Skip to navigationSkip to content
Leo Chiang

Leo Chiang

Engineer at Microsoft, Opinions are all my own

I write software while thinking about the world.

  • Mathematically a ridesharing company can address cost by 1. Doing its best for optimizing resources; and/or 2. Carefully and tactically charge customers more. So 1 is pretty much saturated now and 2 will silently start. When 2 is done, it would be the time to cut internal resources, which means they

    Mathematically a ridesharing company can address cost by 1. Doing its best for optimizing resources; and/or 2. Carefully and tactically charge customers more. So 1 is pretty much saturated now and 2 will silently start. When 2 is done, it would be the time to cut internal resources, which means they no longer need many data scientists and engineers for “optimization” as there is little room for it.

  • Happened to have read another report saying realtors are hiring Uber drivers for tips on potential housing trends. Anyways, I actually feel it’s like a traditional real estate giant expanding globally. Where is the coolness of a Silicon Valley high-tech disruptor?

  • My last Samsung phone also had a screen issue after a system upgrade, and I switched to Apple ever since. It may not just be that the new foldable technology is too complicated to hold, but the company’s coherent pitfall. Let’s see how Chinese companies like Xiaomi and manufacturers do with foldable phones.

  • When Amazon wants to do sth, it can become very devoted like this. However I only use Echo to switch on/off my light because I quickly found it (to my surprise) more privacy invasive than Google Home.

  • I don’t see it’s the same story as Twitter in 2016: 1. Snap CEO hasn’t changed at all; 2. Twitter always matters as a public media tool, while Snapchat is still like a niche player. That matters in stock market performance.

  • This is actually b......t. It has no way to be death penalty.

    No matter how bad CRISPR babies are: 1. He didn’t violate a law, so a verdict, or sentence from the court (not community/academia) will be extremely difficult; 2. He is not arrested yet but just under investigation. If he is a suspect in

    This is actually b......t. It has no way to be death penalty.

    No matter how bad CRISPR babies are: 1. He didn’t violate a law, so a verdict, or sentence from the court (not community/academia) will be extremely difficult; 2. He is not arrested yet but just under investigation. If he is a suspect in a sense by law, police would have been involved. However now he’s “prisoned” inside his university guesthouse, so it is still at the level of academic ethics.

    Moreover, China has no motivation in setting a precedence for a scientist, given their investment and state-wide agenda in promoting the country’s scientific research and high tech. To give a death penalty in China, the judge still needs to cite the established law instead of a state research guideline — otherwise it leads to controversy outside the box and the government doesn’t want to deal with it. Even corruption and bribery don’t result in death penalty unless one is a high-level government officer and huge $.

  • This is a typical Amazon way: more aggressive in pursuing bigger goals. However is it a better way for advertisers? Depends on how math works.

    Google and Facebook only deal with online activities for the advertising money flow to make it easier and more error-tolerant: you can hardly get the ground

    This is a typical Amazon way: more aggressive in pursuing bigger goals. However is it a better way for advertisers? Depends on how math works.

    Google and Facebook only deal with online activities for the advertising money flow to make it easier and more error-tolerant: you can hardly get the ground truth because you don’t know who the user really is so many predictions will get wrong. As a business model, pay-per-click has been working for so many years.

    Now Amazon is pushing it to the boundary: pay-per-conversion looks more appealing to advertisers. But they will pay for their samples too. Who owns the risk?

    In the pay-per-click era, if a data scientist has a bad prediction model, the advertiser would just go away. Excuse is easy: we don’t design algorithm for a single advertiser, but for overall better results. Now that the stake is higher as a reality check (against bogus science :-) ), the reliability and generality of algorithms/data is in the spotlight.

  • Despite of Crazy Rich Asians’ flopped box office in Asia, the two Chinese-descent directors, Jon M. Chu and James Wan, happen to have shot their (probably) most popular movie yet (Aquaman by James Wan).

    Both around 40 years old, Chu grew up in Silicon Valley and graduated from USC cinema arts, acutely

    Despite of Crazy Rich Asians’ flopped box office in Asia, the two Chinese-descent directors, Jon M. Chu and James Wan, happen to have shot their (probably) most popular movie yet (Aquaman by James Wan).

    Both around 40 years old, Chu grew up in Silicon Valley and graduated from USC cinema arts, acutely finding that the US market is ready to embrace Crazy Rich Asians at this moment; born in Malaysia, Wan has lived in several countries, and also as one of the screenwriters of Aquaman, he seems to be telling this: a true king may not be the most purebred, but is the one who has a wide range of experiences and thus can find the synergy between worlds.

  • Well, it strangely reminds me of MoviePass, which was once a buzzword but is not doing well. And I don’t agree with the analogy to Amazon Prime because Prime only provides add-on services like 2-day shipping and you still pay full price for your online purchases. The margin of a Uber ride after paying

    Well, it strangely reminds me of MoviePass, which was once a buzzword but is not doing well. And I don’t agree with the analogy to Amazon Prime because Prime only provides add-on services like 2-day shipping and you still pay full price for your online purchases. The margin of a Uber ride after paying out to driver may not be much higher than movie theater operations, when product cost is considered.

    Of course, Uber does it for the future: imagine there will be so so so many people who ditch personal vehicles and rely on Uber. Wait, the Internet scale model doesn’t seem to work here because Uber will pay even more to the hourly-salaried drivers, not counting the additional demand boosted by limitlessness. But the catch is: it will actually scale well when a) driverless cars take over, or b) Uber raises the price or add a lot of restrictions when subscribers really grow fast (assuming people can’t leave Uber because they don’t have their own cars anymore?).

    No matter driverless is coming or not, everything could become a subscription service as long as someone else is willing to pay for the gap. In Uber’s case, it is up to the shareholders when it goes public — “too big to fail”, or seen “swimming naked when the tide goes out”.

  • R.I.P. Who else founded one of the biggest high-tech companies in an area and bought the region’s NFL franchise when its former owner threatened to move the team? Doesn’t sound like the Silicon Valley way though...