Free software company devoted to AI/Math

Posted 26 Dec 2000 at 13:56 UTC by CarloK Share This

The only one I know is Spinninglogic, but it's just born...

I am wondering about European or American companies who develop free software tools for AI or mathematical modelling.
Do you know some?

Why is this here?, posted 26 Dec 2000 at 16:10 UTC by sneakums » (Journeyer)

Isn't this kind of query what IRC and specialist mailing lists are for?

your answer is too easy, posted 26 Dec 2000 at 16:31 UTC by CarloK » (Journeyer)

If you know one, why don't you write here a name instead of a general answer?

It's not an answer, posted 27 Dec 2000 at 09:33 UTC by sneakums » (Journeyer)

My point was that Advogato articles are intended for thought- and comment-provoking discussions of a topic, not to save its users some legwork and time using search engines.

Just do it, posted 27 Dec 2000 at 13:22 UTC by CarloK » (Journeyer)

I know your point.
Trust me... I wasn't able to find any companies like that.
I used search engines and so on.
Advogato is a free software community so I think it *could* be a good place to ask for and, most of all, I tought my article contains an implicit question:
"There are a lot of free software/open source companies in a lot of fields... do you know some in AI? I didn't find one. Is there a special reason?"

What area of AI/math?, posted 28 Dec 2000 at 12:53 UTC by Rhys » (Journeyer)

I take your point about it being tricky to find out about stuff like this on IRC/Google. The fact is that this sort of topic is usually dealt with in academic environments, and usually (but not always) academia and open-source don't mix and match much (PostgreSQL being an example to the contrary; for others, see below).

I know a bit about speech recognition, and if pattern-matching is the sort of AI stuff you're interested in, have a look at Cambridge University's excellent HTK, a suite of ANSI C programs designed for speech recognition algorithms, but with applications elsewhere in AI. The history of this is that it was originally developed by Cambridge's Engineering department, spun off to an outside company with the same personnel, said company was bought by Microsoft last year, and

Without going into a deep discussion about their rather closed license agreement (somewhere to the right of Mozilla's original one, but it is free for download) HTK is top stuff, very robust, and does actually work very well. You're likely to find it referenced in at least half of the present-day speech research papers out there, and an enhanced (Cambridge in-house) version of HTK has won NIST's Broadcast News speech recognition competition, against competition from some of the other giants in the field.

To go on a slight tangent here - on the subject of speech science, there was an excellent article in one of the UK Linux magazines this month (I forget which) that documented SuSE's efforts to build a Linux that could be installed by the blind, using a Braille reader as an output terminal, and a rather fantastic open-source speech synthesiser (CSTR's Festival) as a screen reader. Festival speaks Welsh, so I like it :)

Mail me off-site if you want any more details about the above. Have fun.


More details..., posted 28 Dec 2000 at 13:53 UTC by CarloK » (Journeyer)

Your two homepages are unreachable so I can't email you directly. I post here some comments.

> if pattern-matching is the sort of AI stuff you're interested in...

My MSc Comp. Eng. Thesis was about the developing of a new algorithm (a variant of adaboost) designed to learning classifiers (in first order logic) by integrating boosting and evolutionary search. It is a new research direction that gave us good results (this research is done by the Machine Learning Group of Universita' di Torino). So the software you described could have something similar.

> there was an excellent article in one of the UK Linux magazines...

This is very good 'cause I suppose it will be free software. I found a bit strange that there is a lot of big and important project in the free software arena (apache, gnome, gnu/linux and so on...) but there are not BIG scientific projects with the same spirit: I mean where developers are wide spread, the project is not lead by any universities...
Maybe we should get inspired by seti@home or something to start up a world wide scientific free software project? Science is full or BIG projects are waiting to be done by a huge number of hackers.

Thank you for your time.

Needs and gaps in AI/free software, posted 3 Jan 2001 at 10:46 UTC by Rhys » (Journeyer)

The relevant router has been rebooted; my email and homepages now work again. That was beyond my control, unfortunately.

You implied that AI was a relatively spartan area for free software. I'd tend to disagree, but it all depends on looking in the more unusual places. Much, in fact most, AI research in UK universities is supported by the relevant government-funded research council. And in most cases, it is a condition of such public funding that your research is, well, public. Company confidentiality is increasingly an issue nowadays, though, and where industry and academia overlap in such a way, you may not have quite as much freedom to publish all your results. But still, at the end of the day, you have to produce reports and/or a thesis, which increasingly these days contain at least some algorithmic information. So this is free for others to use, provided a company hasn't bought your brains and patented the algorithms first though.

It may well be that the relatively esoteric (but not impenetrable) maths behind AI is a bar to many people entering the field with a free software mindset. After all, most free software is still written in C, rather than C++ or assembler, primarily because C is still considered to be the lingua franca of programming. You could I guess construct a similar argument about constructing a free software database system as opposed to (say) a Gaussian Mixture Model-based algorithm for robust speaker recognition. More people feel able to do the former. I think, anyway.

Your comment that a large scientific free software project should be developed is valid, but I'd shy away from trying to plug a gap simply because one exists. You may find, unfortunately, that others have tried and failed for the very reasons above. But that's veering into recent Advogato territory about the critical mass of projects, and I wish you all the best if you do decide to venture out.

After all, we're all supposed to have sentient computers this year :)


New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!

Share this page