Computer Ethics, Fall 2012
Corboy Law 423; Thursdays, 4:15-6:45
Week 10
A few highlights from Sherry Turkle's keynote address at Loyola's Digital Ethics conference last Monday:
1. A member of a group of students who had been regularly texting in class made the following confession:
"I am not as strong as technology's pull"
2. Mark Zuckerberg: "Privacy is no longer a relevant social norm"
3. Another student stated that the way to deal with the loss of privacy is "just to be good"
4. Turkle's assumption for healthy democracies: "everyone has something to hide"
5. A sixteen-year-old girl interviewed by Turkle stated, regarding privacy, "who would care about me and my little life"
Read Baase 4e Chapter 3 on Speech
Read Baase 4e Chapter 4 section 5 which includes patents
Some patent papers
These are also assigned reading.
1. Simpson
Garfinkel, Patently Absurd, 1993
Garfinkel's article is pretty easy reading, pointing out some problems
with software patents specifically. Garfinkel wrote his article 20 years ago; has any progress been made?
2.Richard
Stallman on Patents, 2002
Stallman is against software patents, of course. However, his case here
is
better than many open-source-related arguments; in fact, it is squarely
aligned with the interests of software-development businesses.
3.Paul Graham, a
computer scientist and one of the partners of the venture-capital firm
Y Combinator, wrote a 2006 essay Are Software
Patents Evil?
Graham makes the following claim early on:
One thing I do
feel pretty certain of is that if you're against
software patents, you're against patents in general. Gradually our
machines consist more and more of software. Things that used to
be done with levers and cams and gears are now done with loops and
trees and closures. There's nothing special about physical embodiments
of control systems that should make them patentable, and the software
equivalent not.
Is this true?
Does it matter that Graham is also a radical proponent of using the
lisp programming language, which everybody else stopped using in the
1990's?
Graham also says,
Frankly, it
surprises me how small a role patents play in the
software business. It's kind of ironic, considering all the dire
things experts say about software patents stifling innovation, but
when one looks closely at the software business, the most striking
thing is how little patents seem to matter.
But that paragraph is about software companies being sued by other
software companies, and not "patent trolls".
Graham also makes some other claims, in particular some about the role
of the patent system in business competition generally. Check out what
he says about Reveal.
Is source code speech?
Well, is it?
Cases where it's been debated:
- encryption
- DMCA anti-circumvention (eg deCSS)
The DCMA also has a speech restriction:
(1) No person shall manufacture, import, offer to the public, provide,
or otherwise traffic in any technology, product, service, device,
component, or part thereof, that— (A) is primarily designed or produced
for the purpose of circumventing protection afforded by a technological [copy-protection measure]
If you write an online or print article about how to bypass copy-protection, you may be violating this.
DeCSS case
There are several; the best known is Universal
Studios
v Reimerdes, Corley, and Kazan.
Eric Corley, aka Emmanuel Goldstein, is the publisher of 2600 magazine. In 2000 the magazine included an article about a
new program, "deCSS", that removed the CSS ("content-scrambling
system") encryption from DVDs, thus allowing them to be copied to hard
disks, converted to new formats, and played on linux systems.
Corley:
DeCSS was developed in ~1999, supposedly by Jon Lech Johansen. He wrote
it with others; it was released in 1999 when Johansen was ~16. He was
tried in Norway in 2002, and was acquitted.
Cute story about Jon: In 2005, supposedly Sony stole some of his GPL-covered code for their XCP "rootkit" project. Jon
might have been able to sue for huge damages (though the usual
RIAA-lawsuit standard is based on statutory damages per item copied,
and here only one thing was copied). More at http://news.slashdot.org/story/05/11/17/1350209/dvd-jons-code-in-sony-rootkit
Judge Kaplan memorandum, Feb 2000, in Universal v Reimerdes:
As a preliminary matter, it is far from clear that DeCSS
is speech protected by the First Amendment. In material
respects, it is merely a set of instructions that controls
computers.
He then goes on to consider the "balancing" approach between free speech
and regulation, considering the rationale for the regulation and the
relative weights of each side.
The computer code at issue in this case does little to serve
these goals [of expressiveness]. Although this Court has assumed
that DeCSS has at least some
expressive content, the expressive
aspect appears to be minimal when compared to its functional component.
