A challenge - find the most ridiculous Wolfram Research claim.
Dave
I'm going to set a challenge - all in the name of a bit of fun.
Please find the ridiculous Wolfram Research claim. Provide a link to the
web page where it can be found. (Keep it to Wolfram Research web sites
only).
Here are 3 of my favorites. Can anyone find any better.
1) "Every Demonstration undergoes a rigorous review process that checks
for quality, clarity, and accuracy, so you can count them as academic
publications."
http://demonstrations.wolfram.com/FAQ.html
2) "For academic purposes, Wolfram|Alpha is a primary source."
http://www.wolframalpha.com/faqs.html
3) "The standards of correctness for Mathematica are certainly much
higher than for typical mathematical proofs."
http://reference.wolfram.com/mathematica/tutorial/TestingAndVerification.html
--
I respectfully request that this message is not archived by companies as
unscrupulous as 'Experts Exchange' . In case you are unaware,
'Experts Exchange' take questions posted on the web and try to find
idiots stupid enough to pay for the answers, which were posted freely
by others. They are leeches.
Roman Pearce
every single release is described as "revolutionary". Why release new
versions -- was last year's revolution not good enough ? There are
not enough pitchforks in the world for all these revolutions. Or
maybe they mean "revolution" as in "the earth revolved around the sun".
Dave
To say "every company" is not true. I have certainly come across
companies that delivery more than they claim.
I would agree however, a lot do make silly claims.
But with all the companies that I have dealt with on a *professional*
basis over the years, I can't think of one that makes such ridiculous
claims as Wolfram Research do.
Wolfram Research have some very bright, helpful people like Daniel
Lichtblau who I have a lot of respect for. You wont find him making
silly claims. But some of the rubbish the web site has, is not really
appropriate for a company whose main product (Mathematica) is an
expensive product aimed at a professional market.
William Elliot
> Please find the ridiculous Wolfram Research claim. Provide a link to the web
> page where it can be found. (Keep it to Wolfram Research web sites only).
>
> Here are 3 of my favorites. Can anyone find any better.
>
Yes, Wolfram "Research".
> 1) "Every Demonstration undergoes a rigorous review process that checks for
> quality, clarity, and accuracy, so you can count them as academic
> publications."
> http://demonstrations.wolfram.com/FAQ.html
What demonstration? Has there even been any
orignal work resulting from their research?
> 2) "For academic purposes, Wolfram|Alpha is a primary source."
>
> http://www.wolframalpha.com/faqs.html
>
As they did not discover any theorems, they are not a primary source.
The primary source is the journal that first published the theorem.
> 3) "The standards of correctness for Mathematica are certainly much higher
> than for typical mathematical proofs."
>
> http://reference.wolfram.com/mathematica/tutorial/TestingAndVerification.html
What a joke. They haven't even corrected "primary source" to "reference
source" nor "Research" to "Handbook of Mathematics and Physics".
Welcome to "Wolfram scientific reference and software".
The Big Bad Wolfram is proud to offer the best
touted, proprietory intellectual property poop.
clicl...@freenet.de
Dave schrieb:
> Vladimir Bondarenko is often setting "Simplification Challenges" so here
> I'm going to set a challenge - all in the name of a bit of fun.
>
>
> Please find the ridiculous Wolfram Research claim. Provide a link to the
> web page where it can be found. (Keep it to Wolfram Research web sites
> only).
>
> Here are 3 of my favorites. Can anyone find any better.
>
> 1) "Every Demonstration undergoes a rigorous review process that checks
> for quality, clarity, and accuracy, so you can count them as academic
> publications."
>
> http://demonstrations.wolfram.com/FAQ.html
>
> 2) "For academic purposes, Wolfram|Alpha is a primary source."
>
> http://www.wolframalpha.com/faqs.html
>
> 3) "The standards of correctness for Mathematica are certainly much
> higher than for typical mathematical proofs."
>
> http://reference.wolfram.com/mathematica/tutorial/TestingAndVerification.html
>
Two proposals:
1. (Oleg Marichev) on the Wolfram Functions site: "the formulae were
checked from beginning to end - with Mathematica", or words to this
effect.
... yet the site is as bug infested as a printed collection of
formulae; for WRI's admission of this compare last year's
sci.math.symbolic thread "200000+ new formulas on the web" <http://
groups.google.de/group/sci.math.symbolic/browse_thread/thread/
553e3d1362177004>. And to this date, the site still hasn't been
checked with Mathematica.
2. Wolfram Integrator response: "Mathematica could not find a formula
for your integral. Most likely this means that no formula exists".
... to be able to cite this as an academic achievement enter (1-x)^(a
+b-c) Hypergeometric2F1[a,b,c,x] at <http://integrals.wolfram.com/
index.jsp>, the non-existent integral of which is (c-1)/((a-c+1)(b-c
+1)) (1-x)^(a+b-c+1) Hypergeometric2F1[a,b,c-1,x]. Try to guess when
will they will have put this one into the reach of Mathematica's
lookup table.
Martin.
