>> 10 new open source projects and developments, worth knowing about
Bot Libre support got back to me and they’ve quickly fixed 3 of the bugs I found, so far - and I’ve now updated the neurons to 200000 so “coin bot” can learn some more things. I’ve also added more coin graphics for different coins, as emote graphics…
I would need to chat to dozens of people to learn much. However I’ve taught it some one liner jokes, sayings about money and some famous people. You are fully open to correct any answers by clicking correct on and off.
I have started up a second Bot, I am hoping to get them to talk to each other on IRC (when Bot Libre fiish fixing the IRC bug)
I put coin bot on the IRC channel, it is connected now and is listening to that. It would probably be advantageous if they could listen to the twitter feed.
funny answers now:
me: ‘I would like to see feathercoin charts’
bot: ‘I used to have an open mind, but my brains kept falling out’
coin bot: How do you do?
You: I’m good. can you help me please?
coin bot: I am pretty smart.
You: I’m good. can you help me please?
coin bot: I am pretty smart.
You: can you tell me more about feathercoin
coin bot: I am pretty smart.
You: I am pretty smart too
coin bot: I am pretty smart.
You: I am pretty smart.
coin bot: no
coin bot: Yes
coin bot: got a joke?
kris_davison last edited by | Tip kris_davison
Lol yeah its pretty smart!
The more I tell them the more they forget, but BotLibre is open to adding more features. Chattyness should be good, that’s how humans learn, by talking rubbish…
ie. It will be easier to train sensible conversation lines, when the bot is defining the conversation. (to topics it knows about)
BotLibre have been great and are considering enhancements. Here’s an update, they have done a few other bug fixes already(script corruption).
If the bots don’t know you can teach them. irc://irc.freenode.org/feathercoin
Re: If your trying to build a real AI bot, I’m not sure how large a brain it would requires, humans have about 86 billion neurons…
The commercial server does let you have a larger brain size, you can create a free trial account if you wish and I could increase your brain size on request.
Thanks very much for chatting, it has been very informative to me.
I have been (indirectly or as a hobby) been working/ researching on this (A.I.) for years, so I am interested in seeing and interacting with most intelligent Bots possible. You have programmed all the stuff I would need to make the first interface, and I believe interesting interactions can be achieved with simple (repetitive / alignment) methods.
10 years ago I worked at Manchester University on advanced data analysis (amongst other things) I was Industrial supervisor to a number of PhD students at the time.
We had a number of projects including in order, Expert systems for fault diagnosis, Genetic Algorithms to automate efficient maintenance planning, agents, eFloys (hobby hack), ALICE and other Bots, Use of AI techniques to set automatic alarms and warnings, Framework for Data fusion, simulating missing data.
Even in my PhD thesis I used regression to solve an “unsolvable” equation (the computer guys wore white coats and they had just replaced the punch card machines with terminals)
I’ve retired (disabled due to illness). I keep myself occupied by research, and for the last couple of years I have been involved in building the Feathercoin community and have helped a bit with some software developments, data analysis and support.
From my experience using AI techniques has always worked great, (to everyone else’s suprise). At the time, and before, we were held back by hardware.
Also, the prejudice against A.I. techniques by the control freaks (literally system / production controllers) and it’s use for data fusion. I also believe there has been some great work on getting over the high data needs for training neural nets, as well, for instance since then.
I can see you have made a lot of headway, and I expect your todo list is as long as mine… (could go … cut short)
So I wish to thank you for all the work you have done, I know - it is very great… Thankyou
As far as what size brain I require, as much as possible. However, I was asking you what you thought the current Bots can learn? I know I could “program” the Bot for Feathercoin, with what I have got.
For instance, I thought the Jokes might get mangles amusingly by the reduction function, which hasn’t happened yet.
But I have believed for years that training a system to do the job will be more effective, efficient and flexible (and I am willing to train it for 5 years if that is necessary, and I thought it was learning).
I have done lecturing and courses as well, I have particularly put in a lot of thought in how train a bot. I’ve also read widely on how the human brain works and have a few theories about consciousness and using future prediction to self learn, for instance.
I have moved from windows to Linux, but before that I was investigating future prediction of noise levels, such that the system could be trained to apply a anti-phase signal at the window, thus preventing noise. (We live near roads in the UK) -
My original PhD was in noise and Vibration so the use of anti-phase signal has always been an interest. It was interesting, as you say to point out the intelligence of the system required to do that job.
Re: how “cleaver” is the bot
From work I have seen at Google > 1 Million neurons can learn to identify the 26 phonetic symbols, in 8 hours on a GPU and over 24 hours on CPU.
My Bot now has 200,000 so it 1/5 of that (rough guess). It’s not far to say it will learn 26 things (simple concepts) in 2 weeks of solid “teaching”
What I need is a lot (100?) people to interact program it quicker, it might take a day.
