I think the main question I have about these smaller, "lottery ticket" networks is that they are being trained over and over on the same problem as the bigger network, and are being evaluated on the same dataset as the big network, which leads me to believe that the model will fail to generalize to different but still related problems. Like if the model was trained on Imagenet an the winning ticket was found that had ridiculous accuracy for a relatively small network, I would think it would be heavily overfitted

The article discusses this as its main theme and shows generalization across datasets.

I wonder if a pullback by WeWork would potentially close the "window" for IPOs meaning other tech companies that want to go public but haven't so far will have to wait longer.

Its pretty smart to use neural networks to exploit the non uniformity of the data, but I thought they were going to have replaced just the hash function part of a bloom filter, instead of the entire algorithm. I wonder if it could preserve the speed of the bloom filter while exploiting potentially non uniform data.

There's probably something worth using in this tradeoff space, though this isn't it.

Maybe the network can be used to construct the hash function after it has learned the shape of the data.

This is probably extremely cynical, but does floydhub have a vested interest in promoting genetic algorithms because they can take long times to train, especially with neural networks hyper parameters as the search space. These kind of beefy, complex training processes would be fantastic for its business.

I think this is a reasonable insight into their business model. Especially if they are able to somehow cheaply spin up long-running processes. Not sure how they might be able to do that, though. Conventional tech is all about short term processes.

While I see the point of being financially independent and generating positive-sum wealth, some parts of this argument are bullshit. Most people I have talked to attacking those with lots of money do so because the money gained was made by depriving others of an opportunity to achieve their own wealth. I also disagree with his idea of a journalist reporting for the people being a "status game" and a "nessesscary evil" when he is attacking those who made money illicitly.

Just how some criticize Bezos and Zucherburg for building massive fortunes with either terrible workplace practices or the invasion of privacy of millions of people. not all wealth was gained this way, some was gained through honest hard work, diligence, and prudence, but you also can't blame some for being overly cynical about whether most of those with great fortunes did so be committing great crimes.

This isn't completely on topic but ...

I wonder what happens to a company's culture once the employees see the stock absolutely tanking. The probably aren't fully vested, and so are just watching their payday shrink and shrink.

Even if they are vested, they are almost certainly subject to a 180 day "lockup" period where they can't sell for that long post-IPO.

Github's competitors are already providing the same service for free. Github is dropping their prices from $7 / mo to free to compete.

But they're not. They created a new, feature-reduced repo type to make free/private. The marketing hides this, and you need to dig to the "Comparison" section to see it:

https://github.com/pricing#feature-comparison

Good point, it's not the exact same. But their new offering is much closer to what their competitors are offering, so I'd say it's still a move to try to better compete with the free offerings from other companies.

I don't see this here, but could a student debt bubble result not in a wave of defaults but as a contributing factor in a market slowdown or recession because the average borrower will have less disposable income to spend on other goods and products due to the insane debt payments? A wave of defaults most likely won't hurt any major institution except the federal government, but the lack of consumer spending seems like a potential problem for our consumeristic economy.

I've had similar conclusions when considdering this as well. These loans are not wiped out by bankruptcy so they will be a permanent drain to personal wealth indefinitely. I think we will see lasting effects as more and more young people are unable to invest, save for retirement, and buy real estate but there won't be a sudden shock to the economy like the mortgage bubble.

I think we are seeing this drain already occuring actually. The popular news sources are all reporting that millenials are struggling to buy homes and build savings, though this may be due to low wages rather than debt servicing costs

As much as a Google replacement would be great for competition, it would be extremely difficult. Google was able to scale with the web gradually through anti spam and proprietary software, but a new competitor would struggle to reach such a needed scale so quickly

Tl;dr: High cost of entry for newcomers

People said the same thing about AltaVista.

Agreed but I don’t think this is quite the same case. Back then Alta Vista pretty much only offered search as a value to its users. Google certainly started there but offers a lot more than just search to its users. Only one service had to be overthrown to overcome AV. Lots of services would need to be overcome to overthrow google.

(Note I said “users” not “customers”)

You just need to overthrow the ad business and Google will have a huge problem.

I don't see why this would be smart move for Google/Larry Page at all. If He goes, either he will fail, which would be disaster similar to Facebook for Google, or he will come out clean, which wouldn't change anything. The majority are still going to use google products regardless of public image, so not going best maintains that image.