HOME | DD | Gallery | Favourites | RSS
| Sushimitzu
Sushimitzu
[4443095]
[2007-03-27 12:21:43 +0000 UTC]
(Germany)
# Statistics
Favourites: 39; Deviations: 1; Watchers: 189
Watching: 62; Pageviews: 40316; Comments Made: 28643; Friends: 62
# Comments
Comments: 4564
Thanks for the llama! Enjoyed reading your very fascinating post on transhumanism!
š: 0 ā©: 1
No problem, and many thanks for taking your time to read that! Glad you found it fascinating.
š: 0 ā©: 1
Vielen dank for the llama (Please accept this llama in return)
š: 0 ā©: 1
Ha, nice. 
š: 0 ā©: 0
Hey dude, I saw this TED talk recently that I'd rather like your opinion on.. it also explains why I dislike social media so much.
www.youtube.com/watch?v=iFTWM7ā¦
š: 0 ā©: 2
I attended the GRACE Symposium 2017 yesterday. One event was a discussion on stage between several computer scientists regarding where they want to see software engineering go in the next decade. A few speakers proposed increasing the intentionality, i.e., bringing software engineering closer to the real world application (meaning more domain specific) so the impact of the system is clearer. Some others proposed increasing dependability, i.e., making sure that systems do not go haywire and do not do unforeseen things (basically the worry of "we're growing intelligence that we don't truly understand"). What I disliked about the latter is that making lack of dependability a common fear could undermine intentionality by giving software engineers a go-to excuse to not care about sociopolitical issues and to disregard potential outcomes ("as so often with AI, I didn't know it would turn out this way"). When I mentioned this, some speakers agreed while others claimed the two efforts are not mutually exclusive.
š: 0 ā©: 1
Just wanted to apologize and 'run' (by all appearances), I've read your comment(s) many times and I thank you for sharing them, it's just I'm intermittent on here now and don't have a lot of time but do have a lot of problems.
š: 0 ā©: 1
Don't worry about it. I've learned early that the likelihood of a reply from you decreases as the length of my comment increases. 
š: 0 ā©: 1
But you can be absolutely sure that I do read them.
Thanks, mate. That music was also a long way from what IĀ was expecting... haha.
š: 0 ā©: 0
It's like we're not programming anymore; we're growing intelligence that we don't truly understand.I have to point out that AI developers are not surprised by this. In fact, the motivation for moving from traditional programming to machine learning is that the developer no longer needs to understand the problem enough to be able to formulate it as a traditional algorithm. Instead, the developer just gives the computer examples of which input leads to which output so the computer can figure out the pattern and fill in the gaps for other, yet unknown examples. Mathematically speaking, artificial neural networks approximate an unknown function (which is too complex to be formulated by hand) by generalizing from the given data points.
Without meaning to get into technical details such as overfitting, the more examples (i.e., data points) are available during the learning phase, the more accurate will be the approximation later, making the results more useful. (A running gag of a colleague of mine is that every paper on machine learning ends with the conclusion that the presented artificial neural network works but would perform better if they had more data.) Additionally, the developer might not even know which kind of input data is important (i.e., what discriminates the function to be approximated). In that case, just throwing more data (which may or may not be relevant) at the artificial neural network can help the learning process by increasing the odds that there is a pattern among the data that yields useful results.
So what is the pattern (i.e., the approximated function)? There have been attempts to understand what is going on inside artifical neural networks (more specifically, what kind of input activates them the most at certain stages) by for instance visualizing activation . However, as Tufekci (from the TED talk) points out, it is like trying to figure out a person's thoughts from a cross-section of their brain. Personally, as a computer scientist, this lack of transparency does not cause me concern, because I know it is just an approximation of an unknown function. Hence, I am glad she starts off with the disclaimer that AI is not like Terminator, unexpectedly going haywire.
And yet, she gives the example of YouTube's machine learning algorithm picking up on more and more extreme political opinions as patterns that increases viewer retention. This is not something I can confirm from my own experiences, but even if it is true, there is nothing spooky about it. It is a pattern in data that, if effective when exploited, reveals something about human nature.
