News for hackers. Stuff that matters.

anurag6892

everyone in the org can read them. One can also subscribe to other org's mailing lists.
In practice, each org has a top5 mailing list to which employees send weekly updates. Jensen samples from these mailing lists.
Re:Running a 180B parameter LLM on a single Apple M2 Ultra on Thursday September 07, 2023 @01:46PM (#37422761)
this advantage is not specific to OpenAI right? Any big cloud provider like Amazon/Google can host these open LLM models.
Re:What OpenAI Really Wants on Wednesday September 06, 2023 @02:36PM (#37409196)
higher comp at Netflix
8% range increase expected in EPA cycle
Re:Nvidia on the Mountaintop on Monday August 28, 2023 @06:23PM (#37301136)
I am baffled when people say Nvidia got lucky with the recent AI wave.

Alexnet, one of the first papers in 2012 to kickstart the current AI revolution was using Nvidia GPUs. Nvidia has been prioritizing AI research (both hardware and software) for the last decade.

definitely. this is often the cause of overinflated CEO pay.
why should he be better than replacement level CEO? why can't we assume 225 million is what an average CEO running a Google sized company deserves? Like how we accept 300-500K is what an average Google SWE deserves?
when dealing with 1%, what is the need for accurate measurements? they calibrate his pay to other CEOs in tech and other industries, and that is good enough.
I struggle with this too. My usual exercise routine is to run alternate days. With WFH, I go on 1 hour evening walks on days I am not running. This makes up for the lack of walking at home.
EPA should also add multiple range estimates say at 70mph, 80mph. Lots of people drive 80mph+ on the highway.

And buyers should be made aware they will be using only 70% (10 to 80) of the battery between highway charger stops. Charging gets really slow after 80%.

Sure, but isn't it up to EPA to fix their testing methodology to get better estimates? Why do we expect car makers to do that?
Re:CPU Utilization Is Wrong on Tuesday July 04, 2023 @02:55AM (#36583132)
yes, most instructions are multicycle. but multiple instructions can execute in parallel in the CPU pipeline.

Moreover, while latency is multiple cycles, most ALU units in CPU can start a new operation every cycle i.e. throughput is 1 operation per cycle.

agree with the last part. The average YouTube user seems to be clicking through ads more often than we think. This is similar to credit cards, where banks get most of their interest revenue from few users.
51% of the $12 premium is distributed to creators of videos one watches.