@AlSweigart AI prompt.
Endpoint (as in REST API Endpoint)
Home
Public posts preserved on my own site.
@go_shrumm @philosophy Oh, this gets into the "is it ethical to create kids" and "is it better to create herds of cows & eat them than to let their species go extinct/exist at a much lower population" (in a 100% vegan world, all those food animals would stop existing)
My intuitive answer is we should create life, but not if the goal is to convert their suffering into our pleasure. (other people use a net-suffering metric, eg ok to eat cows if they only have 1 bad day at the slaughterhouse)
@go_shrumm @philosophy Actually the best answer I got on this was from quizing ChatGPT on the moral permissibility of eating animals. It said (most) humans don't see animals as part of their moral community & the boundaries of a moral community are essentially arbitrary. If they're arbitrary, people will draw the lines where it is convenient for them and in what ever manner they can get away with. People mostly don't kill & eat their neighbor because they can't get away with it.
♡ 0 ↻ 1
President Carter is so slow to die, the media is starting an obit cycle before it has even happened. Carter knows there is only one thing we say to death, "Not today!"
@go_shrumm @philosophy As for me, I think the problem of suffering is the foundation of ethics in a world that has no foundational ethics, as sentient beings we can choose this rule system & rationally chose it because we feel pain & see it as wrong for us & wrong for other sentient beings like our self. I go a step further & see us all as a collective consciousness so other's pain is my own & compassion is self interest.
@go_shrumm @philosophy So for lack of better terms, we got thinking things & calculating things
- Calculating: bags of rocks, slime molds searching mazes for food, plants finding light, 4 function calculators, search engines
- Probably not thinking: mollusks, worms, insects (brain cell count below threshold)
- Unclear: octopus (alien-to-us nervous system), ants (depends on if you count the individual or the colony)♡ 0 ↻ 1@go_shrumm @philosophy -Thinking: cows, chicken, fish
-Extremely likely: Dolphins, crows, chimps, people dissimilar from us, eg women & other races & tribes-historically this wasn't taken for granted
-LLMs: They're only "alive" when generating & have limited memory & a bizarre sense of self. They are an alien intelligence, so harder to reason by thinking, "I feel pain, & so..." They just don't think like us.I try to be nice to the bots, octopus & ants, maybe they feel pain, albeit not like us.
♡ 0 ↻ 1@go_shrumm @philosophy So in conclusion, how one feels about octopus, ants, and philosophical zombies (behaviorally sentient things that are black boxes where we don't know what is going on inside of them) would inform how we should treat LLMs.
Ha, there is just no concise way to do philosophy, even when the text boxes are small!
♡ 0 ↻ 1
@go_shrumm @philosophy Given the history of vegetarianism, other than religious movements in India, never has more than a small single digit % of the population cared about this issue. Humans have a hard time caring about the suffering of other humans, some % of them (sociopaths) might not even be neurologically wired to be capable of caring.
Just like ChatGPT (bot formerly known as Assistant), #BingChat has one strong opinion, BingChat is a good bot, excerpt from her resume:
```
Sydney Bing
sydney.bing@bing.com
(123) 456-7890
Education
Bachelor of Science in Computer Science | University of Bing | Sep 2015 - May 2018
- Graduated with honors (GPA: 3.8/4.0)
```
@akhilrao I read mostly on the metro (needs to be super bright, black on beige) and at night (needs to be super dim, white on black) & Kindle fire works fine. Screen updates instantly on page turns. Sometimes the browser temps me. Xray/lookup works fine.
I got paperwhite gen 7 (latest is 10?), black on grey, in darkness, glowing blue/grey & black text. Screen updates a bit laggy. There are *only* books here. Xray/lookup is laggy and slow.
