strin

Many U.S. businesses never survived local competition in China. Kudos for Apple to stay in business and remain a significant share of the Chinese cell phone market.

“One of my philosophies of building companies is the importance of velocity,”

True. And the key to keep (increasing) velocity is momentum. I believe the fundamentals to that is infrastructure. Once you get to the "flywheel stage" of an AI company (http://nicodjimenez.github.io/2018/01/25/stages.html), experimentation becomes so easy that new models can be rolled out nightly.

Democratization should allow users to "own" their models. This is not the case in Cloud AutoML. Users cannot download their models and host them elsewhere. This dependency means Google can have control over the business's AI capabilities.

Levels of driving automation is an ill-defined concept.

For example, L4 is defined as "mind-off". According to Wikipedia, "the driver may safely go to sleep or leave the driver's seat. Self driving is supported only in limited areas (geofenced) or under special circumstances, like traffic jams. Outside of these areas or circumstances, the vehicle must be able to safely abort the trip, i.e. park the car, if the driver does not retake control."

However, the difficulty of driving varies so much from case to case. City is significantly more complex than urban areas. If some company geofence the cars in urban areas and achieve level 4, it might be less impressive say a Level-3 system that works in complex city scenarios and extreme weather conditions.

> there's a lot of other factors to work out like scheduling, routing, supply-demand management, optimal vehicle distribution, pooling etc

These things come along the way. If there is a service that's 10x cheaper than Lyft, I could imagine most people would churn to that service. The technology shift from human to completely autonomous driving is so huge that it could just disrupt the current market structure.

Another competitive advantage of China is data. China has 1.4B population, and nearly everyone in the country owns a mobile phone. This will contribute data 3-4 times larger than US. Considering second-order interactions (network effect), the data could be an order of magnitude larger.

Chinese citizen also seems less concerned about their privacy. I've talked to one of the largest chat platforms in China, and they seems very open to share their conversation datasets with third-parties for dialogue research.

In high-value verticals like healthcare, there is also less regulation compared to U.S. For example, it's easy to start a AI + radiology company in China and have access to hospital patient data, without going through tedious approval process.

Their high-speed rail train is not maglev. There is one built by Germany in Shanghai, but it was more of a demo than an actual infrastructure.

Great work.

I've also been working on a similar open-source project "Sharead".

https://github.com/strin/sharead

It has a chrome extension that uploads Arxiv papers, and you can manage papers with tags.

It also automatically converts pdf to HTML using a library called pdf2html:

https://github.com/coolwanglu/pdf2htmlEX

I own a chevy volt and it never felt as good as it should be. Initially it has a burning smell on gas mode. I took it to the dealership for maintenance and they said GM's official solution is to install a "carbon filter" to the venting. It has an integration with Apple CarPlay that works most of the time but has some small glitches.

It all felt like... Nokia - you know the times when you buy those smartphones expecting them to be really intelligent but they turn out to be awefully hard to use... until iPhone came along.

Most industries tend to consolidate over time with a few big players. For technology it's particularly true given the winner takes all phenomenon. Unless there is a fundamental new technology / hardware platform, there's little chance small startups can go into existing ones.

However, I am more interested in seeing technology disrupting old heavy industries. Please go outside the silicon valley and there are plenty of those.

Take my money!

The home IoT world is broken - I got lots of devices (Sonos, Hue, Alexa, Roku, ...) and they don't play well with each other.

Shoutout for data and API integration layer for Home IoT.

Marketing is crucial when you have the right product. If a startup spends tons on marketing and there are product deficiencies, then users are going to come and go.

I wonder how crowdsourced paper reading can change peer review. For sure, we need experts to tell the quality of a paper. But most interesting papers coming out of Arxiv got read by experts anyway.

The live demo of a neural network to tell romantic stories about images, thanks to the open-source project by github.com/ryankiros/neural-storyteller. Drag and drop images to input and see what happens! Share interesting stories with your friends.

Live demo! http://beta.moxel.ai/models/strin/colorization/latest

We need more demos of AI models: there is what people claim their model does, and there is what the model actually does.

A random idea: in this article, they talked about a Feature Store shared across teams. Feature extraction also seems a major bottleneck in their modeling. It seems a very natural and valuable thing to have a platform to share "generic features".

Yann LeCun said once a deep neural network predicts many targets, say 1000 classes in ImageNet, it is possible for the model to learn quite generic features. So it makes sense to pre-train on a large amount of data and a reasonable number of targets, and then share the learned feature extractors with others.

Could this be a business? Or a community? Thoughts?

Maybe they trained on global data and it biases against sample portion of the dataset that represents Singapore?

How could this be 10X better than products in the already crowded cell phone market?

Interesting. That means your data gets extra Great Firewall protection :)

Google's PageRank code is a trade secret.

(Chinese citizen here)

I do know a decent number of top Chinese VCs looking into investments in Silicon Valley. Those funds are typically started by Chinese entrepreneurs who made a fortune in things like the internet wave or real estates.

There do exist branches of Chinese government, or nation-owned organization making investments in US. But they are not the top VCs, and I haven't seen any of them taking a stake in a top US AI startup.

There are a couple of reasons why Chinese money is flooding into silicon valley. First, there is too much cash in the domestic market and not enough good tech investments to make. Second, there is a fear that Chinese economy might crash someday so having oversea investment diversifies the risk.

Awesome project.

Despite the flaws, the nice thing with VNC is its universality to support any apps on a computer. Using HTML5 in a browser limits the scope of things we could encapsulate as environments, and makes it less "universe".

However, there is a difference between the universality of the tech stack and the exposed interface. In my opinion, the future universe would be rich clusters of RL environments with unified API, each of which implemented using different underlying technology to meet the desired synchronicity and frame performance.

HTML5 could deliver one of such clusters.

PDF here: http://online.liebertpub.com/doi/pdf/10.1089/space.2017.2900...

Interesting Venn diagram in Fig 2. If "COST OF TRIP TO MARS" = "INFINITE MONEY", shouldn't "CAN AFFORD TO GO" be an empty set?

My theory is based on the Internet.

First, people are now exposed to vastly more information that in 80s or 90s. This forges the diversity in taste and segregation of pop culture.

Second, trends come and go at faster cycle than before. If something on Youtube goes viral, this could be in pop culture for weeks and then vanished in the shadow of another fad.

At first, I thought this is a VNC connecting to a Win93 in a virtual machine.

Then I realized everything is written with web technology.

I highly recommend the book "Master Switch" by Tim Wu on this topic. (https://www.amazon.com/Master-Switch-Rise-Information-Empire...).

It argues that every information networks in the history- telegraph, telephone, radio, cable - follow the pattern of consolidation and disintegration. The new inventions always had the chance to disrupt the old industry, but our modern network - the Internet - might be an exception. Because the Internet is the master switch of all things digitized.

Sure, Google can monetize from these products. But they don't seem to choose to capture those values yet, as long as "Don't be evil" credendum still holds.

> ... the company hired Russ Salakhutdinov, a Carnegie Mellon professor whose expertise is in an area of artificial intelligence known as “deep” or “unsupervised” learning, a complex branch of machine learning in which computers are trained to replicate the way the brain’s neurons fire when they recognize objects or speech.

The author of the article clearly made a mistake saying "deep learning" is the same as "unsupervised learning".