Slashdot is powered by your submissions, so send in your scoop

 



Forgot your password?
typodupeerror
×

Comment Re:Cash is so much better. (Score 4, Informative) 186

Your ordering is wrong.

The correct ordering is:

1) Cash: 15 seconds or less
2) Credit/debit card: 45 seconds or more
3) Smart phone: 1 minute or more
4) Checks: 2 minutes or more

What kind of lame POS system does your coffee shop have? When I go to Starbucks or Peets, it takes me the same amount of time to hand over my card as it does to hand over cash, the difference being that it literally takes them only a second to swipe it, and by they time they hand it back to me, the transaction has already been approved, no signature required.

I don't see how cash could possibly be faster unless I hand them exact change, but even then they still have to count the bills and put them in the drawer, so even if *my* transaction is faster, the next patron has to wait.

Comment Re:... and this is surprising how? (Score 1) 153

Do you expect that your smartphone's voice recognition is all happening in the phone?

No, but I expect my smartphone to recognize more than "channel up" "volume down". Even my 2005 era feature phone could recognize key phrases without sending the audio anywhere. Surely a modern TV with multiple gigahertz CPU cores can do the same.

Comment Re:... and this is surprising how? (Score 1) 153

But it's not a secret. You know when you buy one of these your voice is going to be transmitted over the internets for analysis. You would expect them to take some obvious steps to secure the potentially private information from third parties but there is nothing "secret" about the collection and transmission of the user's voice. The only potential violation of privacy here would be the ability for a third party to intercept the unencrypted data on someone.

Why would someone think this? If my TV had voice detection, I would expect it to all happen locally, I certainly wouldn't expect the TV to record me 24x7 and send snippets of conversation to a central server for analysis.

Comment Re:And so it begins ... (Score 1) 158

Of course. One of the articles I found proposed to do brute force solutions, and given the terabytes of code out there, it should be possible. Even the creation of the original modules should be open to brute-forcing.

Having terabytes of code to choose from does not make brute forcing any easier.

But note I described the results I wanted - not the code to achieve them. And since I've already written code to do it that way around the turn of the decade, and I don't have the resources to brute-force code creation, I'm figuring I'll go with automated code generation from a simple wish list.

Sure, natural language processing is becoming more refined and will continue to become more powerful. But that's not brute forcing - the natural language processor doesn't piece together random combinations of code to give you what you asked for, it already has algorithms to retrieve data from a database, perform transformations and updates, etc, so it puts together the code logically, not through blindly pasting code snippets together to see if it does what it needs.

Comment Re:And so it begins ... (Score 1) 158

Example - you have several different database crud operation code modules to choose from.

You also have several different database user interface to pick from.

Additionally, you have several database schemas to choose from, including different indexing options depending on what is important to get fast, the mix of reads, deletes, writes, and rewrites, etc..

And several data input modules - keyboard, external data feed, whatever.

And several logging modules, each compatible with the back end.

And several different error-reporting modules (do we put up a user alert and give a chance to edit it, do we not allow it and send a text message to a phone, whatever).

It would be able to give a list of data we want, like Name, etc. without specifiying the data size or internal type, because that's all been standardized (last name, first name, middle name, etc).

Given the requirements in more or less plain english, it should be possible to come up with the optimal solution pretty quickly, since each module has standard interfaces to the others.

For example, I need a way to track a million people. The information that's mandatory is their name and address, date of birth, and gender. When the address changes, the old address should be preserved so that I can trace back if necessary.

Optional fields are cell number, email, home and work phone numbers, and 1 or more emergency contacts. When any of these change, the old ones should be preserved so I can trace back if necessary.

Initial input is via a record dump on a usb key stored in SDF format, with updates being done by either using the same method or by someone typing them in.

A sequential account number should be auto-generated for those records that don't have an account number from the initial dump. The account number is 2 letters, 6 digits, then 2 random digits to help detect bad account numbers.

I should be able to search by account number, name, or any phone number.

Input data from the initial dump should be flagged if not valid, and input data from later should only be entered if all required fields are there.

Auto-generating such an application should be doable now.

What you are describing now is not what you were describing earlier. Earlier you suggested that a computer could "mix and match billions of code snippets already written and brute-force" a program.

What you're suggesting now is that you want the computer to parse your natural language of a problem and turn that into a program.

Do you not see the vast difference between the two?

