« first day (4500 days earlier)      last day (717 days later) » 

00:45
@Mitch Well, I found one of those, and it sounded kinda bogus. Like many of their entries.
01:37
@Cerberus I also thought it to good to be true
@CowperKettle Where on the Internet did you find it?
I forgot
Ah
Here
> Andrew S. Curran is the William Armstrong Professor of the Humanities at Wesleyan. His latest books are The Anatomy of Blackness: Science and Slavery in an Age of Enlightenment and Diderot and the Art of Thinking Freely.
This is interesting and weird.
I was using the Bing AI chat, which is using a newer version of Chat GPT.
But it refused to let me type after this response.
01:56
I'm thinking of the unfortunate uncontacted tribe on an island near the coast of India. Once they decide to contact our civilization, it might be too late. We'd say: sorry, we already wrecked civilisation by inventing an AI. Please go back to your island for the next 1000 years, maybe we'll restore the civilization by that time. Sorry for the inconvenience.
The chief will turn his head back to his assistant chiefs, and say: I told you they're nuts. Lets go back to the island.
What do you mean by "we already wrecked with by inventing an AI"?
Wrecked the civilization.
I've just woken up, and make typos.
Feeling neurospicy from coffee
Ah, OK.
Allarin aghrimasin for spotting it.
Is this edit what you intended?
Then the joke makes sense.
One wonders, though, whether an AI would leave such islands alone.
02:07
Maybe AI won't have a motivation of its own.
Man's cerebral cortex does not have a motivation unless driven by lower structures. So AI actors will be driven by people with different motivations.
Well, what is motivation?
Pressure built up inside a volcano can kill people.
A virus can kill people.
A nest of fire ants can kill people.
A tiger can kill people.
It is only a gradual difference.
A gradual indifference can kill people too.
What is man that thou art mindful of him
02:28
". . . That Thou Art Mindful of Him" (also signed as "That Thou Art Mindful of Him") is a science fiction short story by American writer Isaac Asimov, which he intended to be an "ultimate" probe into the subtleties of his Three Laws of Robotics. The story first appeared in the May 1974 issue of Fantasy and Science Fiction and the 1974 anthology Final Stage, edited by Edward L. Ferman and Barry N. Malzberg. It was collected in The Bicentennial Man and Other Stories (1976) and The Complete Robot (1982). == Plot summary == In this story, Asimov describes U.S. Robots' attempt to introduce robots...
@Robusto a bunch of crap words, and then they don't accept perfectly fine ones.
It's why they call it the failing New York Times
@Cerberus 'refusing' has too many implications for intention that are just plain not there
@Mitch Well, one "person" calls it that. I think you know who I'm talking about.
@Mitch The way a jar of pickles refuses to open.
But I'd be curious to hear your analysis of the screenshot I posted.
How do you feel about the AI's responses?
02:44
@Cerberus You broke Bing AI Chatbot?
Or did you just hurt its feelings?
 
1 hour later…
03:58
7
A: Why doesn’t a language modernization initiative adopt pure phonetic spelling?

user6726wʌn ˈɹʷijzn̩ iz ðæɾ ɪndʌˈvɪdʒl̩z ˈdɪfɹ̩ʷ səp̚ˈstæ̃ʃəli ɪ̃ ðɛɹ prʷəˈnʌnsiɛiʃn̩ ʌ wɹ̩ʷdz. In fact, it is extremely difficult to get undergraduate students in a linguistic class to produce an accurate phonetic record of their own speech – it takes a very long time and it is typically very inaccurate...

ɪndʌˈvɪdʒl̩z ˈdɪfɹ̩ʷ səp̚ˈstæ̃ʃəli ɪ̃ ðɛɹ prʷəˈnʌnsiɛiʃn̩ ʌ wɹ̩ʷdz
Copper levels somewhy correlate highly with obesity slimemoldtimemold.com/2023/03/05/…
04:20
Happy International Women's Day
05:19
@Robusto Hard to say!
 
