Forgot your password?

Comment: Re:Like the world needs more web monkeys ... (Score 1) 226

by byteherder (#48407127) Attached to: Coding Bootcamps Presented As "College Alternative"
The CS graduates at my college have required courses in discrete math, data structures, algorithms, compiler theory, operating systems theory, assembler language, graphics, database theory, microprocessor systems with electives in parallel programming, functional languages, machine learning, information theory, complexity theory, computer architecture and artificial intelligence. Are you telling me that all those courses can be crammed into a 19 week boot camp? We don't have courses like JavaScript, web design, html, css, agile and ruby. If you couldn't pick them up on your own, you weren't smart enough to be in the program.

Comment: Re:That's true, but... (Score 4, Insightful) 212

by byteherder (#48357279) Attached to: New Book Argues Automation Is Making Software Developers Less Capable
That is the point, you DO need to know what is going on underneath.

There is functional requirements and non-functional requirements, both are important for the projects to be successful. I was on a team creating a moderately complex system. One of the programmers checked in a perfectly correct functional code but did not meet the performance requirement. The conversation went something like this.

Me: Your code works fine but I need it to be 5 times faster.
Coder: [Looking at me like I just turned green and grew horns] Can't we just up the hardware requirement.
Me: Sure, if you want our customers carrying around laptops the size of suitcases.
Coder: All I do, in the code, is call the library functions.
Me: The library function is totally inefficient for the algorithm you are trying to implement. You need to recode the function manually.
Coder: But how do I do that.
Me: You need to write it in OpenCL and use the GPU.
Coder: [Turns white as a ghost]

If you want your skills to be more than sorting a list or changing the color of the font, you have to know what is going on underneath.

Comment: Re:Can this stuff be farmed out? (Score 1) 125

by byteherder (#48263241) Attached to: 16-Teraflops, £97m Cray To Replace IBM At UK Meteorological Office
The test for speed is not 16Pflops of raw computation but 16Pflop on the Linpack test suite. And no the cloud cannot do 16Pflops as they measure it on a supercomputer. You may be able to spin up more nodes to get more cpu power but I cannot spin up 100 new network connects and get 100x the bandwidth. Or get the sub microsecond latency of a supercomputer no matter how many connections you have.

Supercomputer are in a class by themselves.

Comment: Can this stuff be farmed out? (Score 1) 125

by byteherder (#48259915) Attached to: 16-Teraflops, £97m Cray To Replace IBM At UK Meteorological Office
There is a reason that organizations by supercomputers. There is no cloud in the world that can do what this computer does. None. Nada.

Cloud computing can run multiple copies of Office and host a website but when you need real horsepower, you get a supercomputer.

Comment: How fast is fast enouch (Score 1) 291

by byteherder (#48209955) Attached to: Will Fiber-To-the-Home Create a New Digital Divide?
Guys, do we really need 273 Gbit/s to the home that fiber provides. Isn't this just a bandwidth pissing match. My neighborhood has more bandwidth than yours.

I, for one, would be happy with 1GigE connection. Extra bandwidth would just bring diminishing returns. What education, what healthcare and social good needs that kind of bandwidth? So now you get your Youtube videos a millisecond faster.

Think of all the children that grew up using 28.8k modems. Oh, the humanity.

Comment: Re:Errors, what do we do (Score 1) 230

by byteherder (#47101583) Attached to: The Flaw Lurking In Every Deep Neural Net

We have trusted the computer to do math perfectly.

Its none ideal to have a single computational route and unit on safety critical systems.

I agree, the space shuttle had backup and redundancy for all the safety critical systems.
'We have trusted the computer to do math perfectly.' and then along can the Pentium Bug.

Comment: Errors, what do we do (Score 1) 230

by byteherder (#47099165) Attached to: The Flaw Lurking In Every Deep Neural Net
We know there will be errors with the neural nets. There will be edge cases (like the one described with the cat), corner cases, bizarre combination of inputs that result in misclassifications, wrong answers and bad results. This happens in the real world too. People misclassify things, get things wrong, screw up answers.

The lesson is not to trust the computer to be infallible. We have trusted the computer to do math perfectly. 1 + 1 = 2, always, but is not so for neural nets. It is one thing if the neural net will not tag the photo of your cat on Facebook even if there are 100 other pictures of your cat on your account. It is another if your photo get misidentified as being a terrorist on the "kill on sight" list.

The question is what do we do with the errors?

Comment: No Deterrence (Score 2) 108

by byteherder (#46844285) Attached to: Apple, Google Agree To Settle Lawsuit Alleging Hiring Conspiracy
If I was a CEO of a Tech company, I would be calling up all the other CEOs and saying, "Let's do this again". Why? $2.7 billion reason why.

Apple, Google, Intel, and Adobe conspired to hold down salaries of their employees, saving their respective companies $3 billion. When they got caught, they paid a $300million settlement and walked away. Net saving $2.7 billion. No admission of guilt, no one goes to jail, no one gets fired.

There is no deterrence for them to not do it again. No penalty, just a slap on the wrist. The penalty has to be at least the damage down in real terms. When you conspire to do something illegal and they only penalty is that you make $2.7 more in profit, you will never, never stop this behavior.

Comment: Re:Million Dollars (Score 1) 467

by byteherder (#46774509) Attached to: Survey: 56 Percent of US Developers Expect To Become Millionaires

Not exactly instant if it takes 30 years.

Not exactly instantly but it is typical of how people purchase a home.

Plus you're going to end up paying 3 million when interest is included with that principal, so your plan actually requires someone to pay 3 million in order to have 1 million.

Actually you will only be paying $616,560.88 total interest with a 3.5% mortgage over 30 years, $1,616,560.88 in total payments. The upside is that you are not paying rent and you get 30 years of appreciation in the home. Look at the price of home 30 years ago and compare them with today. That $1 million home could be worth $3-5 million.

Comment: Re:A million is easy (Score 1) 467

by byteherder (#46774405) Attached to: Survey: 56 Percent of US Developers Expect To Become Millionaires

Except 401k, IRA, and S&P500 index funds don't pay interest.

S&P 500 does not pay interest but it does pay dividends, currently around 2%. 401k and IRA can pay interest depending on what they are invested in. You are never going to reach $1million if you rely solely on traditional savings accounts paying 1%. If you limit your investment choices to only the safest ones, you will always get the lowest returns.

Compound that all you want and you're still going nowhere in a hurry.

Compounding is exactly what I want. 10% compounded over 7 years doubles my investment, 7 more years and it doubles again. With enough time all those doubles will turn a small investment into a large one.

"Don't worry about people stealing your ideas. If your ideas are any good, you'll have to ram them down people's throats." -- Howard Aiken