« first day (5072 days earlier)      last day (144 days later) » 

CowperKettle
CowperKettle
01:58
Word of the day: cryopeg
user image
CowperKettle
CowperKettle
02:35
AI begins its ominous split away from human thinking - is this really so, @Mitch?
Cerberus
Cerberus
@CowperKettle GPT has always approached things the same way Deepmind's AIs have, by infinite trial and error.
The author says that is 'new' for the latest version of GPT, which it isn't.
CowperKettle
CowperKettle
So I thought..
Cerberus
Cerberus
> Loz leads the New Atlas team as Editorial Director, after nearly two decades as one of our most versatile writers. He's also proven himself as a photographer, videographer, presenter, producer and podcast engineer. A graduate in Psychology, former business analyst and touring musician, he's covered just about everything for New Atlas, concentrating lately on clean energy, AI, humanoid robotics, next-gen aircraft, and the odd bit of music, motorcycles and automotive.
Conrado
Conrado
They keep on saying they've woken up the Macrobes.
CowperKettle
CowperKettle
02:55
Ludwig Wittgenstein's Philosophical Investigations - Audiobook
►
I'll put this on my jogging mp3 player.
Sounds like a nice song for running in the park
Cerberus
Cerberus
How far can you run until the song is finished?
 
1 hour later…
CowperKettle
CowperKettle
04:17
> Transformers and cortical waves: encoders for pulling in context across time. Authors suggest that neural cortical waves could implement a similar encoding principle to transformers' encoding vectors. arxiv.org/abs/2401.14267
@Cerberus Probably 70 km or more )))
I would fall of exhaustion far earlier
My longest run thus far has been 30 km.
Cerberus
Cerberus
Extremely long.
CowperKettle
CowperKettle
04:32
It took me more than 3 hours to run for 33 km.
There's a record on Strava, but I don't recall the exact date.
Cerberus
Cerberus
04:45
Very impresisve.
 
3 hours later…
Xanne
Xanne
07:35
Wordle 1,199 3/6

⬜🟩🟩⬜⬜
⬜⬜🟨⬜⬜
🟩🟩🟩🟩🟩
Vikas
Vikas
07:59
> For most pictures, hang them so that the center of the picture is at eye level, which is usually around 57 inches from the floor. Larger pictures, you can hang them a bit higher, so that the top of the picture is just above eye level.
Is the text about "Larger pictures" poorly written here? It says "bit higher" which I assume - bit higher when compared with smaller ones - so how would it end up having the top of picture around eye level? Wouldn't it make the center above eye level instead?
 
5 hours later…
SmokeDetector
SmokeDetector
12:46
[ SmokeDetector | MS ] Manually reported answer (93): Meaning and usage of the phrasal verb "to bridge out"‭ by Brian Phan‭ on english.SE
Mitch
Mitch
13:14
@CowperKettle Yes, kind of maybe, but really no, not really.
1) the title might as well be click bait for a listicle ... "Here's 10 ways that AI will strangle the life out of you without you realizing it. #7 will shock you (literally!)"
2) the guy writing it and the (online = really easy to produce) magazine I've never heard of before, but minimal googling says that they're not experts in the subject.
3) I read the article and could give a play by play for every sentence but Yay You're lucky I'm not going to bore the crap out of everyone with that (even though it is a mental exercize I did when reading it)
4) Some sentences say things I agree with (and wish I could say) and other sentences are the usual misunderstanding of AI and 'what things mean' and other just plain wrong.
5) a lot of AI talk (and this article in particular) is simple analogical reasoning where the reader has to come up with the analogy (ie the reader is filling in the gaps in the direction of their innuendo but they're not saying what their innuendo is, so they can't be held to making a definitive statement..
OK I'll do a few sentences.
The title: 'ominous split' is heavy handed editorializing/click bait "Oh No! Unspecified AI Fear Mongering'. A more factual but less flashy title is "What's new in ChatGPT o1, just released".
First he introduces things by talking about chess and go and the AI methods for them (standard AI 101, good things to know about AI history)
"ChatGPT and other Large Language Model (LLM) AIs, like those early chess AIs, has been trained on as much human knowledge as was available: the entire written output of our species, give or take." That's patently false. Not everything is on the internet and not everyting on the internet is publicly available (which is all that OpenAI (and all the other LLM makers have access to).
Robusto
Robusto
13:29
@Mitch Not everything on the Internet is accurate or worth referencing, either.
Mitch
Mitch
"But LLMs specialize in language, not in getting facts right or wrong" - this is something that is very true and not said enough by AI writers. LLMs don't compute on facts, they just compute on word sequences (which often include facts) and probabilistically will return facts. Things that people have unfortunately labeled 'hallucinations' are simply errors. That is, a word sequence produced that doesn't correspond to a fact. 'Hallucination' is just a fun way to anthropomorphize LLMs.
@Robusto Project Gutenberg is entirely fiction.
Also, I love Wikipedia but...
... but fuck that shit, it ain't reliable.
Robusto
Robusto
#travle #656 +0
✅✅✅🟩🟩🟩🟩✅
https://travle.earth
Mitch
Mitch
"Language is a collection of weird gray areas where there's rarely an answer that's 100% right or wrong – so LLMs are typically trained using reinforcement learning with human feedback. " That's a big effing non-sequitur.
RLHF was used to improve "AI alignment" (a quasi BS term) but basically used to help avoid returning answers (factual or not) to things like how to build a bomb.
...which, coincidentally, is not made possible by these LLMS, just slightly easier. If the info isn't already there in the training text, a statistical rehashing of text about it isn't likely to come up with a method.
"OpenAI has built in some 'thinking time' before it starts to answer a prompt. During this thinking time, o1 generates a 'chain of thought' in which it considers and reasons its way through a problem." No, that's wrong (or rather I get waht he's trying to say but he's not saying it well or using words the way they are supposed to be meant.
Robusto
Robusto
Wordle 1,199 3/6

