Announcing: Slashdot Deals - Explore geek apps, games, gadgets and more. (what is this?)

Thank you!

We are sorry to see you leave - Beta is different and we value the time you took to try it out. Before you decide to go, please take a look at some value-adds for Beta and learn more about it. Thank you for reading Slashdot, and for making the site better!



AI Experts Sign Open Letter Pledging To Protect Mankind From Machines

scruffy Re:I no longer think this is an issue (258 comments)

You misunderstand how AIs are built.

The AI is designed to improve/maximize its performance measure. An AI will "desire" self-preservation (or any other goal) to the extent that self-preservation is part of its performance measure, directly or indirectly, and to the extent of the AI's capabilities. For example, it doesn't sound too hard for an AI to figure out that if it dies, then it will be difficult to do well on its other goals.

Emotion in us is a large part of how we implement a value system for deciding whether actions are good/bad. Avoid actions that make me feel bad; do actions that make me feel good. For an AI, it's very similar. Avoid actions that decrease its performance measure; do actions that increase its performance measure.

The first big question is implementing a moral performance measure (no biggie, just a 2000+-year old philosophy problem). The second big question is keeping that from being hacked, e.g., by giving the AI erroneous information/beliefs. Judging by current events, we don't do very well at this, so I can't imagine much better success with AIs.

about two weeks ago

AdNauseam Browser Extension Quietly Clicks On Blocked Ads

scruffy I don't use an ad blocker, (285 comments)

but NoScript seems to block most of them anyway. I don't mind seeing a few ads, but I'm going to try to control what programs run on my machine.

about a month and a half ago

Confidence Shaken In Open Source Security Idealism

scruffy Open Source Tradeoff (265 comments)

Yes, the advantage of open source is that good actors can read the code and find and fix security flaws. The disadvantage is that bad actors can also read the code and find and exploit security flaws. One would hope good actors would outweigh the bad ones, but my fear that that governments and organized crime have become bad and worse actors in a big way. Even when a particular flaw is fixed, we all know that there are still flaws to be found and exploited in any big software project, and nowadays the big-time software exploiters have the budgets and the manpower to take advantage.

That said, that doesn't mean closed-source is any better (a different tradeoff), but it would be foolish to think that open-source software is not being exploited for its open-source properties.

about 3 months ago

Experiment Shows People Exposed To East German Socialism Cheat More

scruffy Re:But was it really unethical ? (619 comments)

I can't speak for Kilobug, but my answers would be:

1. It depends on your values. E.g., how much do you value your own welfare compared to family, friends, co-workers, fellow citizens, and those other people? If you want to be conscious about it, you need to think about what you value and how you might have done things differently in that light.

2. I probably thought I was I a deotonologist, but if you carefully study your own and other people's decisions, the vast majority are consequentialists with values that tend to selfishness. WItness how many Americans are angry about the Central American children/teenagers trying to get into the US.

3. As others have commented, doing a full analysis is time-consuming and uncertain (hence "maximum expected utility"). Most of the time, one has to follow rules that generally (so one believes) that have good consequences. And generally, virtue and duty are good rules. But people make up all sorts of rules with little sense behind them. My grandmother thought opening an umbrella indoors was bad luck, but I am a little skeptical about that one.

about 6 months ago

US Nuclear Missile Silos Use Safe, Secure 8" Floppy Disks

scruffy Security by Obscurity (481 comments)

Treating it as a maxim rather than as a caution.

about 9 months ago

Ask Slashdot: Books for a Comp Sci Graduate Student?

scruffy Rattiest Books on My Shelves (247 comments)

Knuth's books are very book, but they don't get much use from me. Instead:

Introduction to Algorithms by Cormen et al.

A good statistics book. Mine is an old thing: Mathematical Statistics with Applications by Mendenhall and Scheaffer.

A good operations research book (linear programming, queueing theory, Markov models/decision processes, and the like). Another old thing: Operations Research by Hillier and Lieberman.

Other than that, it's books that are/were used often for programming reference: Common Lisp: The Language by Steele and LaTeX: A Document Preparation System by Lamport look the most worn.

Hopefully, someone will come up with something a little more recent than the "old things" I mentioned above.

about 9 months ago

How Ireland Got Apple's $9 Billion Australian Profit

scruffy Tax Corps Based on the CItizenship of Their Owners (288 comments)

Really, the "location" of these mega-corporations is a sham.

Instead, figure out (or estimate) what percentage of the shares are owned by US residents. Multiply that percentage times the corporation's profit times the corporate tax rate and that is what they should pay.

Note: Any public corporation knows who are the immediate owners, so that they can send out shareholder info. However, a shareholder might be another corporation which is owned by other corporations, etc. Hence, the need to estimate (along with following the money as much as possible).

about a year ago

DARPA Tackles Machine Learning

scruffy The Reasons for "Herculean effort" (95 comments)

Raw data need to be cleaned up and organized to feed into the ML algorithm.

The results of the ML algorithm need to be cleaned up and organized so that they can be used by the rest of the system.

No one (currently) can tell you which ML algorithm will work best on your problem and how its parameters should be chosen without a lot of study. Preconceived bias (e.g., that it should be biologically based, blah, blah) can be a killer here.

