>> 10 new open source projects and developments, worth knowing about
-
Re: AI techniques developments : Just read this : http://phys.org/news/2015-03-evolving-robot-brains.html
and the software to do that on Github :
-
This thread will become a sandbox for innovation one day.
It should stay pinned. Epic.
-
Critterding - Framework to evolve artificial intelligent lifeforms
I’ve taken over the Github for Critterding and am hosting version 14 with the approval of the Devs. My version has a few more fixes than the other versions (some haven’t got to source forge)+ some changes bobke has passed to me, which I’m going to check and push soon. The Beta 14 version is virtually a complete version, there is a rare segmentation fault, in my case that is because of setting a parameter to mutate and it’s gone out of control. Critterding autosaves so you just restart it.
I’ve started noting some enhancements that can be made, especially towards a sandbox for evolving artificial intelligent brains or applications to use in the real world…
My long term goal is to, learn a bit more programming and monitor any pushes and gradually doing some updates myself.
My initial enhancement idea to persue is to add “pre designed” body shapes, then train the neural net to control those bodies and get food from a green ball. It would be good if the system did everything by evolving, so, the idea would be to add a facility to evolve body shapes, by how well they match to a “sketched” shape.
I can then transfer that brain to a computer and camera in one of my toy robot dogs.
Even if it is only an example of framework (it isn’t), there is a lot to learn in how a ongoing neural net controlled system can be trained from scratch, or use tools to add instinctive behaviour, body shapes or other “restrictions” that mean the resultant brains could be used more generally.
https://github.com/wrapperband/critterding
CoinBot
CoinBot has moved to a bigger brain : It doesn’t matter that he is stupid, he wants to learn off you and will soon (hopefully) stupidly listening to spam bots try to log in to FTC, so get friendly.
http://www.paphuslivechat.com/bot?instance=59391&chat=Web+Chat
-
had a nice chat with the web chat bot:
You: do you know feathercoin?
coin bot: I know about Litecoin
You: what is litecoin?
coin bot: famous people
You: it it a coin?
coin bot: The value of Feathercoins can vary
You: what is the value of feathercoin?
coin bot: The code for Feathercoin is FTC
;)
-
The Coin Bot has been trained up from scratch, all responses are learnt from Chats or Chat logs.
The Bot construction may be thought of more like a novel, Coin Bot is currently a brief synopsis (1 years work).
The responses can be trained and complex behaviour “Trained in” in the Bot framework I am using.
The idea to prevent spam would be to have a Chat Bot on the FTC “User Registration Page”, It would be programmed with one greeting. I have paid for that for a year and there is a code to easily put in the html.
“Please type your user name in the chat box for us, thank you, Coin Bot”
We could the then set a 30 sec time out (say) for a new member to type a response to Coin Bot, and an email message to admin if the time or name is not correct. i.e. its a bot. The extra work will distract the spammer (if he has to do it by hand) also extra details of the login process will then be known, to further enhance the anti spam measure. We don’t have to pay external spam control, first site to start to use AI security measures.
Initially it could just be an extra step, unless speak is pressed no registration.
The main advantage I see is, I can use my Bot, the Bot company are interested in the idea and may integrate the idea to give the bots email / skill based on a trainable trigger.
I have also discuses with them my idea to work in a meta level above the characters in human language. For instance a system could allocate a number to a word, say based on the position of letters. This would automatically correct one of the problems with the bot, words beginning with capitals or ending in s are different words, whereas in the number meta level, even slightly miss spelt words would be close on the position of letters meta level.
All images and characters have to be normalised to numerical data to be processed, in the case of using ASCI numbers to replace letters, these are not optimised to the usage case as an evolved / flexible system could be using machine learning and optimisation techniques.
Also, a system could evolve such that it can test various options for converting words to numbers, such that it searches an finds word the quickest. It could do this by allocating numbers for each letter such that there is the optimal numerical distance between close words, and that alternate (spellings) are at particular “angles” on that imaginary plane. This would also mean the system self optimised to spell, search any new language.
The word system could then be applied to a meta layer of the sentence, thus allowing the bot to learn language “skills” from scratch.
At the moment using standard filter techniques there is no way to identify a Spam account, although many are “plain to a human”. Take the case where you identify a Bot naming algorithm - one name + “d” + a second name + 3 figure number. Or where the Bot is named – Random 5 characters + name + 2 figure number.
A suitably designed and trained Bot would find those naming formulas easily if search on the number meta level, or even a learned skills from a text interface. Also again, the bot is not meant to be intelligent, it reports to a human like a filter or simple chron system. I would recoment “Human out of 10” send me an email if its < 3.