Computer code primarily is a set of instructions which, when read
by the computer, cause it to function in a particular way, in this
case, to render intelligible a data file on a DVD. It arguably
"is best treated as a virtual machine . . . ."
[the decision cites
Lemley & Volokh, Freedom of Speech
and Injunctions in Intellectual Property Cases, Duke Law Journal
1998. However, the sentence in Lemley and Volokh's paper explicitly
refers to executable object
code, not source! "The Bernstein court's conclusion, even if upheld,
probably doesn't extend past source code to object code, however. We think most executable software
is best treated as a virtual machine rather than as protected
expression." Judge Kaplan apparently did not grasp the distinction.]
Note that this virtual-machine argument renders irrelevant the Bernstein
precedent! Actually, the virtual-machine argument pretty much presupposes that you have come down
solidly on the side of code-as-function instead of code-as-expression.
Also note the weighing of expression versus functionality, with the
former found wanting.
Do you think that Judge Kaplan was stricter here than in the crypto
cases because crypto was seen as more "legitimate", and deCSS was
clearly intended to bypass anticircumvention measures?
The district court issued a preliminary injunction banning 2600.com
from hosting deCSS; the site then simply included links to other sites
carrying it. The final injunction also banned linking to such sites. Furthermore, the decision
included language that equated linking with trafficking.
Universal v Reimerdes, Appellate Court
The Appellate decision was similar to Judge Kaplan's District Court opinion, though with somewhat more on the
constitutional issues, and an additional twist on linking. Also, note
that one of Corley's defenses was that he was a journalist, and
Writing about DeCSS without including
the DeCSS code would have been, to Corley, "analogous to printing a
story about a picture and not printing the picture."
However, in full context, that idea was harder to support. Corley's mistake was in describing DeCSS as a way to get free movies. What if he had
stuck to the just-the-facts approach, and described exactly how easy it
was to copy DVDs without actually urging you to do it? Is this similar
to the "Grokster" workaround?
Both the DC and Appellate courts held that the DMCA targets only the
"functional component" of computer speech.
One argument was that the CSS encryption makes Fair Use impossible,
and that therefore the relevant section of the DMCA should be struck
down. The appellate court, however, ruled instead that "Subsection
1201(c)(1) ensures that the DMCA is not read to prohibit the 'fair use'
of information just because that
information was obtained in a manner made illegal by the DMCA". Subsection 1201(c)(1) reads
(c) Other Rights, Etc.,
Not Affected. — (1)
Nothing in this section shall affect rights, remedies, limitations, or
defenses
to copyright infringement, including fair use, under this title.
This is an interesting argument by the court! Literally it is correct, but the practical problems with Fair Use access go unaddressed.
Some notes on the free-speech argument:
Communication does not lose
constitutional protection as "speech" simply because it is expressed in
the language of computer code. Mathematical formulae and musical scores
are written in "code," i.e., symbolic notations not comprehensible to
the uninitiated, and yet both are covered by the First Amendment.
The court also acknowledged Junger v Daley (above).
However:
As the District Court recognized, the
scope of protection for speech generally depends on whether the
restriction is imposed because of the content
of the speech. Content-based restrictions are permissible only if they
serve compelling state interests and do so by the least restrictive
means available.
A content-neutral restriction is
permissible
if it serves a substantial governmental interest, the interest is
unrelated to the suppression of free expression, and the regulation is
narrowly tailored, which "in this context requires . . . that the means
chosen do not 'burden substantially more speech than is necessary to
further the government's legitimate interests.'"
That is, the DeCSS code may be said to be "expressive speech", but
it's not being banned because of what it expresses.
Unlike a blueprint or a recipe, which
cannot yield any functional result without human comprehension of its
content, human decision-making, and human action, computer code can
instantly cause a computer to accomplish tasks.... These
realities of what code is and what its normal functions are require a
First Amendment analysis that treats code as combining nonspeech and
speech elements, i.e., functional and expressive elements.
As for hyperlinks (in the section "Linking"),
a hyperlink has both a speech and a
nonspeech component. It conveys information, the Internet address of
the linked web page, and has the functional capacity to bring the
content of the linked web page to the user's computer screen.... The
linking prohibition is justified solely by the functional capability of
the hyperlink.
