CarbonCycles

Intel has an identity problem and greatly over-estimates its capabilities (or influence...or maybe they have jumped the narcissistic shark all together)?!?

TSMC & SAMSUNG have large backing and incentives from their respective governments....no publicly backed corporate entity can match those pockets.

Intel's designs (i.e., architectures), products and manufacturing capabilities are from the early 90s. They have failed to innovate partly because their manufacturing capabilities are pretty poor, which have a high impact on their architectures and product development. They need to decouple those two functions as AMD did when they split into GlobalFoundries and acquired ATI...which was probably the smartest thing AMD ever did.

Also, Intel needs to get off their highly egotistical perspective that chips can only be processed using PhDs and Postdocs...their entire mentality is elitist driven by a highly academic white castle mentality.

I've started to adopt a hedging strategy (i.e. short plays) with IBM marketing...seems like everything they touch dies (horrendously).

ETA..in case anyone is interested, the author of the article is a theoretical physicist, which lends more credibility to the article.

https://en.wikipedia.org/wiki/Sabine_Hossenfelder

I'm excited to see the performance improvements as well...may be one day we will see a version w/out the GIL.

The following notes are interesting: https://lukasz.langa.pl/5d044f91-49c1-4170-aed1-62b6763e6ad0...

That was one of the better letters written by execs....also a generous package.

I feel bad for the folks who have been impacted.

ML applications continue to push the boundaries in the security space...

I have no doubt the new headsets are phenomenal, but it's not in my character to spend that much time with a VR headset. Not to mention, have they solved the motion sickness that came w/ the earlier headsets?

I honestly don't get it; however, I have two solid data points that lead me a very befuddled conclusion.

One of my buddies is an orthodontist who hates his current life but loves video games...he spends most of his free time plugged into the VR headsets playing first player shooter games. Okay..I kind of get that.

My other buddy who has a post graduate degree in genetics who is also an introvert mentioned how he loves Meta's universe. He loves the fact that he can hang out in a virtual theater with his other buddy who lives half way across the CONUS watching 3D movies together. That I don't get.

After typing all that out...I guess I really don't get it. Move along...nothing to see here :D

Not a fan of cloud desktops or any "virtualized" desktop. Experience is typically subpar and the worst part is that it requires a stable internet connection. What's the point of that when many of us are working remote and mobile?

Lots of excellent advice...here is something I like to do.

1) Listen w/out interrupting...look for that natural pause. Also take note that if there is no natural pause, that in itself is a clue to: 1) how the person comprehends information 2) treats other ppl as they may not respect other points of view otherwise they would stop with a pause.

2) Ask if you can repeat what you heard...this forces you to become an active listener as you try to retain salient points of the conversation; lets the other person know that you are listening, which has an interesting psychological effect; and finally validates both what you heard and what they said.

Some tips that might be useful

I like your blog on how you use EDA, but I'm not sure I'm getting the Machine Learning piece. It would be nice if you guys went into more details, but I appreciate how you were able to tie together different data sources and walked ppl through the analysis!

Why do you say that? I think each field has its strengths and weaknesses. For example, mission critical applications still tend to avoid using DL algorithms since stability can only be proven in the sense of Lyapunov. There is also RL, which borrows many fundamental concepts from control theory.

I think they have very very high potential for simple dynamical (i.e. continuous time) systems. I enjoyed the presentation given at NeurIPS (2018) and look forward to seeing them applied to more real-life scenarios.

As someone who comes from a control background, I'm really happy to see this topic get coverage on Hacker News. Also, really glad to see ppl recognizing where control theory really shines (and where machine learning has its limitations)

This should be interesting on how it unfolds...

Seeing the same thing in readers at the gas pump...Austin and surrounding areas have been hit hard w/ them.

That article isn't very clear on what she is advocating.

From the article: This led Anandkumar to challenge AI’s reliance on matrix methods. She deduced that to keep an algorithm observant enough to learn amid such chaos, researchers must design it to grasp the algebra of higher dimensions. So she turned to what had long been an underutilized tool in algebra called the tensor. Tensors are like matrices, but they can extend to any dimension, going beyond a matrix’s two dimensions of rows and columns. As a result, tensors are more general tools, making them less susceptible to “overfitting”

She mentions using higher-order tensors, but if I remember my advanced mathematic classes, tensors use matrix operations extensively.

Is she possibly referring to array programming languages?

https://en.wikipedia.org/wiki/Array_programming

Congrats!

I think it's going to become a cat and mouse game. AI generated images (e.g., deep fakes) are already being used in very nefarious ways such as job interviews, applying for gov't documents via video, etc.