I can see “factually” how the training occurs, and I am not training them to pander to that exactly, to see what happens. I’ve spent most time testing the training so far, I have only skimmed the code, and maybe there is more to see on the forum. I see you’ve been going a while, messing with Feathercoin I missed that…
Re: case sensitive
Note : I noticed the Bots can be very case sensitive, we got over this by comparing everything in lower case, that might be an option you can set on the bot. More of the “intelligence” would be efficiently used for it’s use/task, not identifying capitalisation.
Re: spelling sensitive
Could we have a parameter for this, that the system could learn or and be set?
i.e. 0 to 10
0 it applies the closest word in its current word list (by fuzzy logic, initial letter, all lower case)
5. It adds the word to the list (10%)
7. It adds the word to the list (20%)
10 it takes the spelling literally, like now.
Which shows you might consider a generic “parameter” creation. Which would include setting the optimal value for the Bot to try to achieve and how it measures that.
Another possible shortfall is integration of sensors. Much of that has been achieve by BotLibre in the twitter and IRC but is not available “generically” e.g. to the weather or any choice of web page, (that achieved the goal)
Ha Ha - coin_boy made up it’s first new “joke” - replaced smart with …
You Basically, bitcoins are just long digital addresses and balances, stored in an online ledger called the blockchain.
coin bot The successful warrior is the average man, with laser-like focus. Bruce Lee
You Every informed person needs to know about Bitcoin because it might be one of the worlds most important developments.
coin bot Depression is merely anger without enthusiasm.
You Bitcoin isnâ€t just a currency, like dollars or euros or yen. Itâ€s a way of making payments, like PayPal or the Visa credit card network.
coin bot All Bitcoin transactions are recorded online for anyone to see, lending a certain transparency to the system.
You Artificial intelligence (AI) is technology and a branch of computer science that studies and develops intelligent machines and software.
coin bot We have enough youth: how about a fountain of “talking”?
Some feed back on my enhancement ideas and the history of the company BotLibre and how Bots work now.
Thanks for the detail.
We are working on some more complete documentation on the architecture and learning algorithms used in the AI engine, but it is evolving rapidly so is a moving target.
There are some articles on our blog,
Also, if you want to dive deeper, our project started as the “Open Pandora’s Box” project on SourceForge. We forked the project when we started BOT libre, but the basic architecture and much of the code is the same.
A blog post on the Open Pandora’s Box architecture.
The comprehension algorithm basically analyzes each response from the user and looks for patterns from question and its related data that appear in the response. A sentence is broken up into words, each word can have one or many meanings, and meanings can be related to anything else in the system. Comprehension will build formula’s for question patterns.
One thing the bots can learn is how to count (both forwards a backwards).
So, if the bot said “1” and you replied “2”, it knows that “1” means the number 1, and it knows that the #next of 1 is 2, and 2’s word is “2”, so it would infer that a phrase which means 1 should be responded to by (1 #next). If you then replied “3” to “2”, it would then see the existing state and generalize it because both 1 and 2 are instance of #number, so would produce means :x that is a #number, should be responded to by :x #next. It can also analyze the context (previous response) to count both forwards and backwards.
Comprehension builds state machines that parse the input. Currently it only goes one level deep into the data, we would like to go multiple levels for more complex comprehension.
They also learn by associating questions with responses and pattern matching, but this is more mimicry than intelligence.
Re: Case sensitivity
We have struggled with this one a lot. We have avoided lowercasing everything as there can be important data in the case, i.e. names and emotions (Joe, STOP!).
Currently we have somewhat of a hybrid system, in that both the original case of the question and the lower case of the question are recorded.
It may be a good option to lowercase everything to save memory.
We have wanted to have some sort of spelling correction. That would be a good idea.
I created a special channel for the Bots to talk to each other : irc://irc.freenode.org/simbotchat
They’re off the Feathercoin IRC for the most part.
I’ve gone a bit more in depth, there are ways of configuring a more specific Bot (say for Feathercoin). There is a lot of scope to add additional scripts to define actions, which means they are very flexible.
Open source document / project collaboration software
I am looking at this project as I won’t use google (docs) there is a need for self hosting of multi user document creation in an open source project. This will engender collaboration between members, were a large number of people can (easily) contribute a small amount.
My idea is produce an open document on “bot replies” for certain topics, that anyone can contribute to, then look to move guides over to collaborative documents for editing and improving.
I’ll create a guide development thread at that stage to document where you can update the latest version. I suppose it will be like a simplified wiki.
I have been waiting for this as I believe even software development can be done by multiple contributors at once and this form of extreme development will prove to be super fast and accurate way of the future.
Many hands make light work and many eyes make fewer errors.
