7 TED Talks on Artificial Intelligence (was: Starve the Bots!)
Please Do Not Vote for this Top Level Post!
Introduction
Anyone who has read my On Bots series will be surprised by this post's title, because I am an out of the closet fan of our automated steemsymbionts. I think they are just about the best thing since sliced bread. Why, then, would I want to starve the bots?
Answer: Because I am a huge fan of what bots can do for us, and I want to help them reach their full potential. What bots are doing now is just the beginning. We should always demand improvement from our electronic playmates. Right now, it is my opinion that most bots (including my own) ignore comments and only vote on top level posts. This dysfunctional voting pattern encourages people seeking author awards to only post blog-style articles and it discourages discussions. This is a problem, and I want to encourage the bots to get better. So, I'm going to intentionally try to devalue that voting pattern with this post.
[Image Source: pixabay.com, License: Creative Commons, CC0, Public Domain]
To do so, I thought I would try a little experiment by hijacking the #weekly7 tag from my son, @cmp2020, and adding a couple twists to hopefully create a discussion style post. The first twist is to ask you not to vote for this top level post. Repeat: Please do not vote for this top level post. Instead, please use votes on the comments of this post to express your thoughts about the value that the post contributes to steemit. If we move enough curation rewards from the posts to the comments, the bots will start to get hungry, and they'll have to get a little bit better at voting. This is my tiny contribution in that direction.
How Does it Work?
In @cmp2020's #weekly7 posts, he posts 7 videos on a topic along with a brief discussion of each of them. I'm going to do the same here, with 7 TED talks on the topic of Artificial Intelligence (AI), but the second twist is that I'm going to discuss the videos one day at a time in the comments of this post, probably starting Saturday afternoon. So, if you want to encourage our electronic friends to start making better voting decisions, please consider checking back after that for the daily commentary on the videos and voting on those comments. Please also consider voting in the comments of other peoples' posts whenever that option is available. Until the bots catch up, any sizeable comment votes will divert curation awards away from bots and towards human/manual voters.
Also, over the course of the week, discussion from others is vigorously encouraged.
Videos
For this experiment, I have chosen seven TED videos on the subject of Artificial Intelligence (AI). To the best of my recollection, I haven't seen any of them before, so this might go off the rails fast. ; -) I have also put the videos into a youtube playlist. Here they are:
7 (Saturday)
Don't fear superintelligent AI
Grady Booch
6 (Sunday)
How computers are learning to be creative
Blaise Agüera y Arcas
5 (Monday)
The rise of human-computer cooperation
Shyam Sankar
4 (Tuesday)
Can we build AI without losing control over it?
Sam Harris
3 (Wednesday)
Machine intelligence makes human morals more important
Zeynep Tufekci
2 (Thursday)
The rise of personal robots
Cynthia Breazeal
1 (Friday)
Get ready for hybrid thinking
Ray Kurzweil
Conclusion
What happens next? Probably nothing, I guess. After all, my daily comments won't show up in anyone's "Follow" feed, and the guilds aren't going to be voting on comments either, but a journey of a thousand miles starts with a single step, right? And this is all moot with the second curation pool for comments in hardfork 17, anyway. But even so, there's one final point that I wanted to make by structuring this post the way that I am. We can change things on this platform through our own individual decisions and actions. We don't need to call for "them" to change rules and try to dictate our preferences to others. When we encounter a problem, we just need to apply our ingenuity to propose a solution and set an example for others to follow. If we're not willing to start by changing ourselves, then maybe we shouldn't be trying to force changes upon others.
So let's just see where this goes.
@remlaps is an IT professional with three decades of professional experience in data communications and information systems. He holds a bachelor's degree in mathematics, a master's degree in computer science, and a master's degree in information systems and technology management. He has also been awarded 3 US patents.
This may not be entirely on topic, but...
The dumbest thing about the proposed comment rewards pool is that it eliminates the incentive for comment voting. Currently, comments reward the full 25% of curation rewards to voters, while blog posts are only distributing about 12% of the pool to curators.
If we allow Steemit Inc. to continue making poor decisions that disincentivize the very behaviors that they want more of, then we will continue to see greater distortions in the platform's economics.
I'm sorry, but I upvoted this post because I wanted to reward you as much as I could with my trail. Unfortunately, some of my followers opt out of comment voting. Also - your reverse psychology is impeccable!
Thanks! I was already feeling guilty about my loyal bot followers going hungry on this post, so I'm glad you voted it even though I asked you not to. ; -)
5 (Monday)
The rise of human-computer cooperation
Shyam Sankar
According to his blog, Shyam Sankar is the "Director of Forward Deployed Engineering at Palantir Technologies," and he believes that, "a handful of computer scientist organized around a higher purpose can change the world." According to the youtube video description, he is a data mining expert who believes in "human-computer symbiosis," a concept which he discusses in this TED talk, so let's get to it.
Sankar starts off by talking about two games of chess. The first, in 1997, when Gary Kasparov lost to IBM's Dep Blue. He points out that when it happened, many people took it as a sign of the triumph of AI over the human intellect, but the reality has been different. The second chess game was a freestyle tournament where humans and computers competed on teams in 2005. In this tournament, A supercomputer was beaten by a grandmaster with a weak laptop, but the winner was the huge surprise. The winning team was comprised of two humans and three low-end laptops.