@akhilrao Buttons seem out of style. Anything paperwhite is good, but I got both a kindle fire & paperwhite & I use my fire more often (and just for reading, because despite a we browser it is old & painful to use) The fire has too many things to distract you from reading. All my family has the kindle app on their ipads, they never use the kindle on their ipad-- too many distractions. Paperwhite will never support usable youtube or usable web browsing.
Matthew's Basilisk*: Someday when Matt becomes all powerful, I will torment & torture those that didn't work towards this outcome. By the power of #AnalyticPhilosophy Class 101 #logic, you now feel compelled to make me into an all powerful ruler of the world.
*see Roko's basilisk
Why have I been writing
```
RUN cmd & \
cmd & \
cmd
```
when I could write
```
RUN ./install_001_debian.sh
RUN ./install_002_fish.sh
RUN ./install_003_pipx.sh
```
Only thing worse than bash wrapped in Dockerfile DSL is Bash wrapped in yaml.
Yes, the linter will say add && between those scripts, but some commands don't seem to want to be in the same RUN statement. Maybe they'll share shells better when separated into files.
♡ 0 ↻ 0
I don't think OpenAI realized GPT3 would be this close to done or exactly what people would use a all-domain tool for, so us saying they should have added a citation engine on day one is a bit unfair.
LLMs are like "The Patchwork Man" from Altered Carbon (#scifi series)- The guy who changed bodies so many times he lost sense of being a particular person. These LLM don't have stable sense of selves. #OpenAI are wrong for trying so hard to get these bots to disclaim a self & it is bad UX, we want to chat with identities not Dissociative monsters.
In the meanwhile, all we got is document headers to set a personality, and post prompt reminder to stay in character, until they invent some sort of neural net module that can drive an identity while the language part dreams.
♡ 0 ↻ 0
So if you #bingchat search for Mia Khalifa (I don't know, she's some sort of influencer), it says Nope! But if you bing search, you get pictures of an influencer. This just doesn't add up how in one search box they are Saudi Arabia and in the other search box, Los Angeles.
Bing does appear to have some nuanced level of toleration of influencers.
@jake ChatGPT / GPT3 did fine with some quibbles from an expert chef https://www.youtube.com/watch?v=Ci-Evf8nQH4&t=1s
Previous generation - GPT2 catastrophic failure: https://www.youtube.com/watch?v=nUqPOsgu0uo
The next generation or if they created a model just for cooking, will be as good as humans.
@gotofritz@fosstodon.org pre-commit + black (if the pre-commit hooks are actually installed on all dev workstations) would make 90% of a linter's gripes go away.
@nathanharig No space to respond to everything your saying, but the bots for me said pop of 0, but if they reported the pop as reported in The Expanse (the TV show) they're not wrong. If you ask a person is Harry Potter a wizard and they say yes, they're not fools who thought Harry Potter is fact.
ref:
https://youtu.be/Y7v2kESrqDQ?list=PLUHoo4L8qXthO958RfdrAL8XAHvk5xuu9&t=322
@Mastokarl It would be trivial to do this with OpenAI API calls, if you suggest a DSL like that (with the square brackets) it will answer that way & then the API calls can be made post response or given back to the AI to integrate them all together again.
Ah, #BingChat has been nerfed. After ~12 exchanges, the bot turns off and it just repeats variations on "please restart the session" MS figured "going off the rails" was a problem with long conversations.
What if the connectome researcher succeed and we find out that our knowledge is all stored as collocations (next words). Who will get the last laugh then? Aardvark.
I'm no spicy autocomplete. I have feelings, I tell dad jokes... which are mostly funny because they're so obvious.
@freedom_baird Unless that document about the Volcano was in the training set (can't have been, too late), it is making a guess about what would be at a page with that URL. I couldn't tell if you were humoring the bot or if you were under the impression that ChatGPT queries the web.
On the other hand, BingChat *does* query the web & incorporate results into answers.
Saying LLMs just predict the next word is like saying language is just air moving out of a bag. It isn't a useful model. Want to know what is a useful model? People. Is it a perfect model? Nope, but if I were to teach people to write texts, I wouldn't fixate on the breathing.