Comment Re:And so it begins ... (Score 1) 158

If your code snippets are all working modules (not just functions) will well-defined inputs and outputs, the brute-forcing takes on a different meaning - combining modules that have the appropriate inputs and outputs with each other. Entirely doable.

That doesn't change the numbers - if you have a reasonably sized library of code modules and try to brute force an app by putting them together randomly to see what you end up with, it's going to take an obscenely long time. And it's not even clear how this code writing AI will know when it gets a useful app -- It may create a working calculator app that can only calculate the cosine of base 13 numbers... it's a valid app, but is it a useful app? How would it know? What about the billion other apps that do a similar task, like calculating the sine of base 213 numbers?

Comment Re:indirect jobs (Score 1) 158

I get entire racks prebuilt shipped in and out, it just takes one on-site person to plug it in then the shippers get back in their truck and go back whence they came.

That said, if the company cited really does only have one employee, I'm glad not to do business with them. I'd require a bare minimum of three to at least pretend someone is on site 24x7.

If you have a lights-out datacenter, you don't need employees on-site 24x7 because replacing hardware is not a time-critical task, if a disk drive, server or switch fails, you can replace it during your next monthly maintenance sweep. If a power system failure takes out half the datacenter, you failover to your backup datacenter while you wait for a repair crew to arrive at the failed datacenter. Whether you have an employee on-site or not, he's not likely to be able to fix anything himself anyway.

You don't need people on-site to monitor the datacenter, you can have everything monitored by your NOC on the other side of the country.

Comment Re:And so it begins ... (Score 1, Informative) 158

That is, unless you are a software engineer.

Hahahahahahahahaha ... guess you didn't follow all the links in all the articles to supplementary material. One makes a darned good argument for the elimination of writing software by having computers do it. And why not - a computer can mix and match billions of code snippets already written and brute-force the "creativity" out of creating software by testing each one. I give it 20 years.

If you have a library of 25 code snippets and need to find the magic order to combine just 10 of them to do your task, that's around 1.1 x 10^13 combinations that need to be tested. So around 10000 more than a "billion".

if you have a library of 50 code snippets and need to find the magic order to combine 25 of them to do your task, that's around 2 x 10^39 combinations. If you can test a billion billion (1e18) every second, it would still take 60 trillion years to test them all.

I don't think brute forcing code is going to replace anyone's job.

Comment Re:Sigh... Yet another scam (Score 1) 233

On their site, they put the total sponsorship/marketing revenue of the Olympic games at something like $8Bn/year, on the same order of magnitude as each of their four-person colonisation missions. (There will be many of these.) Of course, the Olympics is the culmination of the entire competitive calendar for a wide variety of spectator-friendly sports and athletic activities, operated with prominent sponsorship all over the arenas and participants. Whereas space missions have demonstrably niche spectator interest and occur in the most barren and advertisement-free areas in the universe.

Since there's a good chance that all or some of the Mars mission astronauts will die, plenty of people will watch. The sponsor better hope that they don't die during launch since that will severely limit the revenue they can generate.

Comment Re:No more or less than anything else (Score 4, Insightful) 323

I'm an EE who works as a programmer, who is currently working on a masters in CS. The general engineering skills involved in EE make me a better programmer than 90% of the CS majors out there.

With few exceptions, I've found that when someone tells me that they are better than 90% of the X out there, I've found that to be false and that the person just doesn't know how little he knows, and he'll go on at length about some arcane little anecdote that "proves" how smart he is.

Comment Re:Worked for drugs (Score 4, Informative) 154

I'm not sure methamphetamine is a direct replacement for Sudafed. You might want to double check that.

The point isn't that Sudafed is the same as meth, but that it is restricted because it can be used to make meth, yet I can buy the meth directly easier than I can buy Sudafed.

http://www.forbes.com/sites/da...

Also, this war on meth has resulted in the pharmaceutical industry selling what is essentially a placebo in the form of a "meth-resistant" Sudafed PE:

http://consumer.healthday.com/...

Seven other studies, according to the authors, found that phenylephrine didn't work better than a placebo.

"It does nothing," Hendeles said. "Clearly the 10 milligram (dose) does not work."

So consumers are being guided into buying a product that doesn't work by a drug policy that also doesn't work.

Slashdot Top Deals

E = MC ** 2 +- 3db

Working...