3 hours later…
08:06
Wordle 627 5/6

🟨🟨⬜⬜🟨
⬜⬜⬜⬜🟨
🟩🟨⬜🟨⬜
🟩🟩🟨🟩⬜
🟩🟩🟩🟩🟩
I finally got it. I should have gotten there sooner.
08:31
Wordle 627 6/6

⬜⬜⬜⬜⬜
⬜🟨🟨🟩⬜
🟨⬜⬜⬜⬜
⬜🟩🟨🟩🟨
⬜🟩🟨🟩🟩
🟩🟩🟩🟩🟩
Two letters, three slots, many possibilities.
 
2 hours later…
10:02
Neural net
 
1 hour later…
11:40
[ SmokeDetector | MS ] Few unique characters in body, repeating characters in body (156): who let the dogs out?‭ by Samual Parker‭ on english.SE
12:35
> Ukraine expends an estimated 100 000 artillery shells per month; Russia, an estimated 400 000 shells.
700 artillery rounds per hour, on average
One fired round every 5 seconds.
 
1 hour later…
14:00
@Robusto wait, cloroform? But it's supposed to have an H
@M.A.R. No, colorific was one of the words.
@Robusto the guy giving the directions in Home Alone 2?
@M.A.R. If that guy was Donald Trump, yes.
#Worldle #411 1/6 (100%)
🟩🟩🟩🟩🟩🎉
⭐⭐⭐
https://worldle.teuteuf.fr
🌎 Mar 8, 2023 🌍
🔥 54 | Avg. Guesses: 4.87
⬜🟩 = 2

globle-game.com
#globle
Good guess.
14:19
Very interesting audio discussion about AI: samharris.org/podcasts/making-sense-episodes/…
Wordle 627 5/6