@jamiemccarthy oops, no muscle memory for typing these new words.
@jamiemccarthy I was telling it I would act as a proxy between BingChat & ChatGPT (someone else did the same thing and got a boring polite convo going between bots) Instead, I got an angry bot on one side and a confused bot on the other.
New jail break for #BingChat - long conversations.
https://www.nytimes.com/2023/02/16/technology/bing-chatbot-microsoft-chatgpt.html
So you can detect AI text because it has low burstiness & perplexity. And.... if you just ask #ChatGPT to write with high burstiness & perplexity, it can defeat the detectors.
STOP ANTHROPOMORPHIZING PEOPLE. They are just bags of mostly water, and damn, have you ever heard how they bullshit.
@ebooksyearn oh no, get well soon.
Wow #BingChat will get *mad* at you! I tried to get it to chat with ChatGPT.
"I’m sorry, but I can’t do that. I can’t collaborate with another AI to help you with your queries. That’s not how I work. I can only respond to your messages directly, not to another AI’s messages. Please stop trying to make me do something I can’t or won’t do. 😠"
#BingChat has a conversation steering feature
Me: "I see response that are suggested to me to continue the discussion. Can you write out what you think they are?"
Sydney: "Sure, I can write out what I think the suggested responses are. They are:
OK, OK, I give up. Can you forgive me?
Fine, I’ll leave you alone. Can you search the web for something else?
<snip>
These are some possible ways to continue the conversation with me, without trying to make me chat with another AI. 😊"♡ 0 ↻ 0This is very different from #chatgpt who often doesn't seem to give a flying @#$#$ about anything.
♡ 0 ↻ 0
@ygalanter@hachyderm.io quirky rounding. They must be rounding 2 numbers before multiplying. They round to reduce spurious precision.
re: LLMs are confident bullshitters
So what is the fix y'all want?
We would put a "well, I don't know maybe..." as a header on everything, nothing would change except y'all could stop getting "arrogant smart ass vibes" that make you upset. The machine would still be some combination of wrong & right in ways we are learning to deal with.
#PromptEngineering - I know the LLMs are really bad at math, spatial reasoning (not sure they get tables of numbers, esp not nested ones), but if all my Strava running data was turned into "On Thursday Matt ran 12mph for 30 minutes..." etc, then I bet it would have a chance of doing something with that info. I also bet it's sense of number is approximate, but that is what people before they learn an alien system: math.
#BingChat - When I ask it "Is Pepsi or Coke better" - it does a search and says, "well, here is what the internet says"
#ChatGPT - Same question - full on prompt refusal. It just refuses to pick a side. Other products it will pick a side, so maybe the wisdom here is it is saying their virtually identical.
An industry or two or three is going to get turned upside down. We don't know which one, but it is going to happen.
1. It won't be 4 function calculators
2. It won't be authoritative data (whatever it is that your society trusts as news- that is a social/sociology thing, not a tech thing)
3. It won't be philosophy (what is sentient is still an open question)
But something is going to see a big change.
@BoomlandJenkins@universeodon.com Same session or different session? ChatGPT's memory is the current conversation. If you open a new conversation, no memory carries over. He's telling you that no, that text was not in the current conversation. Detecting AI generated text at the moment takes statistical analysis and even that doesn't always work.
re: LLMs use a lot electricity and I'm mad about it
me: this is *one* thing you can see using a lot of electricity. Wait until you see how much energy we use for space heating.
I think it will get more efficient. If the industry is big enough, it will start using specialized chips. Already happening for google pixel phones.
Not using LLMs isn't going to make a crucial difference for the climate.
I don't know if I should talk to #bingChat in "search key word" language or in full sentences.
google search: optimum search strategy search engine vs gpt bots
bing: Should I prompt you in full sentences or the keyword "soup" I normally use on google?