az226

10 ms·

121.
by az226 2y ago
At big tech company I used to work for ($3T) they did this in 2017, and my manager did not give a single offer to a man in years even when everyone said hire, and the next 14 of 14 offers were to women, several minorities, despite many havi
122.
by az226 2y ago
The honest message wound have been: Hi all, I wanted to share some changes we're making to our hiring, development and procurement practices. Before getting into the details, there is some important background to lay out: The legal and
123.
by az226 2y ago
I didn’t find it. I dug up the real/raw numbers and did the math.
124.
by az226 2y ago
We don’t need this. We can easily unpack Nvidia’s marketing bullshit. 5090 is 26% higher flops than 4090, at 28% higher power draw, and 25% higher price.
125.
by az226 2y ago
The 5090 TOPS number is with sparsity at 4bits, so it doubles the value compared to the 8bit sparse number for 4090. The real jump is 26%, at 28% higher power draw and 25% higher price. A dud indeed.
126.
by az226 2y ago
Flops went up 26% and power draw 28%. So the biggest benefit is PCIe 5 and the faster/more memory (credit going to Micron). This is one of the worst generational upgrades. They’re doing it to keep profits in the data center business.
127.
by az226 2y ago
For sure.
128.
by az226 2y ago
Yes, but while many on Altman’s circuit pledged to donate to OpenAI, they never actually did. Musk was like the only main one.
129.
by az226 2y ago
It’s also not in the non-profit’s interests to convert the shares to ordinary shares at trust me bro conversion rates.
130.
by az226 2y ago
That’s what they want to convert into ordinary shares into the for profit company. If you cash them out at the last round price, I bet they wouldn’t be happy. They want to convert at a trust me bro conversion rate. “Independent financial ad
131.
by az226 2y ago
They just raised $6B. That’s like at the tippy top of a VC round in human history.
132.
by az226 2y ago
Exactly. The bespokeness they mention is opposite. When investor demand is crazy high and you can raise the largest VC rounds in human history, the argument that they can’t raise money unless they change to for profit is laughable at best.
133.
by az226 2y ago
The argument was because they’ve converted the nonprofit to a for profit, which enriches the employees and investors and doesn’t serve the nonprofit or its mission, the nonprofit was only so in disguise and should be viewed as having been a
134.
by az226 2y ago
“Our plan is to transform our existing for-profit into a Delaware Public Benefit Corporation (PBC) with ordinary shares of stock…The non-profit’s significant interest in the existing for-profit would take the form of shares in the PBC at a
135.
by az226 2y ago
Only a matter of time. The costs are aggressively going down. And with specialized inference hardware it will go further down. Cost of coordination is also large. Immediate answers are an advantage/selling point.
136.
by az226 2y ago
The problem is data. GPT-3 was trained on 4:1 ratio of data to parameters. And for GPT-4 the ratio was 10:1. So to scale this out, GPT-5 should be 25:1. The parameter count jumped from 175B to 1.3T, which means GPT-5 should be 10T parameter
137.
by az226 2y ago
I thought that was a fake.
138.
by az226 2y ago
GitHub is scared. It’s losing mind share with startups who are leapfrogging them, Cursor, Poolside, Augment, Cognition, Codeium.
139.
by az226 2y ago
Mousetrap.
140.
by az226 2y ago
https://x.com/natfriedman/status/1823143111205490937
141.
by az226 2y ago
Show the code
142.
by az226 2y ago
Or a workflow that pauses charging a customer when a cancellation meeting has been scheduled. And lets the customer know it won’t be charged.
143.
by az226 2y ago
Yes, but it was trained on 2018 data and earlier.
144.
by az226 2y ago
What’s the context window?
145.
by az226 2y ago
This is wrong.
146.
by az226 2y ago
Inference speeds go brrrr as well.
147.
by az226 2y ago
How common is it that people come to SF and work illegally on their startup and then raise money?
148.
by az226 2y ago
Just gonna leave this here: https://github.com/kyegomez/tree-of-thoughts/issues/78#issue... Also this part from the reply before editing it away: They get mad that my repo and code is better than their's
149.
by az226 2y ago
Based.
150.
by az226 2y ago
Can I hire you to make Flash Attention a reality for V100?

More ›