@DavidDeutschOxf
A breakdown for non-physicists in 3 points:
1. The geometry we study in school is called Euclidean. The distance between any two points is the line between them, and the sum of angles of a triangle is 180 degrees. This is only true on a flat surface such as a sheet of paper or a
@marshray
@DavidDeutschOxf
Sure! Imagine you're waving your hand in a jar of honey. The more you move the more resistance honey is opposing to your hand. If you were swimming in a pool of honey, it would take you more time to swim the same distance than in water.
Imagine we thought the universe is
@vietdle
Here's my meatbag attempt at breaking Q* down:
1. A* is a path-finding algorithm. That's what League of Legends (e.g.) uses to move units from point A to point B even though there's a wall between the two. It takes a starting point, a destination, and it searches for a valid
@nikitabier
Don’t forget to visit X regularly to understand the Manager Brain.
In this example, the Manager Brain thinks that all engineers are the same, share the same brain and they all comment on Hacker News.
@VictorTaelin
@DavidDeutschOxf
On behalf of the Terran nation we are sorry to disappoint. But we expect our Singularity soon, so check back in 20 years ... or maybe we'll just come to visit you.
@nearcyan
Seems like
@TheMonitoro
survives to see another day.
We're also working on no-code tools to make datasets from any web data so you can create GPTs or fine-tune LLMs with minimal effort data-wise.
@rrhoover
@Manitcor
Person of Interest, a series. You have to survive until season 3 but then there’s a mind blowing arc about AI that becomes the whole series.
In it a good and a “bad” AI are fighting through their human puppets.
@BHolmesDev
You start to see this kind of stuff when old-timers retire and the new generation is about to rediscover lessons from the past.
The upside is, there are playbooks written for this since long ago. Just wait 3-5 years for someone to rediscover it under a different name.
@Rainmaker1973
Brian Greene is a brilliant science educator. I strongly recommend his courses on Space-time, Einstein's works, Special Relativity and so much more.
@villageglobal
has been instrumental to where we are today at
@TheMonitoro
. More than just investors, Village is a family & community we're proud to belong to.
@adwane
you're a mentor, a model I look up to, and your energy and enthusiasm is a real inspiration.
Go Villagers! 🚀
🎉 Today we announce Village Global III, our new ~$250M seed fund dedicated to backing the most tenacious founders at Day 0.
We're also thrilled that
@reidhoffman
has taken the role of Chairman. 1/5
@wesbos
@syntaxfm
Did you know that Mongo's $out aggregation stage clears the collection you're saving to?
I didn't, not the first time I used it. I cleared 80M records in one swift move, 0 backup.
It wasn't user data though, but a dataset to train AI that took me months to build and to recover.
@fuckpoasting
It seems to be shrinking steadily since peaking a couple of generations ago.
Except this time it's accelerating faster than the shrinking of north pole glaciers.
@DavidDeutschOxf
A breakdown for non-physicists in 3 points:
1. The geometry we study in school is called Euclidean. The distance between any two points is the line between them, and the sum of angles of a triangle is 180 degrees. This is only true on a flat surface such as a sheet of paper or a
@ClementDelangue
We need more of this.
The ongoing safety psyops are getting incredibly well entrenched in the public mind, we need more voices decrying this hypocrisy.
"Oh there are killer robots around the corner. But not mine, mine are totally-safe-not-killer robots."
Hey techies! 👋
I'm amping up my visibility game for 2024 and looking for a buddy to trade tips, cheer on, keep each other on track and take over the tech scene next year.
Who's in?
Working on AI-powered data extraction for
@TheMonitoro
👀
Yeah. You don't need code to setup monitoring around the clock and infuse it with GPT-4 or Llama smarts.
And it only runs the AI model if the underlying data (not the whole webpage) has changed, keeping costs low.
Llama2 pruning done right:
> ~50% reduction in the number of parameters.
> Up to ~50% faster training vs. bnb 8-bit quantization.
> Up to ~1.25x inference speed-up.
by Hicham Badri from
@Mobius_Labs
We've just released our latest blog post on using low-rank structures to prune Llama2, showcasing Hicham Badri's excellent work (who has opted out of X). Dive in here:
After endless chatters with clients, mentees & some who claim to know better, I've reached a verdict: Data, my dear, you're misunderstood.
So, I've cooked up a light read - a gentle first waltz to ease you in.
#Data
#AI
#Scraping
#Dataset
#WebData
@typedfemale
This is so cool, now if only there was a similar function for generating sliding window batches (overlapping batches, shifting one item at a time)
@francoisfleuret
@pmarca
Let's do that!
- A DGX GH200 is purported to offer 0.5 exaflop at FP16.
- An RTX 3060 is 50 tflops FP16. It's the most popular gaming GPU on last month's Steam survey[0] and arguably the oldest that you can train modern models on without fiddling.
That would place the DGX at a
Taking LLMs from proof-of-concept to production - 4 takeaways from Ting Wang, Director of ML & AI at
@Wayfair
, and
@MarcKlingen
, Co-Founder of
@Langfuse
. [1/5]
I am observing the same thing, incomplete answers or placeholders in the code.
This happens regularly despite explicitly asking it in the system/request prompts to answer exhaustively.
Anybody else noticed something similar or was it always like this?
@bindureddy
I like to follow two parallel development cycles:
- Short iteration (every 2-4 weeks), mostly hacks to sustain customer request inflow.
- Long iteration (2-4x a year) with a proper refactoring based on the cycle's learnings.
Why not fix things "properly" from the start?
As we
I’ve given a short interview about
@SawalniBot
, the first Moroccan-speaking AI, and its critical contribution to the cultural sovereignty of Morocco and its social and economic development. Read more in this article.
Among the dangers of AI is that LLMs dual-trained on code and biology could enable computer viruses to jump to DNA substrate. Imagine getting a cold that compromises your immune system and makes it start mining Bitcoin
And deployed! My blog now offers a feed in all 3 formats (RSS, Atom, JSON)
Thank you for reading! 🙏
Add to your reader and stay tuned for a future write up, including the code to setup your own feed on a
#Sveltekit
website.
#buildinpublic