What if one simply printed the site name, without
the link: eg cs.luc.edu? For links, one can argue that the expressive
and functional elements -- what the other site is, and how to get there
-- are inseparable.
The non-linking rule may become more of an issue as time goes on and the US attempts
to remove from the DNS system sites which provide illegal access to
copyrighted material. In the future, identifying a new IP address for,
say, the now-seized megaupload.com may be suspicious.
Gallery of DeCSS: http://www.cs.cmu.edu/~dst/DeCSS/Gallery
Check out these in particular:
- haiku page
- page1.gif
- mathematical description
- DVD logo
Does the entire gallery serve to establish an expressive purpose?
If you want to play DVDs under linux, I recommend VLC Media Player at
http://videolan.org/vlc (which
generally needs libdvdcss).
Sita Sings the Blues: moved to Week 4
Patents
Baase 4e §4.5
Do patents help advance progress? or hinder it?
Patents are pretty clearly a market
regulation
with the sole goal of improving innovation in technology. Inventors
don't "deserve" to profit from their ideas; we simply want to make sure
they are motivated to continue. In other words, this is purely
utilitarian.
Patents are intended to cover INVENTIONS rather than IDEAS.
If you have an idea to sell hamburgers with salsa,
or newspapers & beer together, or to create a website
where people can post their own stuff, that's an IDEA.
It can't be protected: everyone else is entitled to
copy it freely.
If you decide that a certain vertical market can use an XML-based word processor, is that an invention or is that a business idea?
- patents cover the idea, not the expression
- the time period is more limited
What do these have to do with computers? There are several issues about
whether the patent system for software
in fact helps anyone, and whether software patents fulfill their
constitutional mandate to benefit society as a whole. These relate to
the situation where patents are dealt with only by serious developers.
Currently in the US, patents last 20 years, and generally cannot be
renewed or extended (but see below regarding pharmaceutical patents).
During that period, the patent owner can enforce their patent, but
patent litigation is relatively expensive. Formerly US patents lasted
17 years from the date of granting,
but to comply with the WTO treaty on Trade-Related Aspects of
Intellectual Property Rights (TRIPS) the term was changed to 20 years
from application.
There is also an analogue to the copyright
situation, where individuals
can download software that may violate patents in some countries. This
creates a situation somewhat related to file-sharing: end-users make
the decision. Unlike file-sharing, there is no group analogous to the
RIAA that is going after infringers.
Look at ubuntu software installation? What about MP3 players?
35 U.S.C. §101 (patent-eligibility law):
Whoever invents or discovers any new and useful process, machine,
manufacture, or composition of matter, or any new and useful
improvement thereof, may obtain a patent therefor, subject to the
conditions and requirements of this title.
The meaning of "process" is critical here: does it mean any procedure or method? Or does it
mean "industrial process"? Historically, it pretty clearly was intended
to mean the latter.
Pharmaceutical patents
Pharmaceutical patents are sort of the poster child for Why Patents
Are Good For Us. Here the patent system IS effective at encouraging
investment:
- Development costs are HUGE (~200 million) (maybe 100 million for a
small, straightforward trial, but some trials need LOTS of patients,
and/or expensive tests).
- Copying is not much harder than for software
One weirdness: patents "for the use of";
someone can, if drug X is in the public domain, patent
the use of X to treat disease Y (this must be in some legal sense a
"new" use of X).
In practice this is not
much of a problem, because generic manufacturers can still make and
market X for its old purpose, and doctors can prescribe it for its new
purpose. Such prescriptions are sometimes said to be "off-label"; they
are an important way for drugs to get to people who will probably be
helped by them, but for which no company has yet done clinical trials,
and never will.
Some specific drugs:
cancer monoclonal-antibody drugs
These protein compoundss are antibodies that are highly specific to a
certain substrate; they have very narrow targets. From http://en.wikipedia.org/wiki/History_of_cancer_chemotherapy:
Another branch in targeted therapy is the
increasing use of monoclonal antibodies
in cancer therapy. Although monoclonal antibodies (immune proteins
which can be selected to precisely bind to almost any target) have been
around for decades, they were derived from mice and did not function
particularly well when administered to humans, causing allergic
reactions and being rapidly removed from circulation. "Humanization"
of these antibodies (genetically transforming them to be as similar to
a human antibody as possible) has allowed the creation of a new family
of highly effective humanized monoclonal antibodies. Rituximab,
a drug used to treat lymphomas, is a prime example. -- Wikipedia
The
point is
that some cancers can be specifically targeted by certain antibodies,
because they have specific antibody receptors not present in
non-cancerous cells. The receptors involved tend to be very
idiosyncratic.
