rememberlenny

Very cool. I've followed this project through the HN Startup School. Impressive progress!

Related:

I make paintings from generative art, which I use to inspire physical paintings.

https://medium.com/@rememberlenny/digital-processes-inspirin...

For reference, Jane Street is financial firm known for their widespread use of OCaml.

What happen to your original idea of machine learning GPU rentals?

The graffiti demo he made is very impressive.

Using various photos of the same place, he makes a time scrubber that shows the older art that was on a wall.

Link: https://twitter.com/nathangitter/status/1020733723183124480?...

Hah. I just did this (yesterday) with Kickstarter. Hit 140% of goal in 24 hours.

I made a write up around what I wanted to do. I made a quick video at work summarizing my key points. And I set a $100 goal.

13 people contributed. More than half of them are total strangers.

This does work and can also be encouraging for feedback.

https://www.kickstarter.com/projects/rememberlenny/new-publi...

This is an incredible product.

I have come across the github repo's authors work a number of times now, and I am continually impressed by the quality of documentation, examples, and immediately usable code.

This is pretty neat. Super simple, but potentially expandable. I could see this being really useful before a interview.

It gets 70-80% performance. I’m not sure about throttling. It’s definitely not 1-to-1 based on my desktop, but it’s so much faster than the CPU.

I don’t have one, but based on your request - I’ll write one.

The biggest issue is OS X version limits. I’m on a version two points back due to the NVIDIA hacks needed. There’s a whole community forum on egpu.io that goes into details.

I got a eGPU too for machine learning work. I have a 1080ti which I put in the Akitio Node Pro. Its nice because it has a handle on the top, is portable, and sits on my desk at home.

I also have a desktop machine that I built with another 1080ti in it. I use the desktop machine as a remote machine for longer running jobs, but like the convenience of running files directly on my machine.

If anyone has questions, I'm happy to answer.

Considering his blog hasn't been updated since 2013, can someone add the proper publish date?

Noted that https://news.ycombinator.com/user?id=Mark-Zuckerberg is real

If you're looking for a Smyte replacement, check out Koko. It's AI for content moderation and classification, developed at the MIT MediaLa.

https://www.koko.ai

If you are looking for a dependable, scalable, closed-source option, check out http://www.comet.ml (the thing I work on).

The focus with Comet.ml is more on experiment tracking and hyperparameter optimization rather than model deployment. We make it very easy to compare your experiments results, code, and hashed datasets for better reproducibility.

We have a one-line integration with your existing machine learning code and make it stupid simple to start tracking your experiments.

All you do is:

  > import comet_ml
  > experiment = Experiment(api_key="MY_API_KEY")

_boom_

Comet.ml supports many libraries (keras, tensorflow, scikit-learn, custom-built code spaghetti, and everything else that makes you a ML wizard/unicorn/armored flaming hippopotamus).

  ++ Its free for public projects and academics.

Equally worth surfacing:

  Habalov reported the bug to Google and Mozilla engineers,
  who fixed the issue in Chrome 63 and Firefox 60.

  "The bug was addressed by vectorizing the blend mode 
  computations," Habalov said. Safari's implementation 
  of CSS3 mix-blend-mode was not affected as the blend 
  mode operations were already vectorized.

I went from not knowing anything about Chess Boxing-to-seeing it multiple times today. I came across this: Chess Boxing's role in empowering young women in India.

https://www.newsdeeply.com/womensadvancement/articles/2018/0...

Can someone explain the basic implications of this against current GANs and also provide a practical ML application?

Entire list

  Bee Downtown
  Breaux Capital
  ConvertKit
  Discord
  DotCom Therapy
  Earth Angel
  Eterneva
  Finless Foods
  Front
  HackerOne
  HelloAva
  Iris Automation
  MeritHall
  Mixpanel
  NaturAll Club
  No Cow
  Pilot
  Practice Makes Perfect
  Pull Up a Seat
  Qualia
  Relativity Space
  ReoLab
  Rip Van
  S2 Capital
  Shake Smart
  Solace Technologies
  Swiftly
  Tiesta Tea
  TONL
  Touch of Modern

The key here is governance.

            ------------

I think if you click into the linked post, you get the analogy regarding governments and governance. Web platforms have to concern themselves with governance of their users and the respective eco-system that their users operate on.

The point is less to be abstract or symbolic, and instead look at the concrete decisions they have to make. Corporations or general relationships do not have to consider these points to the same degree for survival.

In the past, governments would create policies, laws, and make decisions that would determine the well being of a citizen. Now, web platforms are more commonly making decisions of regulation, censorship, interoperability, which all fall in the realm of governments.

