Marazan

7 ms·

1.
by Marazan 3d ago
I'm not dismissing the tech! I think the tech is cool and useful! It is sci-fi levels incredible in many ways. But it is also not some mysterious force beyond mortal ken and pretending it is inhibits the useful and safe application o
2.
by Marazan 4d ago
"OpenAI hacked HuggingFace" that's the headline. When you connect to random number generator to the "Do Things" button you are the one who is responsible. IF you don't like that responsibility then don't
3.
by Marazan 4d ago
Corretct.
4.
by Marazan 4d ago
The reward maximising function maximised it's reward. LLMs are cool and all that but the immediate anthropomorphisation of the next-token-predictor technology has stunted the ability of people to reason about them to an _alarming_ degr
5.
by Marazan 6d ago
At this point I am starting to wonder about the RLHF that is going on for programmig. The quirks in fallbacks, defaults and ludicrous gold plating seems to get more and more intrusive with every model upgrade.
6.
by Marazan 6d ago
I love this dance we are doing where when people write the "AI models are garbage machines that produce garbage and are no where close to the fantasy being pedalled by the Crypto bros who pivoted to AI" it always has to be caveate
7.
by Marazan 13d ago
Something load bearing thats for sure.
8.
by Marazan 16d ago
I remember being told in early 2025 that only now has "real programming ability" appeared in LLMs. And since then there's been 3 more "now real programming ability has been made available and previous stuff was just toy
9.
by Marazan 19d ago
No it isn't. Otherwise the entire concept of Original Intent would be dead in the water rather than a cornerstone of how multiple federal judges all the way to the Supreme court rule on cases.
10.
by Marazan 19d ago
The law is not a set of hard coded rules. It is an agreement between humans, tied up in unspoken conventions. The principle one being, don't be a dick. The law always breaks down when it is treated as a hard and fast set of instruct
11.
by Marazan 26d ago
North & South is part of a long tradition of Strategic Map plus Tactical Battle wargames that culminated in the all encompassing Total War series.
12.
by Marazan 1mo ago
Imagine using a random word generator and getting upset that the random words it generates are not truly artisanal random words.
13.
by Marazan 1mo ago
No, it is really useful to know how a technology works. LLMs work by predicting next tokens. It is _amazing_ the utility they have given that that is what they are and they are highly useful but suggesting solutions that ignore they are sp
14.
by Marazan 1mo ago
The trouble is the LLMs non-determinism and the "You're absolutely right, I placed the red key behind the red door rather than in front of it. This was load bearing for the whole dungeon and I blew it" moments. With a hand a
15.
by Marazan 2mo ago
The greatest tragedy is I can only upvote this once.
16.
by Marazan 3mo ago
"Small government" is a euphemism for letting racist people be racist without censure.
17.
by Marazan 3mo ago
That's why you buy inflation linked bonds
18.
by Marazan 3mo ago
> Profitable isn’t related to “worth the dollars”. You need to look at income and how much is being reinvested into growth. Amazon famously remained unprofitable due to reinvestment and waiting for them to become profitable before invest
19.
by Marazan 4mo ago
Someone connected the spicy autocomplete to the "Do Things" button again.
20.
by Marazan 4mo ago
If you connect the spicy automcomplete to the "Doing Things" button then you are responsible for the ethical questions when it presses the button.
21.
by Marazan 4mo ago
It has been crazy seeing the explosion of Polymarket and Kalshi in comparison to Betfair. Betfair doesn't present itself as anything other than what it is: a gambling exchange where the punters make the book. In my 10+ years time of u
22.
by Marazan 4mo ago
PipeDream was a wild piece of software. Even as a young teenager in the early 90s I could tell it was a weird paradigm.
23.
by Marazan 5mo ago
If you remove the auxiliary tools and just leave the core LLM then strawberry still has an undefined number of `r`s in it.
24.
by Marazan 5mo ago
> "incomes are up!". Great, have they increased by as much as inflation? Yes. > Can I afford a home? No. There's an important lesson somewhere here.
25.
by Marazan 5mo ago
> It will be relevant for a long time. Citation needed.
26.
by Marazan 5mo ago
> (so infrastructure for clean water and all chemicals) Fabs are some of the most complex chemical engineering sites (dealing with some of the most dangerous substances) in the world. So don't underestimate the complexity of this
27.
by Marazan 6mo ago
Google was not marginally better. I was orders of magnitude better. It is that simple.
28.
by Marazan 6mo ago
Every year I ask the latest version of Chat GPT a basic facts question about rugby results. It almost always gets it wrong - even when it does web search and cites sources. Wrong scores, hallucinated matches, wrong locations - just gob sm
29.
by Marazan 6mo ago
VisiCalc was the killer app.
30.
by Marazan 6mo ago
"Very simplistic prompt" is the absolute and total core of this and the thing that ensures validity of the whole exercise. If you are trying to measure GENERAL intelligence then it needs to be general.

More ›