None of
these drugs would exist in the US marketplace if it were not for
pharmaceutical patents.
On the other hand, the US Food and Drug Administration, which
regulates new drugs, is arguably a massive government intrusion into
the free market. Why shouldn't patent law intrude as well?
imatinib/gleevec: leukemias,
stomach cancers. It is used to treat cancers where the cells involved
have a specific receptor. Time magazine called it the "magic bullet
against cancer" in 2001, when it was approved, though that was an
overbroad assessment. It was also the subject of a patent lawsuit in
India in 2007; the case was referred by the Madras High Court to the
WTO.
It is the first member of a new class of
agents that act by inhibiting particular
tyrosine kinase enzymes,
instead of non-specifically inhibiting rapidly
dividing cells. -- Wikipedia
rituximab/rituxan: binds to
the WBC surface protein CD20. Used to treat leukemias/lymphomas, also
some autoimmune diseases such as lupus
cetuximab/erbitux: metastatic
colorectal cancer, head&neck cancers. Binds to EGFR receptor; it is
an EGFR-inhibitor.
trastuzumab/herceptin:
breast cancer. Monoclonal antibody that interferes with HER2/neu
receptor. In some breast cancers, the HER2 receptor is, as wikipedia
put it, "stuck in the 'on' position".
Antibiotics
nobody makes these, actually. The FDA requires not just proof of
effectiveness, but proof that the infection that was cured was in fact
resistant to existing antibiotics. This makes clinical trials very expensive.
Protein pump inhibitors
(PPIs): used for various stomach-acid problems, including ulcers
omeprazole/prilosec
lansoprazole/prevacid
esomeprazol/nexium
HIV: protease inhibitors moved HIV from a short-term acute illness
to a long-term chronic illness. This made drug development profitable
again. When AZT
(one of the first effective anti-HIV drugs) was first applied in the
early 1990's to HIV patients, though, it was an off-label use.
red/white-blood cell drugs:
filGRAStim/neupogen
makes more neutrophils/other WBCs. Used for cancer/chemo/BMT patients
erythropoietin
makes more RBCs: kidney disease, cancer, cancer treatment
diabetes
exENatide/byetta: this often means the patient can
avoid taking insulin. It is offcially for type-2 (non-insulin-dependent) diabetes. From byetta.com: Byetta
may also be used for other purposes not listed in this medication guide.
In 1984, Congress passed the Drug
Price Competition and Patent Term Restoration Act, also known as
the Hatch-Waxman Act. This allowed generic drug makers to use a patented
drug in their own
FDA application, so that approval would be in place as soon as the
original patent expired (normally after 20 years). However, it also
gave pharmaceutical developers a chance at a patent extension for up to five extra years,
subject to the following:
- The extension application must follow the date of FDA approval
(which of course is after all the clinical trials are complete).
- The maximum length of extension is five years additional from
the
time of patent filing, or fourteen years from the date of FDA approval
(whichever is less).
The idea here is that if it took a company fifteen years to bring a
drug from discovery (and patent) to market, they will get ten years of
patent-protected sales to recoup their investment rather than five.
Third-world issues with pharma patents
Basically, if a country is too poor to afford to treat all its
citizens with a new drug, many may die. This has definitely been the
case with some HIV drugs. As a result, the Third World has long argued
that it should be exempt from pharmaceutical patents.
In the mid-1990's, the WTO basically agreed, and allowed (I'm not
sure of the exact terms) third-world countries to manufacture generic
equivalents of first-world drugs for use within their own borders only.
The manufacturing must be for the government's own use (eg in
distributing to its people; the manufacturing can't be a for-profit
initiative of a private company in that country). Also, a good-faith
attempt must have been made first to negotiate for a reduced-rate
license to manufacture the drug, and such profits as are made must be
given to the patent holder. In practice, large US pharmaceutical
companies often enter into vastly-reduced-price licensing arrangements
with third-world companies.