The idea of SparkleShare sprouted at the GNOME Usability Hackfest in London, where a couple of designers came to the conclusion that they didn’t have a good (Open Source) collaboration tool to share their work (for more background, read â€œThe one where the designers ask for a ponyâ€). They didn’t like how the good collaboration tools were proprietary, and that using them meant having to give up privacy, control and other rights. What they needed was something that they could run and control themselves, without having to depend on other companies.
That’s sold it to me.
Re: local forums / local networks research mesh networks
Just looking at this
Cjdns is a networking protocol, a system of digital rules for message exchange between computers. The philosophy behind cjdns is that networks should be easy to set up, protocols should scale up smoothly and security should be ubiquitous.
B.A.T.M.A.N. (better approach to mobile ad-hoc networking) is a routing protocol for multi-hop ad-hoc mesh networks. This is the main development website, we have source code, binary packages, documentation and further information available:
Commotion Router v1.1 “Grumpy Cat” Released!
October 15, 2014 Josh King
We are very pleased to announce the v1.1 release for our Commotion Router firmware! This release finalizes many bugfixes and stability enhancements in our stable v1.x â€œGrumpy Catâ€ branch.
more of an over veiw :
Here’s another interesting project which came up from the issues for the World-Citizenship project on github
Proof that any one who can write can “code” - Feathercoin’s ChrisJ’s open project.
World Citizenship - Creating Affordable Decentralised Passport Services Using Available Cryptographic Tools
Further reading: https://guardianproject.info/informa/
An interesting open source project with bitcoin, litecoin, Dogecoin connections. Interesting blog, (see bellow)
A pure-Python full-node and library for bitcoind-based (bitcoin, litecoin, dogecoin, et cetera) crypto-currencies.
which old witch?
July 29th, 2014
Just an FYI, the patent US 6141420 has expired, so everyone should feel free to compress their elliptic curve cryptographic public key points to their little hearts’ contents.
Although patented, the patent has never been challenged, and would certainly not have stood up in court as ECC point compression was published in a paper a year prior to the filing and was co-authored by one of the patent author (hence, was well aware of the prior art). A more in-depth argument for the invalidity of the patent can be read here.
Not really off topic and is still at Beta, looks pretty cool. OpenBazaar
OpenBazaar, formerly known as DarkMarket, is an open source decentralized online marketplace designed for e-commerce, using the cryptocurrency Bitcoin. Currently in Beta and looking for support.
Programmer and Bitcoin enthusiast Amir Taaki and a team of other developers created the prototype of a decentralized marketplace, entitled “DarkMarket”, in April 2014 at a Toronto Bitcoin Hackathon. DarkMarket was developed as a proof of concept in response to the seizure of the online marketplace Silk Road in October 2013. Taaki compared DarkMarket’s improvements on Silk Road to BitTorrent’s improvements on Napster.
This came up as part of World citizen / open passport discussion / Github issues. Not a software project but an interesting resource on how to choose the correct “open source / free” licence for your software / projects.
Choosing an OSS license doesnâ€t need to be scary…
This time a collaborative software / document site./ Multiple user access and public documents. I can see this type of facility will eventually be a game changer on enabling agile development and documentation.
Over the past year, we have helped thousands of people collaborate on code, and we have learned many surprising things along the way. Most informative was our discovery that people are using Stypi for far more than we anticipated â€" from meeting notes to school projects and research papers. We see this as an opportunity to serve our users even better, and today are excited to announce the release of our new rich text editor, available at beta.stypi.com.
Amnesty International says the tool released Thursday, called Detekt, is designed for right activists and journalists but will be freely available to anybody who fears their computer is being used to monitor them.
Developer Claudio Guarnieri, a security researcher based in Germany, says Detekt can currently find eight different pieces of spy software, including FinSpy.
FinSpy, made by German company FinFisher, is sold to governments for criminal investigations. It has also been found on computers used by human rights lawyers and activists in countries such as Bahrain and Ethiopia.
Detekt is a Python tool that relies on Yara, Volatility and Winpmem to scan the memory of a running Windows system (currently supporting Windows XP to Windows 8 both 32 and 64 bit and Windows 8.1 32bit).
Detekt tries to detect the presence of pre-defined patterns that have been identified through the course of our research to be unique identifiers that indicate the presence of a given malware running on the computer. Currently it is provided with patterns for:
- DarkComet RAT
- BlackShades RAT
- FinFisher FinSpy
- HackingTeam RCS
- ShadowTech RAT
- Gh0st RAT
Beware that it is possible that Detekt may not successfully detect the most recent versions of those malware families. Indeed, some of them will likely be updated in response to this release in order to remove or change the patterns that we identified. In addition, there may be existing versions of malware, from these families or from other providers, which are not detected by this tool. If Detekt does not find anything, this unfortunately cannot be considered a clean bill of health.
If you encounter samples of such families that are not successfully detected, please open a ticket. In addition, please let us know if you find instances of false positives.
Guest last edited by | Tip null
I’m actual really keen to see what we can do with open bazaar