Daniel Lichtblau
> Dave schrieb:
>
>
>
> > Vladimir Bondarenko is often setting "Simplification Challenges" so here
> > I'm going to set a challenge - all in the name of a bit of fun.
>
> > Please find the ridiculous Wolfram Research claim. Provide a link to the
> > web page where it can be found. (Keep it to Wolfram Research web sites
> > only).
>
> > Here are 3 of my favorites. Can anyone find any better.
>
> > 1) "Every Demonstration undergoes a rigorous review process that checks
> > for quality, clarity, and accuracy, so you can count them as academic
> > publications."
>
> >http://demonstrations.wolfram.com/FAQ.html
>
> > 2) "For academic purposes, Wolfram|Alpha is a primary source."
>
> >http://www.wolframalpha.com/faqs.html
>
> > 3) "The standards of correctness for Mathematica are certainly much
> > higher than for typical mathematical proofs."
>
>
> Two proposals:
>
> 1. (Oleg Marichev) on the Wolfram Functions site: "the formulae were
> checked from beginning to end - with Mathematica", or words to this
> effect.
>
> ... yet the site is as bug infested as a printed collection of
> formulae; for WRI's admission of this compare last year's
> sci.math.symbolic thread "200000+ new formulas on the web" <http://
> groups.google.de/group/sci.math.symbolic/browse_thread/thread/
> 553e3d1362177004>. And to this date, the site still hasn't been
> checked with Mathematica.
It's a bit tricky to check "all possibilities" unless one knows in
advance what they are. That said, and knowing a bit about how the
people behind that site do what they do, I suspect that much of it has
been checked with Mathematica. Such checks are neither proof nor
foolproof.
> 2. Wolfram Integrator response: "Mathematica could not find a formula
> for your integral. Most likely this means that no formula exists".
>
> ... to be able to cite this as an academic achievement enter (1-x)^(a
> +b-c) Hypergeometric2F1[a,b,c,x] at <http://integrals.wolfram.com/
> index.jsp>, the non-existent integral of which is (c-1)/((a-c+1)(b-c
> +1)) (1-x)^(a+b-c+1) Hypergeometric2F1[a,b,c-1,x]. Try to guess when
> will they will have put this one into the reach of Mathematica's
> lookup table.
>
> Martin.
I am a bit puzzled. I tried this integral in Mathematica, the
Integrator, and Wolfram|Alpha. The first gave the below result within
a second.
-((Gamma[c]*(-1 + Gamma[-1 + c]*HypergeometricPFQRegularized[
{-1 - a + c, -1 - b + c}, {-1 + c}, x]))/
((1 + a - c)*(-1 - b + c)*Gamma[-1 + c]))
The Integrator responded with a box:
"You need to use Mathematica to evaluate this expression. Download a
trial version of Mathematica [here]."
I need to look into that, as I was not aware it might do such a thing.
W|A gave the same result as Mathematica.
Daniel Lichtblau
Wolfram Research
Dave
> Vladimir Bondarenko is often setting "Simplification Challenges" so here
> I'm going to set a challenge - all in the name of a bit of fun.
>
>
> Please find the ridiculous Wolfram Research claim. Provide a link to the
> web page where it can be found. (Keep it to Wolfram Research web sites
> only).
>
> Here are 3 of my favorites. Can anyone find any better.
>
> 1) "Every Demonstration undergoes a rigorous review process that checks
> for quality, clarity, and accuracy, so you can count them as academic
> publications."
>
> http://demonstrations.wolfram.com/FAQ.html
>
> 2) "For academic purposes, Wolfram|Alpha is a primary source."
>
> http://www.wolframalpha.com/faqs.html
>
> 3) "The standards of correctness for Mathematica are certainly much
> higher than for typical mathematical proofs."
>
> http://reference.wolfram.com/mathematica/tutorial/TestingAndVerification.html
>
Here's one more of my favorites:
5) "Wolfram Research is transforming the way the world publishes
technical documents with Wolfram Publicon"
http://www.wolfram.com/products/publicon/index.html
Does anyone feel Publicon has transformed the way the world publishes
technical documents?
clicl...@freenet.de
Daniel Lichtblau schrieb:
> On Jun 13, 5:17 am, cliclic...@freenet.de wrote:
> >
> > Two proposals:
> >
> > 1. (Oleg Marichev) on the Wolfram Functions site: "the formulae were
> > checked from beginning to end - with Mathematica", or words to this
> > effect.
> >
> > ... yet the site is as bug infested as a printed collection of
> > formulae; for WRI's admission of this compare last year's
> > sci.math.symbolic thread "200000+ new formulas on the web" <http://
> > groups.google.de/group/sci.math.symbolic/browse_thread/thread/
> > 553e3d1362177004>. And to this date, the site still hasn't been
> > checked with Mathematica.
>
> It's a bit tricky to check "all possibilities" unless one knows in
> advance what they are. That said, and knowing a bit about how the
> people behind that site do what they do, I suspect that much of it has
> been checked with Mathematica. Such checks are neither proof nor
> foolproof.