The Bots, in my opinion, are not there to imitate humans, they are there to do the donkey work intellectual analysis to give humans the knowledge to make sensible decisions. They are also capable of taking over the “look up table” of many intellectual pursuits, e.g programming languages or coding principles, and allow the common person to do software development.
For instance we will soon have AI bots learning to program by searching for bugs on Github? If googles AIs were redirected from spying they could easily be doing that now. Who wants to go through every line of code on Github to see if anyone made a similar mistake to this issue that has been corrected?
or who wants to go through every user on this forum and guess / estimate if they are a spam bot?
My (proposed) bots would do that all day long, no complaints…
Also we need to rush before the spam bots get AI.
coin bot: I am the Coin Bot
You: Wellenreiter said he talked to you
coin bot: I know about Litecoin
You: That’s what he said you said
coin bot: There ain’t no such thing as a free lunch.
You: what about free software?
coin bot: Feathercoin bypasses the old banking system by using peer-to-peer technology.
You: what did he say?
coin bot: He said “frostfire: Sideways-Sanae was here yesterday”
You: Wellenreiter was here yesterday
coin bot: did you talk to him about Sideways-Sanae?
You: You talked to him about FTC
coin bot: I know about Litecoin
You: what do you know?
coin bot: We economists don’t know much, but we do know how to create a shortage.
-
Just a note on the current state of machine learning i.e. great if you are Google. Researchers have trained a Artificial Intelligence to patch security issues by cutting and pasting code from binaries. The binaries don’t have to be programmed in the same language, the machine learning extracts and uses the binary patches by learning from their actions.
Two points.
1. Being able to replace binery code on the fly negates all source code and current software security measures, i.e. if security measures can be added (real time) they can be taken out.
2. Machine learning will very soon be able to (i.e. already can) decompile all binaries software, to any language = end of proprietary software?
Update :
Adobe Machine Learning program - Computer program fixes old code faster than expert engineers
-
I see that the next step an Machine Learning taking over “management” level jobs has already happened at Pay Pal, with seller / buyer dispute resolution.
Companies like Moka are also resolving legal disputes “online”. The main rouble is this software is not open source and currently our lives are being run by Artificial Intelligence but we have no ability to monitor or decide how far that goes ie when a human should step in and over ride the decision.
A classic example is my local Tesco, a “monopoly” supermarket in my middle class area is for ever trying to reduce the amount of value items. I often have to complain ( disabled = poor), the Tesco “Algorithms” attempt to maximise profit, even when that means selling unhealthy food or removing value products or absolute basics, such as cottage cheese, from the shelf.
http://phys.org/news/2015-07-silicon-valley-company-court-disputes.html
-
Just looking at http://gazebosim.org/ , slight drawback in that it is not available for Ubuntu 15.04 yet, however, just loading Virtualbox to see if it will run in that.
Using Virtualbox is not as useful as having a proper robot / machine learning macro assistant as part of the desktop, so still waiting for Robotubuntu …
Why Gazebo?
Robot simulation is an essential tool in every roboticist’s toolbox. A well-designed simulator makes it possible to rapidly test algorithms, design robots, and perform regression testing using realistic scenarios. Gazebo offers the ability to accurately and efficiently simulate populations of robots in complex indoor and outdoor environments. At your fingertips is a robust physics engine, high-quality graphics, and convenient programmatic and graphical interfaces. Best of all, Gazebo is free with a vibrant community.
-
Get the benefit of ZFS style copy on write and File system snapshots, retro fitable to you Linux FS with Dattobd.
https://github.com/datto/dattobd
The Datto Block Driver (Dattobd) solves the above problems. Dattobd is an open source Linux kernel module for point-in-time live snapshotting.
Dattobd can be loaded onto a running Linux machine (without a reboot) and used to create an image file representing any block device at the instant the snapshot is taken.
After the first snapshot, Dattobd tracks incremental changes to the block device and can therefore efficiently update existing backups by copying only the blocks that have changed.
-
Two amazing programs are now available on Git hub, which are significant developments in the availability of open source machine learning to the “normal person”,
https://github.com/karpathy/char-rnn
char-rnn
This code implements multi-layer Recurrent Neural Network (RNN, LSTM, and GRU) for training/sampling from character-level language models. The model learns to predict the probability of the next character in a sequence. In other words, the input is a single text file and the model learns to generate text like it.
The context of this code base is described in detail in my blog post. The project page that has a few pointers to some datasets.
If you are new to Torch/Lua/Neural Nets, it might be helpful to know that this code is really just a slightly more fancy version of this 100-line gist that I wrote in Python/numpy. The code in this repo additionally allows for multiple layers, uses an LSTM instead of an RNN, has more supporting code for model checkpointing, and is of course much more efficient.
