algoth1

5 ms·

61.
by algoth1 3mo ago
Google doesnt care about prompt injection attacks??? This is insane
62.
by algoth1 3mo ago
As someone who has worked closely to the marketing space, there’s a saying that goes something like: ‘then the marketers found about it and ruin everything’. Quick example: when amazon launched kindle self publishing, there was a golden age
63.
by algoth1 3mo ago
It even refuses to do numerical time-series analysis (this on an empty project). This is something even a non-llm ML algorithm can do. It’s insane
64.
by algoth1 3mo ago
When my mother was fighting cancer, I recall the many disappointments of finding research shrinking tumours in animal models, only to find out the human research showing it didnt work in humans. This was in the 2010s, before llms, but when
65.
by algoth1 3mo ago
They didnt include chatgpt in the comparison chart. That tells a lot
66.
by algoth1 3mo ago
But wasnt claude code leaked? Why wasnt this found earlier?
67.
by algoth1 3mo ago
I wonder if this bug was found via project glasswing
68.
by algoth1 3mo ago
I love how the doctors injected basically water. I imagine the doctor thinking "we did all we could"
69.
by algoth1 3mo ago
I don't work in coding, but I do a lot of complex tasks that can be automated to some extent. In 2019 I spent more than a month painstakingly building an autohotkey script that would interact with a design app to build a Chinese langua
70.
by algoth1 3mo ago
That's a real possibility for a time, but eventually people will look back at fable 5 the same way we look back at gpt2
71.
by algoth1 3mo ago
Oh, i see. That’s entirely correct. I think the pelican test is more of a meme at this point, similar to Ethan’s Otter on an airplane for video models
72.
by algoth1 3mo ago
We at HN: https://xkcd.com/2501/ to basically say that I think you might be considering low-effort what’s actually an attempt at simplifying - which is arguably higher effort
73.
by algoth1 3mo ago
There’s an hypothesis that states we dream so we don’t lose visual processing neural connections. Similar to what happens in blind people: visual processing neurons are recruited to other sensory tasks due to lack of stimulation. My ed. gue
74.
by algoth1 4mo ago
When a xiaomi 15 phone asks to scan your face for phone unlock, it specifically warns that a photo of you may be sufficient to unlock the phone. Not sure how it works with other brands/models, but i don’t think the way they are doing f
75.
Enabling Claude Code to run code when its sandbox fails (windows only)
2 points by algoth1 4mo ago|0 comments
76.
by algoth1 4mo ago
The thing is, llm token frequency was derived from human writings like yours, and rlhf for good writing practices, like the emdash. So getting ’detected’ on good writing it’s unfortunately to be expected. My broken ESL english is much safer
77.
by algoth1 4mo ago
My experience has been the same: flatout refusals no matter how i frame the health questions - very frustrating. Even psychology is out of scope. Pretty useless unfortunately
78.
by algoth1 4mo ago
I guess you could even turn everything into numbers, not a bad idea at all!
79.
by algoth1 4mo ago
Mythical levels of refusal... checks out!
80.
by algoth1 4mo ago
Isnt this the principle behind synthid?
81.
by algoth1 4mo ago
It’s unusable for me due to the refusals. I’m using claude to find patterns in health data
82.
by algoth1 4mo ago
The refusal rate is insane
83.
by algoth1 4mo ago
Getting the 2002 Cool Edit Pro 2.0 vibes!
84.
by algoth1 5mo ago
It probably won’t work either way: https://arxiv.org/abs/2605.13829
85.
by algoth1 5mo ago
I’m a native Portuguese speaker, whom grow up watching morning Brazilian Portuguese cartoons. I would say it’s doable
86.
by algoth1 5mo ago
It was a conversation crossing raw IVG data against a symptom phenotype (dna analysis) which went on for a good while
87.
by algoth1 5mo ago
It has happened to me with very long conversations
88.
by algoth1 5mo ago
Wouldnt it be easier to fine tune a model to convert the Brazilian Portuguese corpus into European Portuguese and then use that corpus?
89.
by algoth1 5mo ago
With a 10min delay, via aistudio
90.
by algoth1 5mo ago
That’s basically the original gpt5 routing idea but done right

More ›