One of her major concerns is the sociopolitical issue that, in order to make the artificial neural networks as useful as possible, companies may wish to acquire as much data as possible. The issue here is actually less about AI but more that companies have all this possibly personal information that they can abuse. But yes, the existence of machine learning does incentivize gathering more information (for reasons mentioned in the second paragraph). This is especially true because of the connection with ads. I will talk about ads later, but I believe Pandora's Box has been opened and the prospect of powerful machine learning algorithms means many companies would keep gathering data even if ads stopped being a thing tomorrow.
What is about AI, however, is that it can be used to find more data (or rather guess fairly accurately) using other, seemingly unrelated data. When the found information is personal, it is once more easily abused. On that topic, Tufekci claims that 1984 is not the correct dystopia for the 21st century, and I agree; modern slavery is not through overt fear. Instead, she believes it happens through subtle manipulation.
While I do not want to deny the existence of subtle manipulation, I prefer to look inward rather than outward. In particular, I always felt that Brave New World is a more realistic dystopian vision than 1984. Not a single evil mastermind forces us into submission but we ourselves do willingly. After all, there is a reason we watch these 27 videos from YouTube's "up next" list. Nobody forces this upon us.
For similar reasons, I think her proposed solution to basically radically alter software development is fruitless. It requires too much force from outside, whereas it is very easy to just fall into the same old patterns by default.
Now, as for ads, they are obviously one of the great motivations to employ machine learning. I personally do not watch TV because I find it ridiculous to sit through 10 minutes of consumerist indoctrination for 30 minutes of (probably garbage) content. On the internet, at least I can install an adblocker. Speaking of, surely you have heard this argument against adblockers along the lines of "content providers need to finance their servers". While I not only find it difficult to feel empathy for companies, I would go a step further and argue that it is morally questionable to not use adblockers. After all, it supports and thus further establishes the practice of showing ads for revenue, which motivates some of the issues mentioned in the talk.
Did you hear about Coinhive? Basically, some websites include a script that makes the visitor's browser mine for crypto currency on behalf of the website's owner. This indirectly transfers money from the visitor in form of electricity and processor power to the website's owner. This is causing public outrage because it is typically done without the knowledge of the visitor. To me, if done transparently, this would actually be a much better alternative to ads because it does not come with the issues raised in that TED talk, does not distract when browsing and does not infest the mind with nonsense. Unfortunately, I doubt it will ever be an alternative but just another means of grabbing money in addition to ads.
As a side note, isn't it funny how both of these means are indirect transactions? It is all about money, so one would expect it involves paying the content provider with money, yet it is done through ads or mining currency. Capitalism being as efficient as always.
Also, obligatory music link . While I am at it, yesterday I went to a club for the first time. Specifically this place called Womb. It is right here in Shibuya, Tokyo, where I live right now. A certain DJ Kyoko was playing among others, and it was nice even just to watch her enjoy her own music so much. Adorable.
š: 0 ā©: 0
š: 0 ā©: 1
It's March 25th which means it's that time of the year again and your special day is here! We hope you have an awesome day with lots of birthday fun, gifts, happiness and most definitely, lots of cake! Here's to another year!
Many well wishes and love from your friendly birthdays team 
Ā Ā Ā Ā
---
Birthdays Team
This birthday greeting was brought to you by: LDFranklin
š: 0 ā©: 0
Reasonably flashy for the occasion!
š: 0 ā©: 0
How have you been doing of late, my friend?
š: 0 ā©: 1
Heya, my favorite Aussie. I'll reply by note.
š: 0 ā©: 1
You forgot to address me as sire.
š: 0 ā©: 1
Why? Who did you marry?
š: 0 ā©: 0
thanks for taking the time to give me a llama!
š: 0 ā©: 1
DA notified me of MoonlitCrescent talking to you about me, so I went ahead and gave you one of them llamas.
š: 0 ā©: 3
ohhh i see
he told me about you and i checked your work out
its amazing!
š: 0 ā©: 1
Appreciate hearing that. 
š: 0 ā©: 0
xD Good to see you're still around here.
š: 0 ā©: 1
I still check this site daily. However, I comment way less. My pageviews are about to exceed my comment count!
š: 0 ā©: 1
Good to know, I will try and remember that 
š: 0 ā©: 1
*sees you seeing this*
š: 0 ā©: 0
Nothing to thank for.
š: 0 ā©: 0
| Next =>