miraculixx

6 ms·

31.
by miraculixx 2y ago
Attention is all you need.
32.
by miraculixx 2y ago
Ok there is correlation. But is there causation?
33.
by miraculixx 2y ago
Let's note that the label you assign this feature is entirely speculative, i.e. it is your interpretation, not something the model actually "knows".
34.
by miraculixx 2y ago
To plan: to think about and decide what you are going to do or how you are going to do something (Cambridge Dictionary) That implies hire-other reasoning. If the model does not do that, which it doesn't, that's quite simply the wr
35.
by miraculixx 2y ago
Came here to say this. Their paper reels of wishful thinking and labeling things in terms they prefer it would be. They even note in one place their replacement model has a 50% accuracy which is simply a fancy way to say the model's re
36.
by miraculixx 2y ago
Yes GPUs process a (one) computational task on a vast array of data in parallel. But it cannot process two independent tasks concurrently (except, perhaps, by reducing compute power for each task).
37.
Ask HN: Why Don't GPUs Scale?
1 points by miraculixx 2y ago|3 comments
38.
by miraculixx 2y ago
MCP is essentially a request-response convention. We already have one like that, and it is very versatile and well-described: REST. MCP is just hype. Hot air.
39.
Ask HN: Can GPUs do more than just "one task at a time"?
2 points by miraculixx 2y ago|0 comments
40.
by miraculixx 2y ago
If we are to get to AGI why do we need to train on all data? That's silly, and all we get is compression and probabliatic retrieval. Intelligence by definition is not compression, but ability to think and act according to new data, bas
41.
by miraculixx 2y ago
If you squint hard enough he's got a point (just not the way he thinks)
42.
Ask HN: Python package for /healthz API endpoints (Flask, Django)?(twitter.com)
1 points by miraculixx 2y ago|2 comments
43.
by miraculixx 2y ago
bogrod now has a console UI. Using bogrod, reporting on CVEs is a breeze. #devsecops
44.
Show HN: Manage SBOMs like source code(github.com)
1 points by miraculixx 2y ago|1 comments
45.
MLOps simplified. Most teams are small, and don't have engineering capacity(twitter.com)
1 points by miraculixx 2y ago|0 comments
46.
by miraculixx 2y ago
Some people have been told "you are better than anyone else" for their whole life. So that's their base line. Guess what
47.
by miraculixx 2y ago
That's a very weird take. Suerely you would expect some skills from a CS graduate?
48.
by miraculixx 2y ago
It's one thing to analyze it. It's an entirely different thing to let a machine of dubious abilities create new DNA.
49.
by miraculixx 2y ago
Reading their blog post I wonder if an LLMs is really the best way to do this. If I got it right, they used the LLM to enumerate potential protein DNA sequences. Does that really need an LLM? Enumeration is not novel, nor are LLMs particula
50.
by miraculixx 3y ago
Next attack will create a CVE and provide a "solution". Nothing spreads faster.
51.
by miraculixx 3y ago
> Most Redis users have never paid the company behind it even a single cent. Me included. Most users never will. That's the fallacy made by MBA types. They dream up some lofty sums "if only everyone paid us money". What th
52.
by miraculixx 3y ago
Wait, no. You can't run it for anything commercial unless you pay up. That's their whole spiel.
53.
by miraculixx 3y ago
KeyDB, a multithreaded drop-in replacement for Redis, under MIT, owned by Snap. https://docs.keydb.dev/
54.
by miraculixx 3y ago
You means this? "2. Additional Commercial Terms. If, on the Llama 2 version release date, the monthly active users of the products or services made available by or for Licensee, or Licensee’s affiliates, is greater than 700 million mon
55.
by miraculixx 3y ago
Great to see we have alternatives. I think Redis management doesn't understand how people choose software these days.
56.
by miraculixx 3y ago
Microsoft Garnet, licensed under MIT. Uses the same protocol meaning Redis client licenses still work.
57.
by miraculixx 3y ago
Their Q&A essentially says you can no longer build anything that is commercial using Redis, except as a partner. That's exactly the opposite of what the SSPL License says. "this definition would include hosting or embedding Re
58.
by miraculixx 3y ago
Reliability, is what you are missing. LLM are not reliable.
59.
by miraculixx 3y ago
Minio?
60.
by miraculixx 3y ago
Apple more likely. Or some other co we don't know about yet. OpenAI is going to crumble. Microsoft will be challenged for trust. We'll see what happens.

More ›