⬛⬛🟩⬛⬛
⬛🟨⬛⬛🟩
🟩🟩🟩🟩🟩
Mitch
Mitch
'thinking time' is just AI baby talk. Tthe engineers thought that the troglodytes would gnash their teeth if they said "passes off to classic logic and planning algorithms, which sometimes take longer than expected". It badly anthropomorphizes these algorithms.
Also, reinforcement learning (in RLHF) was already used in ChatGPT 3. This new stuff isn't RLHF.
"o1... really 'cares' whether it gets things right or wrong": OK just stop it.
"And in that sense, it's the first LLM that's really starting to create that strange, but super-effective AlphaGo-style 'understanding' of problem spaces" - Oh OK, this must be an actual -opinion- article, rather than an explanatory article written in a breezy style. Anyway, AlphaGo doesn't understand anything or 'understand'. Just stop it. And first time? No, ChatGPT 3 already fooled people into thinking it 'undertands'.
"But o1 is still primarily trained on human language. That's very different from truth – language is a crude and low-res representation of reality. Put it this way: you can describe a biscuit to me all day long, but I won't have tasted it." Excellent point. I don't know if I'd have used 'biscuit' but hey I get it he likes biscuits.
See? Not all bad.
"Unshackled from our language and thinking, they won't even notice when they break through the boundaries of our knowledge and discover truths about the universe and new technologies that humans wouldn't stumble across in a billion years." OK OK I give up. You're having fun writing this. 'Unshackled'? Please. I feel like his eighth grade English composition teacher didn't slap him enough.
"Apologies for these rather esoteric and speculative thoughts, but as I keep finding myself saying, what a time to be alive!" Wait. You're kidding me. You were just late-night bull-sessioning this? You're high and you didn't offer me any? Do they have editors at your magaz... Oh. You're the editor. I get it. Good for you. That's nice that you can just say stuff. Isn't that what blogs are for?
He says a whole bunch of stuff about robots and embodied intelligence which sort of make sense but really I have to go rinse out my eyes with bleach because of the 'unshackled' sentence.
So, in sum, to wrap it up, and to summarize, ChatGPT-o1 is not AGI, is not going to take over the world, is not an ominous split in intelligence that will exponentially metastasize to engulf the cosmos.
ChatGPT-o1 should improve sequential logical problem solving that ChatGPT 4.0 had problems with (underspecified high level undergrad math problems), which by the way is -awesome-.
Robusto
Robusto
14:06
#WhenTaken #216 (30.09.2024)

I scored 853/1000 🎉

1️⃣ 📍 93 km - 🗓️ 2 yrs - ⚡ 193 / 200
2️⃣ 📍 2184 km - 🗓️ 1 yrs - ⚡ 147 / 200
3️⃣ 📍 2 km - 🗓️ 0 yrs - ⚡ 200 / 200
4️⃣ 📍 2156 km - 🗓️ 7 yrs - ⚡ 139 / 200
5️⃣ 📍 723 km - 🗓️ 4 yrs - ⚡ 174 / 200

https://whentaken.com
Daily Octordle #980
5️⃣🔟
6️⃣7️⃣
8️⃣4️⃣
9️⃣3️⃣
Score: 52
alphabet
alphabet
14:25
@Vikas Correct; that sentence doesn't make much sense.
I've never been sure what eye level to use. My eye level? The average eye level of visitors to my apartment? That 57" rule seems to have been created on the assumption that only women would be hanging, or looking at, wall art.
CowperKettle
CowperKettle
@Mitch Thank you! It's always interesting to read your comments
Robusto
Robusto
14:40
Daily Sequence Octordle #980
4️⃣9️⃣
🔟🕚
🕛🕐
⓮⓯
Score: 88
Mitch
Mitch
@CowperKettle 🫡
I have other opinions. Like paper or plastic? With plastic it is easier to carry multiple bags at once, and I can use them over and over. But yeah not only is it ugly to see a plastic bag dangling from a tree, it's also disgusting.
So I'm not sure.
Robusto
Robusto
15:06
@Mitch What if those are ghosts? Have you ever considered that?
CowperKettle
CowperKettle
user image
Robusto
Robusto
Tightrope, a daily trivia game | Britannica