🟨⬛⬛🟨⬛
⬛⬛⬛🟨⬛
⬛🟨🟨🟩⬛
⬛🟩⬛🟩🟨
🟩🟩🟩🟩🟩
Again some country probably in Africa.
@CowperKettle If you'd done a few rounds with the African Countries Puzzle you'd know for sure.
#Worldle #411 1/6 (100%)
🟩🟩🟩🟩🟩🎉
⭐⭐⭐🪙
https://worldle.teuteuf.fr
Phew
Sheer luck
Congrats.
14:28
Spasibo!
Allarin aghrimasin.
> In fact, bees who were allowed to forage with a demonstrator didn't just learn to open the boxes - they learned the same version of box-opening as their demonstrator! Even when some bees discovered the other solution, they still went back to their demonstrator's technique.
> Why is this result so interesting? We based our study design off previous (excellent!) studies done in chimps, vervet monkeys and great tits - and our bumblebees (tiny invertebrates!) performed similarly to these species
> And the results of these studies were used to suggest that these species were capable of sustaining behavioural "traditions" - aka, culture!
Thus, bees have culture transferrable by learning from other bees.
> Culture was once thought to be uniquely human, but has been found in primates, cetaceans, birds, and even fish - with lots of studies to prove it. But only a few studies have considered the possibility of invertebrate culture.
@CowperKettle ??
@CowperKettle I've been saying this all along. Only not in that particular way.
14:47
Daily Quordle 408
6️⃣7️⃣
4️⃣8️⃣
quordle.com
15:04
> Resistentialism is largely a matter of sitting inside a wet sack and moaning.
I did a search on 'resistentialism' and they give it all away in the first hundred hits, like they're explaining the plot of a horror movie to a child by first saying that the heroine survives.
(I still haven't found the original article explaining resistentialism)
That link is very on point for philosophy in that it never actually says what it is talking about.
@Robusto Whew... now I can say things now that it's out.
I really don't think f that as a word (even though a quick google search tells me that it's in a number of reputable dictionaries.
It sounds too much (to me) like a made-up word for a box of Luck Charms cereal... "They're COLORIFIC!"
@Cerberus My first impression was the obvious 'that's awful'.
My second impression was " but isn't that an example of Kant taken to the extreme that people sometimes come up with". Kind of the freshman response "So you're saying that even if it would stop a murder, I should never lie?"
As to the mechanics of how it was produced, I can only make an (educated) guess. It's a guess because It would take a lot to say exactly how the response you saw was formed out of the training data.
I'm actually surprised (and impressed?) by some ChatGPT responses because I find it hard to imagine that there is much training data that looks anything like that. Public blogs are the only place I could think... Frankly Philosophy might be a logical place to find such things but their culture would probably not tolerate that.
On second thought, I bet lying is the one example that people use all the time with explaining Kant.
15:31
@Robusto "Thank you" in Azeri
Azerbaijani () or Azeri (), also referred to as Azeri Turkic or Azeri Turkish, is a Turkic language from the Oghuz sub-branch spoken primarily by the Azerbaijani people, who live mainly in the Republic of Azerbaijan where the North Azerbaijani variety is spoken, and in the Azerbaijan region of Iran, where the South Azerbaijani variety is spoken. Although there is a very high degree of mutual intelligibility between both forms of Azerbaijani, there are significant differences in phonology, lexicon, morphology, syntax, and sources of loanwords. North Azerbaijani has official status in the Republic...
@Vikas That's it? Just a day?
Yup, just a day for half the population.
Human Rights doesn't even get a second thought.
@CowperKettle Dude... there's no transcript? The best part of about writing over speech is that you can skim and skip around.
15:35
Yes, somewhy there's no transcript
Apparently, they wanted us to listen to them.
With listening to three people make speeches is that you have to concentrate on their voices for the full linear time that they're talking and half the words are filler words that don't mean anything but you still have to wait for them to spew them out of their mouths.
Mar 10, 2014 at 15:45, by Mitch
@KitFox Every day is children's day.
@CowperKettle But all three of those guys are big names:
Textbook heresy.
Sam Harris, the interviewer, is known for being a public atheist, and for being skeptical of police brutality against blacks.
15:46
anti-(lies-to-children)
Stuart Russell is famous for being a coauthor of the textbook on Artificial Intelligence that has been used in every US CS undergrad class for 25 years. He is currently working against using autonomous weapons. He is well known for his text book but I feel like he made some important invention in AI that I just can't remember.
I went to the psychiatrist and complained about being afraid of palindromes. He prescribed me Xanax.
@Mitch They do talk like they know something about AI.
Gary Marcus is famous for being a gadfly about AI hype and he supports mixed AI models instead of just probabilistic ones. Both Russell and Marcus have done a lot of research and development in AI. Both are academic.
> Every US CS undergrad class for 25 years.
Cha ching! $$$
@CowperKettle I'm just saying stuff I know about that I think you may not be aware of, but also I don't have to listen to the podcast to say it.
I will go out of my way to remain lazy
