KidComputer

They literally lack basic relational understanding so

> knowledge across the whole scene

isn't really correct.

https://arxiv.org/pdf/2208.00005.pdf

Search for yoga on TikTok and Instagram and compare the results. Guess which one returns almost all softcore pornography vs actually educational content. It's not a surprise which one teenagers would find more engaging.

Doubt it, you seem kind of socially inept.

Source? Ben & Jerry's had 9% sales growth last year, out pacing the rest of the market at 4.5%. They haven't reported this year so I'm wondering where you're reading that, I can't find anything.

Deep. Bet you think New England belongs to the British.

I found Kubeflow Pipelines to be worse than vanilla Argo. Not only is the documentation poor for the python DSL that gets compiled to the Argo workflow spec but some operations are downright impossible to express in the DSL. Vanilla Argo + RBAC works wonderfully especially with Argo Server since v3+.

From a comment:

> This repo is attempting to apply Benford's Law to vote count distribution, so that's what actually needs to span multiple orders of magnitude. Precinct distribution is a factor in vote count distribution but it doesn't tell the whole story. I don't see the Milwaukee data in this repo, but take a look at the Chicago data: https://github.com/cjph8914/2020_benfords/blob/main/data/chi... Biden's vote totals are solidly contained within one order of magnitude, the 100-999 range. Trump's vote totals range from single digits into the hundreds, across three orders of magnitude. Jo Jorgensen is mostly in the 0-20 range, across two orders of magnitude.

I'd generate document embeddings for each page then put the embedding vectors in a similarity search database like Facebook Research's Faiss.

Wow it's almost like ice breaks off and drifts away increasing extent as it is melting. A smarter person might look for the total ice mass over time and would find a massive fucking decline. But yeah.

https://www.pnas.org/content/116/4/1095

Inductive biases are not a new topic in ML. Also your examples have little to do with inductive biases as inductive biases are about generalization on out-of-sample data through human based restriction of the model hypothesis space.

This is total horseshit. The supporting evidence for this argument is a bunch recent public software failures that the author says 'must have been due to complexity' but then offers no proof. Sure it could have been due to complexity but it could have due to many other things such as mismanagement.

Then at the end the author says:

> Docker and Electron are the most hyped new technologies of the last five years. Both are not about improving things, figuring out complexity or reducing it.

That's not why these are 'hyped' technologies. These things solve real problems. Electron solves the problem of a cross platform GUI with permissive licensing where there are an abundance of cheaper frontend devs vs QT experts.

Sure, Johnathan Blow said we could just copy programs between computers in 1960s but that's because every computer in the lab was the same back then. But that isn't reality today. Today, if you want to ship some server applications to different business customers you'll find yourself having to maintain N different install scripts for the N different distros your customers have. Or you could simplify and use docker and have a consistent environment for your application.

> They do not need to know, exactly, how X is built, why it was built that way, or how to write an alternative X from scratch.

Why on earth would they? I want them to build Y not build me an X from scratch. But I assume they are competent engineers that can solve novel problems and they would research and build me an X if tasked to it. After all most software engineers got their engineering knowledge from trying to solve problems they were tasked with in their career, not in a classroom.

There is no intelligent discourse in this piece. The author could have looked at why these technologies were created and adopted, maybe have made the argument that their adoption is unwarranted due to X. Supported X with good examples and proof.

Those benchmarks are ridiculously out of date. See https://wrosinski.github.io/deep-learning-speed-vol1/ for updated benchmarks by the same person. PyTorch beats Tensorflow on every single model.

Still waiting for it to support basic, but crucial functionality like tolerations and node affinities. And no, using admission plugins like PodNodeSelector is not viable on managed clusters like GKE.

That's because octave is using doubles. You can do the exact same thing in PyTorch by passing in dtype=torch.float64 into torch.randn.

If I get the horse galloping in a circle its legs end up crossing and clipping through each other.

> It's looking at real life

Collecting observations aka data.