In 2005, the WTO relaxed this rule to allow poorer nations to import
generics that would otherwise be covered by a patent. This is sometimes
described as a "compulsory license". The manufacturer would not be
prosecuted.
This is perhaps the foremost third-world patent issue.
Patents Generally
The fundamental justification for patents is to encourage advances in
technology. Their justification is purely
utilitarian; it is difficult
to give a deontological rationale for not using someone else's
invention.
Ethically, the scope of the utilitarian argument can be worldwide, but
national patent laws are generally intended for national benefit. That
said, treaties have led to fairly uniform patent laws in the
industrialized world, but there are serious concerns in the Third World
about certain aspects of patent law.
Also, a part of the patent process is the requirement that the inventor disclose
the idea to the world. You cannot protect something with both a patent
and trade secrecy. To put it another way, the patent grants you a
limited monopoly in "exchange" for publication.
Software Patents
These are meant to protect software inventions. They do notprotect
software applications as such; copyright is for that. Copyright
protects source code and executable code; it has been extended to user
interfaces but (recently) not
to APIs. Copyright protects only the expression and not the underlying
idea; when user interfaces are copyrighted, someone else must modify
the interface in some ways.
Suppose, however, that in developing your application you come up with
a particularly innovative algorithm, or innovative application of an
existing algorithm, or a novel user-interface gesture (think iPhone),
or new way of presenting data, or some other novel feature. You
may be able to patent it, as an "invention". You may even be able
to patent a new business method supported by
your software. Patents are
for certain features that you consider especially innovative.
If you have a patent, then nobody else can use your idea without your permission, even if they discovered and developed it entirely separately, for the lifetime of the patent.
Why are software patents even controversial? Here are four categories of
software-patent issues:
- problems with the patent system generally
- implementation problems with the patent office as applied to software
- problems with business-method patents, which until 2005
required
involvement of "technological arts"
- fundamental problems with software patents
One example of a fundamental problem might be that algorithms are can
be seen as mathematical facts, not inventions. While novel applications
of existing algorithms can be patented, "novel" is very subjective: one
developer's breakthrough idea might be a standard software-engineering
exercise to another. An
implementation-related problem might be that the USPTO seems to have a
hard time recognizing software patent claims that are trivial instances
of well-known general ideas.
As we consider the patents below, it will often be claimed that the
innovation is "obvious". In fact, there is sometimes (maybe even often!) evidence that the
idea was so obvious that prior publication (required to invalidate a
patent based on "prior art") would have been rejected by any reputable publisher.
Part of the problem, with aspects unique to computing, is that any
invention has two parts:
- realizing you need the invention
- actually getting it to work
In the physical world, the second often tends to be the much harder
component. But in the software world, once you realize you have a need
to do something, working out an implementation (at least at the
"invention" stage) is often
very straightforward. Not always, but often. Once the need becomes
apparent, finding a solution can be obvious.
What often seems to happen is that a patent is granted for an invention
at a time when the inventor had no clear application for the idea. Much
later, someone needed to get
something done, and found a quick obvious solution that turned out to
be the earlier patented one. The Eolas case (below) is a possible
instance of this. Should early inventors be able to lock out later
developers simply because they managed to patent a technique to solve a
problem that was completely obvious later to anyone else faced with
that same problem? Haven't they really patented the problem?
At the risk of being misleading about the term "obvious" as used in
patent law, I will call this the "obvious-in-context"
problem.
Another way to describe this situation is to say that many patented software solutions are ordinary, obvious applications of basic principles of software engineering. The problem may have been new, but the solution is still obvious.
In some cases, the obvious-in-context problem boils down to a general algorithm being applied to solve a specific problem. The Flook case (below) supposedly made such inventions nonpatentable. However, this isn't always what happens in practice.
It is examples of this kind that led me to formulate my (former) three-part test on when it is appropriate to recognize
software patents:
- The Supreme Court has turned down your last appeal
- Federal marshals surround your cabin in the woods
- You are out of ammo
This marks me in some circles as a software-patent moderate.
I am currently replacing it with a new theory: that a good patent system
can have a very positive effect on business and on investment in
innovation, but that a bad patent system can have a very negative
effect. Good patents reward innovation but bad patents actually punish it.
