maestro64

So AI shamed CR into telling everyone how they are testing.

In all honesty, doing the wide range of testing CR does is extremely hard to be a subject matter expert on so may topics, I would put CR testing in the range or user monkey testing (just randomly doing things to see if something goes wrong) and experts testing (where you have complete understanding of the technology and real work uses and failures). I have lots of product testing experience and one thing I have told people about this kind of testing, any test engineer can design a test which every product passes or fails, however, this taught you nothing about the products performance. The hardest thing to do it replicate real world use cases and results. This take years to understand and the only way to do this, is you have to have real world experience and data from products in the field which CR does not have, they only have anecdotal information.

Case and point, CR said they "train the panelists to discern characteristics of the sound" you can not train people to do this, please refer to the whole Yanny/Laurel debate. 

People's hearing is influence by too many things, what you hear is not always what other people hear, this is why engineer use test equipment to determine how a speaker performs in a specific environment. I had friend who was an audiophile, loved his music and his equipment, he would go through great lengths to explain what he was hearing and why it was important and I could not hear what he was hearing or he could not explain it to me. But hook it up to test equipment and we both could see it in the data. 

Their data is good at telling you whether you may or may not have an issue, and how you end experience may be, but it not going to tell you if you have good or bad product.

For all those who are sighting studies, when you read them think about this.

I will share something I learned a long time ago being an engineering and trying to determine root cause of a problem and finding the single root cause. When doing studies and research into a problem you change one and only one variable at a time. If you change one things the problem goes away, you're not done you have to then go back and undo that change and see if the problem re-occurs. I seen plenty of times by undoing the change does not always bring back the initial problem. Also making one change and seeing the problem going away does not mean it may not re-occur. When either of these events happen, means you do not find the real cause of the problem.

This approach works really well in the non-human sciences like electronic or mechanical systems. In the human sciences you can not always undo the change, also it is extremely hard to control change and other variables. What they do in the human science is to use statistics to try and explain what may or may not be happening. Stats immediate tell you they do not have all the data and they are extrapolating.

Exxon went up because price of oil and gas went up. As soon as pricing goes down they well fall back.

So the FBI found some of kiddy porn on a server which the guy set up many years ago to allow people to share data (file sharing) and can not provide he put the porn on the server. They also could not find any evidences this guy took the files from CIA servers and then sent them to Wikileak.

They also claim to track down the Russian hackers who used series of proxy and VPN servers, and the FBI is 100% sure they found the people who hacked the election as they put it. But they can not seem to find a shred of evidence this was the guy who took their precious hacking information and claim he must have used TOR which is also a series of VPN and proxy servers.

Here is the difference, Russian hacking the FBI does not have to provide beyond a reasonable doubt the Russian were the hackers, they can make claims like this since they will never have to stand up in court and back this up. But in the US for this guy they have to stand up in a court and swear they found the right guy and they have real evidence to back up their claims.

I have funny feeling this guy is going to be railroaded so the CIA and FBI can save face.

steveh said:

Microsoft's "saving" Apple was hardly done out of altruism, nor did it cost Microsoft anything in the long term.

In 1997, Microsoft bought 150,000 shares of Apple preferred stock, convertable to common shares of Apple stock at a price of $8.25, redeemable after a three year period, for $150 million. Apple was worth ~ $3B at the time.

By 2001, they'd converted all of the shares into common stock, netting the company approximately 18.1 million shares. 

By 2003, they'd sold all of it.

It was mostly optics.

Yes it may have been short sighted, however, I suspect that part of the deal was that MS had to exit, the agreement most likely had a clause which required MS to sell its stake in Apple after some time or when a milestone was hit. Apple did not want MS holding a vote share in the company too long.