> deriving the mathematical laws that govern what you see

Fitting a model.

> Your neural network doesn't tell you what features make songs distinct

It literally learns better features that you could ever come up with by hand. This is why CNNs do better in computer vision that hand engineered filters.

> I guarantee you they'd do a better job, and their models would work on a commadore64, with real time training.

LOL if you think that a room full of people can listen to TBs of audio data, decide what mathematical functions when combined together are better descriptors of that data than a DL model learning its features.

You don't have the slightest clue what you're talking about.

Yeah neither of those metrics are by which we measure the effects of carcinogens and endocrine disruptors.

That's not even an argument. So people should be allowed to watch child porn?

Plenty of parallels between the two. People are hurt in the filming of both. It's damaging to the survivors. It feeds into the fetishes of those who would watch such a thing (it's a snuff film). In the case of this film I could argue that potential perpetrators could study the film for their own plans.

People lost their lives terribly, that much can be understood without watching it. What more do you really gain by watching the video? Do you have some morbid curiosity that you need to satisfy? Do you think we should disregard the feelings of the survivors or those who lost someone close? How do you think they feel about the video being shared? How would you? Do you think they deserve respect and solidarity? If so, how does sharing the video contribute to that?

You'll find plate models, PGM junk, etc in modern papers on explicit density generative models and factorizing latents on such models.

It's not solved.

After a review of the course materials I'm not sure what makes this course 'Advanced'.

Thanks for the ancedata, my personal ancedata for many years of living in NYC is:

Cabs:

* Cars regulated by state.

* Usually cheaper than Uber rides (in my experience always).

* Know the city like the back of their hand.

* Can file a complaint directly with the state or cab company.

* Accountable through a taxi license.

* Won't defraud you with bogus cleaning fees.

Uber:

* Been in two minor car accidents due to driver switching between Waze and GMaps to find the route or just constantly fiddling with their phone.

* Had an Uber driver show up in a car that looked nearly totaled. Passenger side door was nonfunctional due to damage.

* Poorly maintained cars operated around the clock by operators than get virtually no sleep due to an unlivable wage.

* Had a driver travel at high speeds in poor conditions while talking about how they only get a couple of hours a sleep a night and manipulating their phone.

* Many drivers clearly living in their car, undergarments and clothes on front passenger seat.

* Uncomfortably hit on my girlfriend 100% of the time if she is alone.

* Driver complaints often ignored / go into the corporate void.

This is why we don't trust ancedata.

Small business owners are definitely not drooling at that prospect. They usually have very slim margins and their rent will sky rocket as the landlords get greedy. Only a select few will remain usually due to being hip local places that already have high foot traffic that can successful raise the prices of their 'famous' cheesesteaks or opening the property. But for the average small business owner running the typical depressed looking establishment they'll see their repeat customers disappear one by one. They won't have the necessary capital to reinvigorate their image or pay their tremendously higher rent when the landlord decides they want to attract Starbucks to that location. Look at Bleeker street, the lower east side, Chinatown, the east village all in NYC. All places I've seen personally change rapidly over the years. I've talked to many owners, the story is always the same, we're being forced out of our lease. Being in tech I'm certainly part of the problem. And to be honest most of the new places that come in, particularly restaurants, are often better than what was there before. But let's not say the existing small business owners are drooling, they're almost certainly filled with dread at the prospect.

Miners are making a killing off of the insanely high Bitcoin transaction fees.

> What, besides ethics and money, stops you from running an evolutionary process that selects for intelligence at a vastly higher rate than real time

You can already do that today. So why don't we have a superintelligence? I can train a NTM to learn counting and give me the next natural number in a single step. I can't train it to give me the next prime in a single step, why?. Computability and complexity, both conveniently forgotten in the discussion of superintelligence.

Seems like building a GTA style simulator in UE4 or Unity would be a better solution in the long run rather than hacking GPU resources like an aimbot developer.