The smartphone wars may or may not be an example of a positive
use of the patent system. If Apple invented some breakthrough phone
features for the iPhone, and patented them, then HTC / Google should
perhaps not be allowed simply to copy them in HTC's Android-based
smartphones. But exactly what happened here is less than clear; we will
revisit it below.
Aeroplane Control
A classic "broad" patent is the Wright brothers patent on
"wing-warping" to control flight (to the Wrights, the wings were the
"aero-planes", planing the air, the rest of it was the "flying
machine"). The Wright brothers actually twisted the whole biplane
structure -- using cables -- to bank in a turn. This later led to the
development by others of ailerons, which achieve the same
effect but which mechanically are entirely different. A court ruled the
Wright patent still applied: what mattered was the concept of adjusting wing angles to
tilt the craft.
This is a classic "broad patent" for a major new innovation. The
courts have traditionally recognized "broad" patents, but somehow in
software this distinction is often lost.
History of software patents
For a long time, software was held to be unpatentable,
as mathematical algorithms are unpatentable. Any fundamental
mathematical or physical laws are unpatentable.
1972: Gottschalk
v Benson (wikipedia):
You can't patent a mathematical algorithm
(in this case a number-format-conversion algorithm). More below.
1973: ATT somehow manages to patent the setuid bit, claiming it's
hardware. This patent was dedicated to the public domain in 1979. This
patent is certainly a deep idea: if a certain bit is set in the
filesystem information node for a file (not in the file itself), then
when the file is executed,
it runs with the privileges of its owner and not the user. Before then
(and after; see what Windows does), there were complex ad-hoc methods
for running selected programs with elevated (or alternative) privileges.
1978: Parker v Flook: Algorithms in a patent must be treated as
prior art; the invention must contribute something substantial beyond
the application of an algorithm.
1981: Diamond v Diehr: computer + machine IS patentable. For a
long time after, software patents always described the software in
combination with some hardware device. This patent dealt with the
curing of rubber, using a computer to guide the process.
In the Diehr ruling the Supreme Court said that an invention isn't automatically unpatentable just because it
contains an algorithm.
But PTO & lower courts read in the converse: algorithms are
patentable.
Note that the current business-world baseline thus rests
on USPTO policy and lower-court case law, NOT congress or SCOTUS.
Problem of "non-obviousness":
the rules state that it's not enough to prove it's obvious today.
Uh oh. That becomes an extremely difficult burden.
To be patentable, an invention must be "novel". Novelty is usually
challenged by the presentation of "prior art":
did someone else discover it first? Often there are arguments about
this.
If prior art is published,
it can invalidate a patent. However, if it was used privately, those
users can continue to use their idea without paying royalties to the
owner of the patent, but the patent may still stand. The patent can be challenged on the grounds of
not being novel, but this is harder.
Broad patents for fundamental new ideas, narrow patents for improvements
compatibility issues: What if the default, standard implementation
is patented? Two cases where there was at least some movement away from a
patented format:
GIF => PNG
MP3 => ogg vorbis
software patent v copyright
Supreme Court cases limit the word "process" in USC Title 35, Chapter
10, §101:
Inventions
Patentable: Whoever invents or discovers any new and useful process, machine,
manufacture, or composition of matter, or any new and useful
improvement thereof, may obtain a patent therefor, subject to the
conditions and requirements of this title.
Gottschalk v Benson, 1972
This case continues to be debated, although the Supreme Court has recently reaffirmed it. (Gottschalk is the
name of the patent commissioner; the one-word name for the case is Benson.) The court stated
that "the
patent would wholly pre-empt the mathematical formula and in practical
effect would be a patent on the algorithm itself". But that is
not
entirely clear. The algorithm was for converting so-called binary-coded
decimal (in which decimal digits are represented in sequence each by
four binary digits, so 157 = 0001 0101 0111) to true binary. The patent
was
for the use of this algorithm on a
computer, not universally (though the algorithm makes no sense
except on a computer).
In the decision, the court cited an earlier decision: "[w]hile a
scientific truth, or the mathematical expression of it, is
not a patentable invention, a novel and useful structure created with
the aid of knowledge of scientific truth may be", and "an idea of itself
is not patentable." The court also said,
Here the "process" claim is so abstract and sweeping as to cover both
known and unknown uses of the BCD to pure binary conversion. The end
use may (1) vary from the operation of a train to verification of
drivers' licenses to researching the law books for precedents and (2)
be performed through any existing machinery or future-devised machinery
or without any apparatus.