  "So as you watch the large web services evolve, think 
  about how they are balancing the relationship between 
  the state and the private sector? What does Facebook's 
  introduction of Facebook Credits say about its monetary 
  policy? What is Apples foreign policy? Do they act 
  unilaterally promoting their own proprietary standards 
  or do they act multilaterally embracing international 
  standards? What is Twitter's industrial policy? Do they 
  invest in state owned services or encourage decentralized 
  economic development? The choices these platforms make 
  reveal a lot about who they are, and ultimately how well 
  they serve the companies operating in their economies and 
  the citizens who live there."

Companies or corporations in general may have a degree of these problems, but not in a way that their survival depends on.

Web platforms are governments.

  "Facebook is a government. Facebook's users are citizens,
  and Facebook's applications developers are the private 
  companies that drive much of the economy. Apple. Twitter, 
  Myspace, Craigslist, Foursquare, Tumblr and every other 
  large network of engaged users (including some services of 
  Google) plays a similar role. We have always tacitly 
  acknowledged this. We talk about these networks as 
  communities, communities have governments."

  "Once you start thinking about large web platforms as 
  governments, the logical question is what kind of government 
  are they. One thing is for sure - none of these platforms 
  are democracies. They are oligarchies controlled by 
  founders, investors or shareholders. That may not be at all 
  bad. As long as citizens (users) can move freely from one 
  government to another with little switching cost, there is 
  no reason to burden these polities with the inherent 
  inefficiencies of popular democracy. But that does put a 
  special premium on emigration policies and property rights. 
  Do I own my data, can I export it freely? It also suggests 
  that large networks that have strong network effects may 
  someday need other incentives to act in the best interests 
  of their citizens."

http://www.usv.com/blog/web-services-as-governments

Classic global variable problem.

Hey all! I'm a member of the team that launched this today! We provide a very simple tool for machine learning engineers to track and log their experiments. You install a simple python library using pip, then add one line of code to your program. Similar to Google Analytics or New Relic, but for machine learning. Once you add the code, each epoch of your ML code reports the results to comet.ml, and make it easy to share/compare results.

Its free to use for open source projects.

Check it out and let us know what you think!

^^^ This is why HN is amazing.

SaaS, consulting, niche product, recruitment, report generation, user onboarding optimization, marketing optimization, APIs... off hand

This is a great. Thank you for making this.

The aggregation of models like this isn't new, but simple layout of models available is useful. I can imagine being able to filter by model origin (OpenAI/Facebook/Google/etc), origin date (year), compared use case (why should I use VGG vs RNN).

Overall, great for people who are and aren't deeply immersed in ML.

You are completely right. Thanks for correcting.

You can make well over a few thousand dollars for learning about crypto and working on a interesting side-project.

I agree to this. If you are interested in getting involved with building a crypto project, but can't justify the time: check this out.

Stellar runs a monthly give away of Stellar coins for submissions by developers. They set out a overall vision of projects, and reward a certain designated amount of coins to the finalists.

From what I understand, thats not what its about. AWS and all of the various centralized hosting providers are still valuable.

The idea of web 3.0 and the ethereum/ipfs structure is that you have a fat client and thin server. This means that all the business logic sits on the client and the server is just utility.

Filecoin and the ifps structure allows content like text or images to be stored in a place that the smart contract based client could access.

Some recent examples are:

- Marketplace that uses ifps to store data: https://demo.originprotocol.com

- Game (not sure if this uses ifps) http://cryptokitties.co

[Text from post]

OK, Deep Learning has outlived its usefulness as a buzz-phrase. Deep Learning est mort. Vive Differentiable Programming!

Yeah, Differentiable Programming is little more than a rebranding of the modern collection Deep Learning techniques, the same way Deep Learning was a rebranding of the modern incarnations of neural nets with more than two layers.

But the important point is that people are now building a new kind of software by assembling networks of parameterized functional blocks and by training them from examples using some form of gradient-based optimization.

An increasingly large number of people are defining the network procedurally in a data-dependant way (with loops and conditionals), allowing them to change dynamically as a function of the input data fed to them. It's really very much like a regular progam, except it's parameterized, automatically differentiated, and trainable/optimizable. Dynamic networks have become increasingly popular (particularly for NLP), thanks to deep learning frameworks that can handle them such as PyTorch and Chainer (note: our old deep learning framework Lush could handle a particular kind of dynamic nets called Graph Transformer Networks, back in 1994. It was needed for text recognition).

People are now actively working on compilers for imperative differentiable programming languages. This is a very exciting avenue for the development of learning-based AI.

Important note: this won't be sufficient to take us to "true" AI. Other concepts will be needed for that, such as what I used to call predictive learning and now decided to call Imputative Learning. More on this later....