I basically agree with James about the patent process.

Software or algorithmic patents are not really very powerful, because there
are usually ways to work around them anyway, i.e. to make modifications that
are just barely enough to get around the patent, but still utilize the core
concept underlying the patent.  For this reason software or algorithmic
patents aren't likely to be enforceable in a useful way.

In a business context, they have a use, but only in an annoying way: because
investors like patents, and as a preventative against other companies making
patents that cover your ideas and forcing you to pay large sums of money to
defeat their bogus patents in court.

The patent system began as an attempt to protect the "little guy" from
inventing something and then having a big company steal it and sell it
without giving him any of the money.   It still serves this purpose in many
cases; but plenty of times it now serves as a way of letting big companies
blockade large areas of IP with unenforceable patents that "little guys"
can't afford to violate because they can't afford the legal fees it would
take to get them overturned.

In the case of my Novamente AI design, clearly there are dozens of
patentable innovations in the design, but do I have the $20K and month of
effort (estimated time, spread over a longer period of course) that it would
require to patent each one?  Clearly not.

On the other hand, we *are* patenting some AI algorithms we've devised in
the context of Biomind, our AI/bioinformatics company, because patents are
considered particularly valuable in the biopharma industry.  Frankly, this
is not so much because we crave patent protection, it's mainly because if
Biomind should ever be acquired or go public, these patents will be viewed
(perhaps wrongly, in my personal view) as valuable IP by potential
purchasers of our company or its stock.

-- Ben Goertzel



-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Behalf Of J.Andrew Rogers
Sent: Wednesday, September 08, 2004 12:49 AM
To: [EMAIL PROTECTED]
Subject: Re: [agi] AGI road map revised version



On Sep 6, 2004, at 11:07 AM, Yan King Yin wrote:
> I'm currently in the process of filing a patent of my model.
> I think licences can be obtained fairly easy nowadays because
> the information economy is getting very efficient and so
> innovation is unlikely to be impeded by IP issues. Maybe patents
> will actually facilitate progress if more people can learn to
> utilize them.


You generally can't patent an abstract model, and even if you manage,
such patents generally aren't enforceable -- it will collapse the first
time someone with something more substantial challenges it.  There is
too much vagueness as to what you are trying to patent.  If you don't
have novel algorithms, explicit mechanisms with quantifiable results,
you don't have anything.  The old "you can't patent ideas" saw applies
here.  There are a lot of "abstract AI model" patents that have been
granted by attempting some type of specificity, but since there is no
specificity in practice (i.e. they've never been successfully
implemented), they are something of a legal joke in terms of standing
and it is plausible that an implementation of the abstract idea would
not be subject to the patent in practice.

I don't object to patents per se, but I do object the squandering of
resources on useless patent efforts.  It isn't cheap of time or money,
and for a patent area as complex as AGI, you'll need to be pretty well
funded to get useful core coverage.  My primary objection to useful
patents from my perspective is that the application process is
massively time consuming for something complex and abstract i.e. when
it is something other than an easily describable widget or process that
you can hand to a patent authoring consultant.  I personally find it
insanely tedious, as the patents I've had to author (and I've squashed
many patent attempts on things I didn't consider worth patenting)
generally required extensive involvement on my part, as in effort
measured in man-weeks.  Ugh.

Pick your IP battles carefully.


j. andrew rogers

-------
To unsubscribe, change your address, or temporarily deactivate your
subscription,
please go to http://v2.listbox.com/member/[EMAIL PROTECTED]



-------
To unsubscribe, change your address, or temporarily deactivate your subscription, 
please go to http://v2.listbox.com/member/[EMAIL PROTECTED]

Reply via email to