Sep. 30, 2024

T I G H T R O P E
✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ ✅ 🎉

My Score: 2290
Mitch
Mitch
@Robusto Ghosts may well be involved but those are plastic bags hanging there.
Robusto
Robusto
@Mitch Take a look at them after dark and you'll see what I'm talking about. I'm just saying, they might be ghosts.
Mitch
Mitch
@Robusto OK I'll look.
But I'm pretty sure they aren't ghosts.
The shoes wrapped around overhead power lines? Are those ghost shoes?
Robusto
Robusto
15:13
They might be. You never know.
Mitch
Mitch
In fact I -will- never know. How do you get those down? You just don't.
I once saw a pair of loafers hanging from a power line. I bet that was a ghost.
Robusto
Robusto
Van Morrison - Jackie Wilson Said (I'm in Heaven When You Smile) (Official Audio)
►
Happy music
Human Intelligence
►
Mitch
Mitch
15:46
@Robusto I've seen that guy before.
@CowperKettle Abridged? Holy crap, how long is the full version?
I had that on my bedside table for years. Just looking at it made me fall asleep.
I liked his first major work, Tractatus Logico-Philosophicus. It is mercifully very short (under a hundred pages?), makes sense (when it is not being ludicrously tautological), written in outline form (great for CS nerds!). Did I mention it is really short?
Supposedly (I've never actually read it even in part) Philosophical Investigations (PI) says the -opposite- of whatever TLP says. I've read many short summaries of it (Wittgenstein for dummies kinds of things) and still don't get it.
jlliagre
jlliagre
16:28
@CowperKettle A landslide victory ;-)
Tightrope, a daily trivia game | Britannica

Sep. 30, 2024

T I G H T R O P E
✅ 💔 ✅ 💔 ✅ ✅ ✅ ✅ 💔 🤕

My Score: 1290
#travle #656 +0
✅✅🟩🟩✅✅✅✅
https://travle.earth
Wordle 1,199 4/6

⬛🟨⬛⬛⬛
⬛⬛⬛⬛🟩
🟨🟨🟨⬛⬛
🟩🟩🟩🟩🟩
Daily Octordle #980
5️⃣🕛
🔟🕚
🟥7️⃣
8️⃣9️⃣
Score: 76
Daily Sequence Octordle #980
5️⃣6️⃣
7️⃣8️⃣
9️⃣🔟
🕛🕐
Score: 70
jlliagre
jlliagre
17:00
#WhenTaken #216 (30.09.2024)

I scored 804/1000 🎉

1️⃣ 📍 458 km - 🗓️ 12 yrs - ⚡ 165 / 200
2️⃣ 📍 2559 km - 🗓️ 1 yrs - ⚡ 143 / 200
3️⃣ 📍 1473 km - 🗓️ 13 yrs - ⚡ 137 / 200
4️⃣ 📍 170 km - 🗓️ 5 yrs - ⚡ 188 / 200
5️⃣ 📍 972 km - 🗓️ 0 yrs - ⚡ 171 / 200

https://whentaken.com
alphabet
alphabet
17:48
Tightrope, a daily trivia game | Britannica

Sep. 30, 2024

T I G H T R O P E
✅ ✅ ✅ ✅ ✅ ✅ ✅ 💔 ✅ 🎉

My Score: 2080
Robusto
Robusto
Has anyone noticed that whenever anyone is said to speak another language it's always described as "fluent in [language]"? And that usually means little more than that they can order food in a restaurant or ask for directions to this or that location.
MetaEd
MetaEd
@tchrist next time in the promised land.
Robusto
Robusto
If Jerusalem is the promised land, I think we should aim higher.
 
2 hours later…
CowperKettle
CowperKettle
19:50
user image
MetaEd
MetaEd
20:19
Wordle 1,200 3/6

⬛⬛🟨⬛⬛
⬛⬛⬛⬛🟨
🟩🟩🟩🟩🟩
@Robusto I'm going to say it's not Texas.
CowperKettle
CowperKettle
20:33
user image
Robusto
Robusto
20:46
Mar 16, 2011 at 3:52, by Robusto
"Just move to Texas." Listen to yourself, man! General Sheridan ("Little Phil"), who was active in the latter half of the 19th century, famously said, "If I owned Hell and Texas, I would live in Hell and rent out Texas."
CowperKettle
CowperKettle
What if Climate Change Was a Hoax? | Frankly 72
►
Philip Sheridan
Philip Henry Sheridan (March 6, 1831 – August 5, 1888) was a career United States Army officer and a Union general in the American Civil War. His career was noted for his rapid rise to major general and his close association with General-in-chief Ulysses S. Grant, who transferred Sheridan from command of an infantry division in the Western Theater to lead the Cavalry Corps of the Army of the Potomac in the East. In 1864, he defeated Confederate forces under General Jubal Early in the Shenandoah Valley and his destruction of the economic infrastructure of the Valley, called "The Burning" by residents...
user image
@Mitch I think that any of his work would be too complex for me to understdand, but I downloaded this audiobook to listen while delivering food, because I heard a lot about this philosopher
Maybe I shoudl find a good audiobook biography on him instead.
CowperKettle
CowperKettle
21:18
The Energy Collapse | Louis Arnoux
►

« first day (5072 days earlier)      last day (144 days later) »