This code was originally based on Oxford University Machine Learning class practical 6, which is in turn based on learning to execute code from Wojciech Zaremba. Chunks of it were also developed in collaboration with my labmate Justin Johnson.
https://github.com/commonsense/conceptnet5/wiki
ConceptNet 5
ConceptNet aims to give computers access to common-sense knowledge, the kind of information that ordinary people know but usually leave unstated.
This Python package contains a toolset for loading new datasets into ConceptNet 5, and it serves the HTML and JSON Web APIs for it. You don’t need it to simply access ConceptNet 5; see http://conceptnet5.media.mit.edu for more information.
If you’re interested in using ConceptNet, please join the conceptnet-users Google group: http://groups.google.com/group/conceptnet-users?hl=en
Further documentation is available on the Wiki: https://github.com/commonsense/conceptnet5/wiki
ConceptNet is a multilingual knowledge base, representing words and phrases that people use and the common-sense relationships between them. The knowledge in ConceptNet is collected from a variety of resources, including crowd-sourced resources (such as Wiktionary and Open Mind Common Sense), games with a purpose (such as Verbosity and nadya.jp), and expert-created resources (such as WordNet and JMDict).
-
Interesting to see these projects have developed recently :
-
Interesting developments in combined Man Machine Intelligence?
https://sites.google.com/site/gbialternative1/home
First signs of “open forum” style public action tools becoming available?
-
This looks interesting : Although I initially thought it was for sniping eBay!
http://docs.snipeitapp.com/installation/downloading.html
Snipe-IT Installation Guide
Snipe-IT is a Free Open Source (FOSS) project for IT asset management, to enable IT departments to track who has which laptop, when it was purchased, which software licenses and accessories are available, and so on.
-
esniper is a simple, lightweight tool for sniping ebay auctions. It runs on all flavors of linux, unix and Windows, and should adapt easily to other operating systems.
http://esniper.sourceforge.net/
What is sniping?
Sniping is the act of bidding on an auction with just a few seconds remaining.
Why snipe?
Because it is better than bidding early. For more information on sniping, including reasons for and against.
What’s wrong with bidding early?
In a perfect world, nothing. The world isn’t perfect, though. Most eBay bids are proxy bids. If everybody understood that, figured out how much they wanted to pay, and bid their maximum, sniping would not be useful. However, many bidders have no idea how much they want to pay, so they bid in small increments until they beat your proxy. If you bid your maximum at the end of the auction, you greatly reduce the chances of being bid up by such a bidder.
Is sniping against eBay policy?
No. You may bid at any time in an eBay auction.
-
OnionShare lets you securely and anonymously share files of any size. It works by starting a web server, making it accessible as a Tor hidden service, and generating an un-guessable URL to access and download the files.
It doesn’t require setting up a server on the internet somewhere or using a third party filesharing service. You host the file on your own computer and use a Tor hidden service to make it temporarily accessible over the internet. The other user just needs to use Tor Browser to download the file from you.
Features :
A user-friendly drag-and-drop graphical user interface that works in Windows, Mac OS X, and Linux
Ability to share multiple files and folders at once
Support for multiple people downloading files at once
Automatically copies the unguessable URL to your clipboard
Shows you the progress of file transfers
When file is done transferring, automatically closes OnionShare to reduce the attack surface
Localized into several languages, and supports international unicode filenames
Designed to work in Tails, for high risk usershttps://github.com/micahflee/onionshare/blob/master/README.md
-
BTFS (bittorrent filesystem)
What is BTFS ?
With BTFS, you can mount any .torrent file or magnet link and then use it as any read-only directory in your file tree. The contents of the files will be downloaded on-demand as they are read by applications. Tools like ls, cat and cp works as expected. Applications like vlc and mplayer can also work without changes. -
Open Data Kit (ODK) is a free and open-source set of tools which help organizations author, field, and manage mobile data collection solutions. ODK provides an out-of-the-box solution for users to:
- Build a data collection form or survey (XLSForm is recommended for larger forms);
- Collect the data on a mobile device and send it to a server; and
- Aggregate the collected data on a server and extract it in useful formats.
In addition to socio-economic and health surveys with GPS locations and images, ODK is being used to create decision support for clinicians and for building multimedia-rich nature mapping tools. See the list available tools, featured deployments, and implementation companies for more examples of what the ODK community is doing.
-
Sharetribe is an open source platform to create your own peer-to-peer marketplace.
Read more about it on the Github page. Or, you can set up your marketplace in one minute without touching code - Head to Sharetribe.com.
https://github.com/sharetribe/sharetribe
http://sharetribe.freeforums.net -
@wrapper you really like discussion with yourself!
-
yes, me and 6.5k views…