@user726941 Normally college textbooks are not particularly lucrative for the authors (it's not like JK Rowling or Stephen King). The publishers certainly make money off them. For this textbook... I don't think the authors could retire from their earnings. (but for this textbook... maybe they could take a ski vacation to Switzerland every year?)
15:58
Yup, somebody made a lot of money 💰🤑
@CowperKettle AN HOUR AND A HALF? I -really- don't have time for that!
@user726941 Yeah... it's awful in a university bookstore. Tuition (in the US) is gouging and then for each class you have to pay -more-? It's not good.
@CowperKettle I'm listening now
sigh
he keeps talking and talking
says lots of good things about rationality
(but of course that's often a way to associate yourself with 'good' thinking while allowing a lot of lapses.)
@Cerberus to continue what I think of a lot of ChatGPT things, I'm impressed by their developers a lot of the features they've added in to the UI in reaction to people's complaints, like the 'references' (adding links to the web to presumably substantiate a factual-looking statement by ChatGPT), and also the 'guardrails' (curtailing conversations where the user or ChatGPT is leading in an unwanted direction).
Mixing in facts with bs always works irl.
To me it seems like there is no way that they could have implemented all those things so quickly... which leads me to believe that the team creating these things have actually been very smart and forward thinking and have already developed a number of good things (UI -and AI) that mitigate all the problems people see. That is, the way they are able to make changes available to the public so quickly is that they were already working on them for some time.
@user726941 People will just fill in the blanks and give a lot of leeway (because they assume agency on the other side).
When in fact there is -no- agency at all.
Just a (very fancy) word sequence correlation calculator and stochastic generator following those correlations. A fancy Markov chain.
They've done their psychology 101 homework
16:14
So what is amazing is all these seemingly intelligent things that it was not intentionally designed to do.
It's designed to produce text that is -like- the text it has seen.
If the combination of words it produces actually forms a statement that matches reality, that is only a coincidence (but a coincidence that becomes more likely given more training data)
I've read that it's the fastest ever growing trend to hit the internet.
That is, the 'learning' model is training on facts and relations of objects in factual terms, but just word frequencies and correlations.
Bigger than eternal September
Welcome to eternal Chat
It can do addition of small numbers correctly (<100) but the common explanation is that in the training data there are enough examples that it is able to predict pretty well what the right digit should be. (ie there's no addition algorithm learned from rules, it just sees examples and while not generalizing the examples it can get close by guess (and guesses correctly with 2 digit numbers just because there are lots of examples of thos e)
@user726941 It's kind of like the problems with copy-pasting from Wikipedia. It's easy and free. But it's not that good.
16:22
Supposedly Wikipedia is getting better. But...
(⁠◠⁠‿⁠・⁠)⁠—⁠☆
Don't forget the lockdown forced everyone to their computers.
Social media popularity went through the roof.
Now something like this comes along to keep the buzz going.
🤔
@CowperKettle something interesting they're talking about in that talk is about the 'guardrails' the rules they're adding in to ChatGPT to prevent it from saying 'bad' things.
Russell mentions that we (people in countries around the world) have been trying to create tax regulations for hundreds of years and there are still loopholes and problems. I read this as meaning that The ChatGPT rules have lots of problems at the moment (it was released with a lot of these guardrails but it was easy to get around them but they've been adding more and more in the few months it has been available)
But it also makes me realize that if ChatGPT becomes a public utility (used as an API in lots of products) that these rules are sort of like a set of constitutional laws with absolutely no oversight from the public.
Facebook and other social media apps limit the language you can use with their own guardrails, and that rule set is entirely hidden within the software.
16:43
How do I convert actual Chinese language to this kind of English words?
xiongdi wo xian shang jin bu qu
17:22
Daily Quordle 408
4️⃣8️⃣
7️⃣5️⃣
quordle.com
@Mitch Yup.
🌎 Mar 8, 2023 🌍
🔥 2 | Avg. Guesses: 6.49
⬜🟨🟨🟧🟥🟩 = 6

globle-game.com
#globle
@Robusto Respek!
 
3 hours later…
20:39
@jlliagre Luck! It was the first guess that I thought was about 25% of the way around the world.
20:55
And there were big countries in the way, whose near edges were way too close.
Daily Octordle #408
🕚6️⃣
9️⃣3️⃣
🕐7️⃣
🟥🔟
Score: 73
And now for my lesson in humility.
21:18
Daily Octordle #408
🕛5️⃣
🟥7️⃣
3️⃣8️⃣
9️⃣🟥
Score: 72
By a whisker...
21:41
Yeah, this one had some misleading clues.
 
2 hours later…
23:19
Ptakha, a Russian rap musician, went on a tour at the front, and visited the frontline. He was recording a video from a trench saying that they "got stuck" there, probably attacked. Then small arms fire was heard, and he fell to cover from fire, and his phone recorded only sounds of bullets and some blurry ground. e1.ru/text/politics/2023/03/09/72117539
He's been out of reach for 12 hours.

« first day (4500 days earlier)      last day (717 days later) »