The best results typically come from combinations of ML algorithms through some kind of ensemble learning, so now your have the problem of choosing a good combination and choosing a lot more parameters.

All of the above need to work together in concert.

Certainly, it's not a bad idea to try to make this process better, but I wouldn't be expecting miracles too soon.

about 2 years ago

How Do We Program Moral Machines?

scruffy Re:Why I doubt driverless cars will ever happen (604 comments)

Yes, plus the fact that this kind of decision policy is already evolving with collision avoidance systems in some cars (and experimental self-driving cars). It's not going to be a sudden mystery to be solved 30 years from now.

more than 2 years ago

Rise of the Online Code Schools

scruffy Online doesn't work for average students (so far) (98 comments)

One of the biggest issues for current MOOCs is the large attrition rate (in the 90% range). Assuming that people signing up are at least average intelligence (on average of course), this suggests that average students are unable, for whatever reasons, to complete these courses. Part of it is that the instructors come from elite universities, are used to teaching elite students, and approach the MOOC in the same way, leaving the average student in the dust. Another part is that average students lack the motivation, discipline, as well as the smarts to learn complex concepts without a real-life instruction.

about 2 years ago

The Information Age: North Korean Style

scruffy The Walled Garden (156 comments)

Why doesn't North Korea just use some version of Apple's walled garden? It sounds perfect for them.

more than 2 years ago

Ask Slashdot: How To Become Informed In Judicial Elections?

scruffy Re:Local Newspaper (153 comments)

The local paper will typically have recommendations of how to vote for as well. It's not ideal, but you might trust them enough to weed out the crazies.

more than 2 years ago

The Rage For MOOCs

scruffy Coursera heavy on math (109 comments)

After taking a few courses from Coursera, a high dropout rate is not surprising. The CS courses are mainly math courses in disguise, which works when you are teaching CS students at the high end of the intelligence spectrum, like at Stanford and other top-tier colleges, but simply loses most students otherwise. Even the NLP course was very focused on the mathematical models, much less so on the linguistics.

I suppose many might say it's not computer science without the math, but you can still teach much about computer technology and software design while being gentler with the math.

Personally, I've enjoyed the courses because I like math (except the quantum computation course, which was dreadful), but I know most of our CS students would be buried by the math. For the record, I'm at a state univ with some good research, but nowhere near a flagship. We do want to graduate some students, and the students we do graduate are in demand in our area.

more than 2 years ago

With 'Access Codes,' Textbook Pricing More Complicated Than Ever

scruffy Why I Don't Require Supplements (400 comments)

It's not only the extra cost, but it's also a loss of control over private information of the students.

more than 2 years ago

Why Professors Love (and Loathe) Technology

scruffy What I use CMS for (113 comments)

As a CS instructor, I use Blackboard for homework and program submission, for posting solutions and for recording grades. Nothing else. Making a full-fledged web site out of Blackboard is too terrible to think about.

more than 2 years ago

Poison Attacks Against Machine Learning

scruffy Is the future like the past? (82 comments)

I'm not sure why this would be surprising. ML algorithms work best if the future behaves like the past, if it has the same probability distribution as the training data. Some algorithms can handle slow changes if they can continually get new training data, but large changes is a problem.

more than 2 years ago

Positive Bias Could Erode Public Trust In Science

scruffy Same Old, Same Old (408 comments)

This has always been the case. Science is not a uniform march to the Truth. There is a difference between well-verified and understood results (think engineering) and working at the margins with not much data and the usual human failings (the vast majority of publications). Scientists are humans, not gods. It takes a lot of effort and error to get to the well-verified and understood part.

more than 2 years ago

Rep. Darrell Issa Requests Public Comments On ACTA

scruffy Re:I for one have new hope... (186 comments)

If serving meat had as great of an effect on public health as providing universal birth control, then absolutely.

This is the key point. There is an enormous amount of evidence that birth control improves public health. From a scientific point of view, it is a no-brainer.

About half of all pregnancies in the United States are unintended. Regular use of contraception prevents unintended pregnancy and reduces the need for abortion. Contraception also allows women to determine the timing and spacing of pregnancies, protecting their health and improving the well-being of their children. Contraceptive use saves money by avoiding the costs of unintended pregnancy and by making pregnancies healthier, saving millions in health care expenses. Several contraceptives also have non-contraceptive health benefits, such as decreasing the risk of certain cancers and treating debilitating menstrual problems. Making contraception more affordable is a significant step forward for the health of women and their families.

This quote comes from http://www.rhrealitycheck.org/article/2011/11/03/testimony-of-douglas-laube-md, which also includes citations for the above points. I would also include reduction of teenage pregnancy as an additional plus.

more than 2 years ago

Academics Not Productive Enough? Sack 'em

scruffy Re:publish shit! (356 comments)

Publish any shit you can! That's the best way! unfortunately, that's how academica works.

That is going to be the result of this kind of policy. I know in Computer Science, there are lots of bad conferences and journals that are easy to publish in.

more than 2 years ago


scruffy hasn't submitted any stories.


scruffy has no journal entries.

Slashdot Login

Need an Account?

Forgot your password?