Researchers are finding ways to identify the tale-tell markers that currently give them away, but yes, for the neophyte this is going to be a real issue on what can I trust.

However, the great thing is that you will always have data...the challenge will then become how well do I TRUST my predictions, which I believe will spur some very interesting algorithms such as anomaly detection (i.e., the RGB distributions, spatial-markers, etc are way too distorted if I compare metadata from other pictures of this type).

IMO absolutely not. I just think the entire grant submission/reward process is so broken in Academia that...ah, never-mind. Think I'm just gonna go sit in the corner now....

Academia is such a weird beast...sadly, seeing more and more of this.

I'm surprised we aren't seeing more of this...just another thing to worry about as part of the ML deployment process.

And this is where DS with a deeper more rigorous understanding begin to differentiate themselves by being able to step back and reformulate/re-model the problem into something more like an anomaly detector (in keeping with this example).

I can see the argument where ML Engineers and DS w/out a more advanced statistics/STEM background would fail since they would continue down their list of libraries w/in their prescribed toolbox. Granted many problems can be approximated to be good enough, and let's face it, the FAANG/MAANG/whatever companies aren't running things so super critical where a user getting one extra email or ad presentation will cause them serious injury or death.

Btw, appreciate your comments.

LOL...typically i would agree but i find DL requires a lot more intense engagement (for now) followed by lots of idle time reading hacker news as the models complete training.

DS/ML/BI/WTF EVER has all become overly incestious filled with buzzwords generated by hiring managers/executives who really have no idea what they're talking about when they write the job descriptions (or they're being seriously cheap by blending job functions together such as Head of Data and Analytics...wtf is all that about. From a business function sure sounds great, technically it's a recipe to set ppl up for failure).

I think the only reason why he used Sagemaker is that it provided him an easy out of the box set of tools that required minimum-to-no setup.

If the author took the time, he could have stitched together a completely zero-cost alternative that could have run locally on his box, but that would have gone against the argument that this is so easy that "even a monkey can do this". Don't get me wrong...AWS has some very nice libraries, especially their DL and encoding/decoding options.

I just think it was a bit heavy using Sagemaker especially for such a trivial use-case...maybe even the free Kaggle notebooks would have been a better alternative?

No, I think the arguments you are reading is that this article was poorly written (granted its Ars) and a more appropriate article would have considered broader perspectives.

This feels like something an undergrad would write for their business management class...btw, your chess analogy is flawed in that there are millions of patterns that a computer can simulate for next play. I wouldn't really call it AI per se.

There is a difference between reporting a stat and understanding the stat. Another thing that many ppl (including ppl who work in this field) fail to understand is the theoretical (or statistical) underpinnings of the algorithms they are trying to deploy.

Many assume IID or NID to make the problem tractable, but that's not how the world typically works even on the very very large data scale. More things on why many data science teams/groups/fail because too many ppl treat it as a BI/BA organization...I really should step away from the keyboard now. LOL.

This is a very dangerous article as it tries to make an argument that a person with minimum to no training can start to run these black box systems.

One of the biggest fallacies that I have run into within this space is the failure to understand the problem statement. For example, this article does not mention or address models performances on very unbalanced datasets. Most classic M/S/D-L models will perform mediocrely if care is not given to understanding the statistical distributions underlying the data (or even understand the dynamics of the system such as seasonality).

In addition, the author does not address how biases are introduced by using AWS' (or insert Google, Azure, etc) algorithms...not all algorithms are codified equally.

Finally, this article demonstrates how ppl are trying to trivialize the complexity of very complex algorithms, statistics, etc with a plug-n-play...guess it doesn't help that many companies (I'm looking at you Meta) treat their DS as high caliber business analysts/intelligence units.

Okay, time to step away from the keyboard...

I don't understand the purpose of this article...it's making multiple comparisons but in a way that seems intentionally biased to incite discourse?

It's also hard to ignore that electric motors provide almost near instantaneous torque on demand...talk about a major wooooooo factor that's hard to say no to.

Me and my colleagues (actuaries by training) ran some very interesting what-if scenarios based on our companies insurance packages, and every single scenario indicated that going with the highest deductible (or HSA) was more cost effective for the employee...the employer (with the insurance companies) had stacked the deductibles and cost structure to shift the risk from the company to the employee.

And the medical providers (not the insurance company) will always have discounts on their procedures...just have to ask for them. Makes one wonder how steep the mark-ups are.

No, I also see this in InsurTech. These companies are so deathly afraid of data breaches and system compromises that anything remotely modern frightens them...sadly.

And btw, that description given by your colleague brought cringy mental flashbacks on meetings I've recently attended.