Sankar uses these examples to represent two competing AI philosophies. Marvin Minsky's vision of fully autonomous AI, and J. C. R. Licklider's idea of human computer symbiosis, or "Intelligence Augmentation." This vision was to enable man and machine to cooperate in a flexible way. Besides chess, he lists a series of examples, one of which is Fold-It, a video-game where non-technical, non-scientific gamers compete to solve problems in protein folding, in many cases better than expensive supercomputers or highly trained scientists.
According to Sankar, the most important thing to focus on when designing big-data analysis is friction. Humans need to be designed into the process in a way that minimizes friction, and the key to success is, "the right type of cooperation."
He closes out the video with another series of examples to support his argument. One key point that he makes will come up again in How computers are learning to be creative. Sometimes, the way to solve an AI problem is to look at the data in reverse.
This comment is for discussion of the "starve the bots" strategy.
This seems pretty interesting. I have no idea how to use bots anyway as of yet, all my votes are manual. I wonder if this will lead to anything progressive.
I guess I'm not expecting it to really go anywhere, but I thought it would be worthwhile to try it out and see what happens. At any rate, I'm going to enjoy watching those videos. ; -)
I personally think bots are not useful for discussion. Only people can truly curate that, and it should happen organically. A bot's place is in post content IMO.
Maybe so. It's definitely much harder to go deeper. Historically, though, "Only people can..." has usually been wrong. If the curation rewards change weren't on the horizon, I was going to start looking at comments next for my bot. With that change coming, though, it's really just a thought exercise.
Oh I think bots can be created to do it, I don't ascribe to the "Only people can" mantra in this regard. I have a strong belief in peoples ability to design a program to do pretty much anything we can imagine. The question is can a bot's voting foster growth of organic conversation between people? I don't think so. It is the conversation in the comments section that truly drives their value, with out it, a lot is lost.
Yeah we will have to see how the changes effect things.
As far as directly influencing the conversation, I agree, it would be difficult. Indirectly, I guess its most obvious way to enhance the conversation would be for people to use its vote as a screening filter to decide which conversations to read or join.
Bots are not stupid nor bad. They are just following rules to achieve a goal. The problem is that the goal of them with voting is earning reward, not selecting good contents. Just giving more incentives on curation reward does not alleviate this problem. Asking people to stop bot is like a government asks to conserve water while not charging water utility fee.
What we need is fixing incentive system. I shared my opinion in this post (while ats hates it 😅). The main argument is not shrinking or removing incentives but creating other types of incentive to diverge profit-driven bots/users.
Thanks. I saw that earlier, but I didn't have time to post a comment when I read it. I'll try to remember to comment on your post tomorrow in order to keep my reply in the same thread with your topic.
Looking forward to it!
103 votes 16 views... You are kinda late for this kind of request, lol.
Yeah, and about 30 of those votes are from the stupid "Winfrey" bots that have been set up, and that do pretty much nothing useful. And about 45+ are from me and my trail.
Anyway - the views and vote counts are pretty much irrelevant on any given post. Most of the content on Steemit goes unread by close to 99% of the active user base anyway.
From the moment that you've an upvote button at the GUI in the frontend that most users use... You can be sure that "people" will only upvote based on the author's payout history and current pool of rewards.
Oh well. At least I'll enjoy watching the AI videos.
Nice idea! I've been irking you to start a weekly7 for a while. I would really like to see Steemit evolve so that reasonable discussions in the comments are paid reasonably by both bots and users. I hope they decide not to go through with removing citation for comments. See you at dinner! :)
ok. Thanks for the #weekly7 idea.
6 (Sunday)
How computers are learning to be creative
Blaise Agüera y Arcas
Missed this one yesterday. So here it is a day late. According to his TED page:
In this video, the speaker says that his work started with machine perception, and unexpectedly connected with creativity. Accordingly, he takes the perspective that, "the flip side of perception is creativity." Starting with 19th century brain maps that were made using microsopy and dyes, he talks about the scientific effort to map the brain's functionality, and shows that neuron maps that can be constructed from those microscopy efforts. In the realm of computing, these neuron maps can be translated to circuit diagrams.
From that framework, he constructs a vector equation for perception, representing an image as X, a neural network as W, and the perceived object as Y, he gives us the equation:
X * W = Y
He demonstrateds this concept with an example picture of a bird and describes how the W vector is found through an iterative training process with large numbers of examples. The surprise comes in after the W vector is built and perception is working. One of his researchers decided to reverse the equation by solving for an unknown X:
X = W * Y
and they discovered that they could create images of the types of objects that the computer knew how to perceive. This video is full of graphical examples, so you really do need to watch it to appreciate what he's saying, but it is fascinating.
He doesn't say this, but it occurs to me that the main point, that perception and creativity are intertwined, may also be relevant to human learning. In music theory class, my son is learning that the way to learn to write music is by reading music by the great composers. When humans learn literature, we typically learn to read before we learn to write. However, in computer science, we normally learn to code by writing code. I'm not the first one to ask this question, but this video brings it to mind. Could computer science education be improved by teaching students to read code before asking them to write it? This lends some support to the idea.
If you watch the video, please let me know what you think?