Applications of computer algorithms may indeed be general. In
retrospect, should this be an issue? The end-uses above are all for software using the algorithm; and
anyone could always use a different algorithm.
One interpretation of the decision quote above is that the Supreme
Court is willing to allow patents on algorithms that are an integral
part of an invention, but not
on algorithms that are in effect simply library routines. The Benson
algorithm was clearly in the latter category; if it had been closely
tied to a particular invention -- that is, if there had been a claim limitation -- perhaps it would have been patentable. Does that make any sense?
One school of thought is that a "process" is not patentable unless
there is a physical transformation involved, or unless the new
invention is necessary in carrying out the process; this is sometimes
referred to as the "Machine or Transformation" rule. The Supreme
Court's
Benson ruling was
narrower than that, however. Nonetheless, the court did
say
transformation and reduction of an
article 'to a different state
or
thing' is the clue to the
patentability of a process claim that does
not include particular machines. [emphasis by pld]
At one time, I might have put the emphasis differently: the clue.
However, the Supreme court backed away from that position in the Bilsky
case (below).
The Benson decision closes with an observation that lack of patentability
doesn't seem to have hurt the 1972 software industry any, and a call to
Congress to figure it out (which may have had its origins in an amicus curiae brief filed by IBM,
which was at the time against software patents):
"It is noted that the creation of programs has undergone
substantial and satisfactory growth in the absence of patent protection
and that copyright protection for programs is presently available."
If these programs are to be patentable, considerable problems are
raised which only committees of Congress
can manage, for broad powers of investigation are needed, including
hearings which canvass the wide variety of views which those operating
in this field entertain. The technological problems tendered in the
many briefs before us indicate to us that considered action by the
Congress is needed.
One possible way to understand Benson is that the patent claim limitation -- the application of the invention for which the patent was claimed -- was overly broad. Benson claimed a patent on any use of the algorithm on a computer.
In the 1980 Supreme Court decision Diamond v Chakrabarty, about the
patentability of genetically modified organisms, the following line
from a 1952 Congress report on patent law is cited
anything under the sun that is
made by man [is patentable]
How does this square with Benson?
The Benson decision is frequently cited as evidence that the Supreme
Court did not
intend to allow software patents; the language of the decision -- that
the algorithm was a fundamental mathematical fact -- can be read as
supporting this. Others use Benson as an example
of a case that, with the benefit of hindsight, seems more and more
strongly to have been wrongly decided; that in patenting an algorithm
on a computer, one is not patenting a mathematical idea. Or, at the
very least, Congress should simply heed the Court's call to address
this issue directly. It doesn't help any that the court seems to have
misunderstood the concept of widely varying applications.
Parker v Flook, 1978
In this case, the Supreme Court dealt with an invention that used
a computer to analyze temperature fluctuations within a catalytic
converter, apply a smoothing algorithm to discount transient
fluctuations, and raise an alarm when the averaged temperature was
outside of preset limits (the algorithm apparently also had something
to do with updating those limits).
The only novel feature of the method is a
mathematical formula. In Gottschalk v. Benson, we held that the
discovery of a novel and useful mathematical
formula may not be patented. The question in this case is whether the
identification of a limited category of useful, though conventional,
post-solution applications of such a formula makes respondent's method
eligible for patent protection.
The court ruled that the patent had to be judged as if the algorithm at
stake were prior art; that is, the algorithm cannot be counted as part
of the innovation.
Respondent's process is unpatentable under 101, not because it contains a
mathematical algorithm as one component, but because once that
algorithm is assumed to be within the prior art, the application,
considered as a whole, contains no patentable invention. Even though a
phenomenon of nature or mathematical formula may be well known, an
inventive application of the principle may be patented. Conversely, the
discovery of such a phenomenon cannot support a patent unless there is
some other inventive concept in its application. [emphasis added -- pld]
One difference between Flook and Benson was that in the Benson case there was essentially no claim limitation, while in Flook the claim was limited to a specific application of temperature control.
