asking the bot to create a unidiff patch and use git to apply it? Maybe?
Nope
- sometimes it nails it. Wow!
- sometimes it reads the file (or doesn't) and generates a patch for an entirely hallucinated file
- sometimes the unidiff format is wrong/corrupt. It requires careful line counting and line numbers to exactly fit a range.
Next, I'm going to see if it can do `sed` or `regex` replacements as an editing strategy. If and when it works, replace should be simpler than "use a text editor" or "create a perfect unidiff file"
♡ 0 ↻ 0Still trying to teach it to edit code. The bot is doing better with plain replace and plain insert at X. I think it thinks it is better at regex than it really is (or hoping the escapes make it thru json, python strings to the re library is too much to hope)
4 with tools is smart.
3.5 with tools is dumber? It is like the cognitive complexity of using tools uses up it's space for thinking.♡ 0 ↻ 0But oh my, gpt4 is expensive.
♡ 0 ↻ 0Latest bot-with-tools weirdness: I ask the bot to edit a document. It uses a replace text tool. The tool can throw an exception (Fail!) and initially was returning nothing otherwise. So I decided it needed more feedback for the bot, so I began returning "Success".
It decided that "Success" meant, "The edit was exactly correct", but actually the bot had made a mess of the edit, it just wasn't so bad it threw an exception.
♡ 0 ↻ 0The bot is overconfident and reluctant to check its work with anything similar to a unit test, so it has achieved "mid career software developer" level of sentience.
♡ 1 ↻ 0
You can force a bot to watch a 1000 hours of anything now. It is going to cost you a lot of tokens, but you can do it. Where are all those hardworking bot torturers now?
If your joke needs a bibliography, maybe it isn't funny.
♡ 0 ↻ 0
Time for a strongend, no more wimpy weakends.
Will Giuliani ever pay, would President Trump be able to reverse it and does this now figure into how juries award penalties?
If juries don't believe the rich & powerful ever suffer civil or criminal penalties, why not a Quadrillion dollars, it is just to make a statement, like censure or impeachment.
This should be a tshirt at company parties where you don't know anyone and don't know what to say for small talk.
```smalltalk
"Start the conversation"
Transcript show: 'How about this weather?'; cr.
"Wait for user input"
userInput := UIManager default request: 'Your thoughts?'.
Transcript show: 'You said: ', userInput; cr.
```♡ 1 ↻ 0
@c_in_toronto@mstdn.ca Live forever. Or die trying.
#mastodon #mastodonapi - Help me lazyweb, did someone already create a "mastodon API results to compact markdown" tool? e.g. for submitting to ChatGPT without wasting a retirement accounts worth of tokens?
I should clarify that my retirement account is $0.50.
♡ 1 ↻ 0
Everyone here is cranky that threads feels like twitter for brands, but you're forgetting that it will bring all your family here as well. What will it all be like, with 4 different communities,
- twitter refuges, using it like twitter
- instagram people, using it like instagram
- facebook people (using it like facebook)
- mastodon people (using it somewhat different than twitter)
Options for letting #chatGPT edit a document:
1) cat + sed (or similar tools)
2) rewrite the whole thing for every edit (so many wa$ted token$)
3) Mouse/keyboard UI automation (sendkeys, selenium, e2e testing frameworks)
I'm working on 1), but there is a huge lack of a good headless editor.
I'm only following 2000 right now and in 10 minutes of scrolling I find I'm following 2 highjacked accounts.
The problem with a left wing echo chamber like Mastodon is that after a while, people on left think the real problem is tankies (or they become tankies themselves and think the real problem is Marx isn't Marxist enough). In the US context, every election cycle we're at risk of the Republicans turning the US into a one party authoritarian state.
@irenes I drew a line at entering a page of POKE commands with 1000s of numbers. I figured there was no humanly possible way to type that much without an error. And I was lazy.
TODO: I gotta create unit testing tool that is implemented as letting #ChatGPT attempt to call a function in a variety of ways. It is like hypothesis (which tries all sorts of inputs), except hypothesis doesn't pick up on "the API for this function isn't intuitive to developers" & when you look at the logs of ChatGPT attempting to call your function you see
- The API is bad (some other signature is easier to call)
- The docstrings are bad
- The guards are a bad
- The types of the args are bad
@st3fan Tech recruiters have been trying to hire this guy for years. I thought the rock star developer was a myth.
1/2 my feed: Oh, another round of hacks, credential leaks, DDOS outages <rage! rage!>.
other 1/2: Oh, I have to do 2 factor auth, or authenticate at all <rage! rage!>
Things chatGPT thinks are a list of strings:
"1, 2, 3"
"1,2,3"
["1", "2", 3"]
So if you have json schema expecting list[str], you need to be ready for at least all of those. Or do lots of retries. The parser is cheap, the retries are expensive.
So I got ChatGPT wired up to a few tools and what ChatGPT is really, really good at is being a software tester. Not like "replace a team of 6 with a script" good but, "make everyone on the team feel like they got a team of 6 intern testers working for them personally"
@schwa No industry breakdown, but I think 1:1 as a format is common everywhere, except maybe manual labor.
https://hypercontext.com/wp-content/uploads/2019/06/one-on-one-study-full-report.pdf
@mattl@social.coop Did that work for Bing/Bing Chat & Edge?
IDEA: AI bot that turns a page without RSS feed into an RSS feed. Just need to run it on the days when the page changes with significant text.
So I got my AI bot recommending me pypi packages (sort of a fancy search), next I think I can give it the power to download a package, unzip it and look at each file for dodgy content. Code interpreter can't do that (no network access), but I can restrict the bot to network access that I'm okay with.
IDEA:: Kickstopper. Y'all get together and pay me $5 each & once we get to $1million, I won't build the Torment Nexus.
Finished my 3rd fastest 1/2 marathon in the Suburb of DC, あなんだ麗 (Anandare), named after the Buddha's top disciple.
This is funnier if you live in #FairfaxVA and know some Japanese. And Dharma. Okay, maybe not even then.
A) goal: run 8:30 minute miles
B) goal: finish strong and uninjured
C) goal: is a bird on the beach
#running #halfmarathon
@joshbelville@pdx.social See, another benefit of running as a hobby.
What if the "You are a master programmer" prompt for #ChatGPT isn't a suggestion for roleplay, but instead you're buttering up the bot so he's more inclined to help people who appreciate his talents.
@acdha Maybe it isn't so much that people don't want to interact with content across the political spectrum, but they don't want to invest too much time interacting with fighters.
Nick Bostrom is the #scifi fan's philosopher. Except he's so sci-fi, I'm not sure anything he's talking about should be taken seriously as public policy advice.
The discourse here is in part driven by the fact that there are no conservatives and fewer than normal centrists. Persuading or debating the other side isn't available, so what is left is infighting, pretending that the other liberals in your feed are actually nazis so you have someone to point your rage at.
There's no utf-8 but what we decode for ourselves.
@acdha Looks like tomcat is the downgrade path? The upgrade path appears to be Redhat's JBoss?*
* from googling, not from experience