There has been some criticism of Flook
because the Supreme Court was vague as to the distinction between §101,
on patent eligibility itself, and §103, on restrictions due to prior
art and obviousness.Such a low rate would have caused potential patent plaintiffs to
think twice about filing a lawsuit using a weak patent.
Flook is arguably an important case, but alas often seems overlooked in
the real world. Consider, for example, a patent for the application of
a standard algorithm to a particular problem:
- Steir patent for overlaying a hairstyle on the image of a person
- Pardo & Landau patent for using "topological sort" to establish an evaluation order for formulas
Both of these patents were granted. Arguably, however, both are trivial applications
of standard algorithms. In fact, more often than not patents that use
general, well-known algorithms don't seem to be rejected on Flook grounds, even when there is essentially no inventiveness in the application of the algorithm.
Diamond v Diehr, 1981
On the face of it, the Supreme Court ruled here that just because an
invention incorporates a software component, this does not make the
invention ineligible for patent. The Diehr invention used a
thermocouple to monitor the temperature in a mold of curing rubber,
connected to a computer that read the raw temperature values and did
some computation to determine when the rubber was finished. While the
majority held that the invention was just patentable, Justice Stevens
(of Sony v Universal) wrote in dissent of three problems he had with
the Diehr patent [division into paragraphs added]:
First, there is not a word in the
patent application that suggests that there is anything unusual about
the temperature-reading devices used in this process -- or indeed that any
particular species of temperature-reading device should be used in it.
Second, since devices for constantly measuring actual temperatures-on a
back porch, for example-have been familiar articles for quite some
time, I find it difficult to believe that a patent application filed in
1975 was premised on the notion that a process of constantly measuring
the actual temperature had just been discovered.
Finally, the [PTO] expressly found that the only difference between the
conventional methods of operating a molding press and that claimed in
[the] application rests in those steps of the claims which relate to
the calculation incident to the solution of the mathematical problem or
formula used to control the mold heater and the automatic opening of
the press.
In other words, Stevens felt that the Diehr invention should not be patentable, because the only novel feature was the use of the computer, and algorithms should not be patentable.
The Federal Circuit
Patent cases are technically complex, and pose a problem for judges.
Therefore, in 1982, Congress decided that all patent cases appealed
from federal district courts should be heard by the newly created
Federal Circuit, instead of the Circuit Court that coresponded
geographically to the district-level trial court. The theory was that
the Federal Circuit would thus gain expertise in patent cases, and
would be better able to rule consistently.
There are still arguments in favor of this idea: patent decisions in
different circuits were often significantly different, and
"general-purpose" district-court judges typically did not have the
technical background to understand some patent issues. However, there
is also a theory that the Federal Circuit plan backfired massively. See
http://arstechnica.com/tech-policy/2012/09/how-a-rogue-appeals-court-wrecked-the-patent-system/, especially the article's title.
Overall, the Federal Circuit appears to have been from the beginning
very friendly to patent claims, and unwilling to address the broader
interests of the rest of society in having patents invalidated.
Yes, perhaps we all benefit from Apple's investment in innovation that
might not have been made without the patent system to protect it. But
for every Apple (and there is perhaps only one), there are thousands of
petty patent cases that in effect reduce
the incentive to innovate. If an innovative solution will cost you tens
of millions in unforeseen patent litigation, it's best not to bother.
Here is a quote from the article above:
In the two decades before the creation of the Federal Circuit,
appeals courts sided with patent holders in only 20 to 40 percent of
cases.
... In contrast, the Federal Circuit sided with patent holders more
than 80 percent of the time during its first year on the bench.
It was the Federal Circuit that developed the original patent test for
"obviousness": there had to be documented evidence in the prior art
that suggested that someone was aware that the combination of new
features in the new invention might be desirable. This was known as the
"teaching-suggestion-motivation" test. It led to the patenting of many
software ideas that might be so obvious that nobody would have bothered
to write them down.
The Supreme Court's Benson, Flook and Diehr patents might have been
read as banning software patents. It was largely the Federal Circuit
(with the complicity of the Patent Office) that undermined these
precedents. And it was the Federal Circuit that ruled, in the State
Street Bank case (below), that business methods were patentable.
By making the Federal Circuit hear only patent cases, the judges
focused solely on patent issues. The interests of the public were
nowhere in sight.