>
It is clear that automatic numerical checking can miss the failure of
a formula because some specific parameter value was not among the
numbers tried, but I am not referring to this kind of problem; a
reader will not try to e.g. divide by zero anyway. I am referring to
formulae that fail for an input-space fraction of "non-zero measure",
like the example given in the thread cited above. These can be found
reliably with randomized numerical checking:
Use a Gaussian distribution around zero with half width 10 for Re and
Im of complex numbers as well as for real numbers, for each other
complex number substitute one of r or I*r or (1+I)*r where r is real,
for each other random real substitute a random p/q with integer p and
q /= 0 uniformly distributed over [-10,10], for each other p use pi*p,
repeat until the tuple of parameters is allowed by the constraints,
then plug into the formula and compare the left-hand and right-hand
sides, and let this checker run continuously over all formulae in the
functions site. Automatically flag a formula as "not well-formed" or
with the numbers of successes and failures tallied so far. Display
this information along with the formula on the site.
Is WRI afraid of the thousands of inconsistencies this would uncover
between the numerics of Mathematica and the symbolics of the formula
collection? Why isn't your QA department doing this anyway?
>
> > 2. Wolfram Integrator response: "Mathematica could not find a formula
> > for your integral. Most likely this means that no formula exists".
> >
> > ... to be able to cite this as an academic achievement enter (1-x)^(a
> > +b-c) Hypergeometric2F1[a,b,c,x] at <http://integrals.wolfram.com/
> > index.jsp>, the non-existent integral of which is (c-1)/((a-c+1)(b-c
> > +1)) (1-x)^(a+b-c+1) Hypergeometric2F1[a,b,c-1,x]. Try to guess when
> > will they will have put this one into the reach of Mathematica's
> > lookup table.
> >
> > Martin.
>
> I am a bit puzzled. I tried this integral in Mathematica, the
> Integrator, and Wolfram|Alpha. The first gave the below result within
> a second.
>
> -((Gamma[c]*(-1 + Gamma[-1 + c]*HypergeometricPFQRegularized[
> {-1 - a + c, -1 - b + c}, {-1 + c}, x]))/
> ((1 + a - c)*(-1 - b + c)*Gamma[-1 + c]))
>
> The Integrator responded with a box:
> "You need to use Mathematica to evaluate this expression. Download a
> trial version of Mathematica [here]."
> I need to look into that, as I was not aware it might do such a thing.
>
> W|A gave the same result as Mathematica.
>
I tried this integral a few weeks back, in the context of the thread
"hypergeometric to elliptic and back". I must have entered it as
(1-x)^(a+b-c) HypergeometricPFQ[{a,b}, {c}, x]
which produces the "most likely no formula exists" message now too.
Another example is:
(1-x)^(b-2) HypergeometricPFQ[{a,b}, {c}, x]
Now, does this one exist or not?
Sorry for the bobble,
Martin.
Daniel Lichtblau
I also was referring to that situation.
> Use a Gaussian distribution around zero with half width 10 for Re and
> Im of complex numbers as well as for real numbers, for each other
> complex number substitute one of r or I*r or (1+I)*r where r is real,
> for each other random real substitute a random p/q with integer p and
> q /= 0 uniformly distributed over [-10,10], for each other p use pi*p,
> repeat until the tuple of parameters is allowed by the constraints,
> then plug into the formula and compare the left-hand and right-hand
> sides, and let this checker run continuously over all formulae in the
> functions site. Automatically flag a formula as "not well-formed" or
> with the numbers of successes and failures tallied so far. Display
> this information along with the formula on the site.
This will work quite well if the set of "bad" values lies, say, in a
sector emanating from the origin. I'm sure I have seen similar tests
used for numerical verification of some work on functions.wolfram.com.
If there are situations where "bad" parameter ranges lie away from the
origin, or involve combinations of values in a multidimensional space,
this test may be less useful. That was what I meant when implying one
must know where to look.
> Is WRI afraid of the thousands of inconsistencies this would uncover
> between the numerics of Mathematica and the symbolics of the formula
> collection? Why isn't your QA department doing this anyway?
For that first, it would simply mean removing symbolic results from
functions.wolfram.com that were regarded as incorrect. I think we all
agree that removal of incorrect formulas would be a good thing.
For the second, so far as I am aware the QA people are not asked to
look into work on functions.wolfram.com. That work which gets
incorporated in some way into Mathematica is a different matter. Given
that f.w.c is pretty much pure philanthropy, this makes sense to me:
we all suffer from finite resources.
You can send feedback for the site to comm...@functions.wolfram.com
Alternatively, you can send bug or other feedback via
http://functions.wolfram.com/signin/
Either of these should suffice for reporting bugs, I think.
> > > 2. Wolfram Integrator response: "Mathematica could not find a formula
> > > for your integral. Most likely this means that no formula exists".
>
> > > ... to be able to cite this as an academic achievement enter (1-x)^(a
> > > +b-c) Hypergeometric2F1[a,b,c,x] at <http://integrals.wolfram.com/
> > > index.jsp>, the non-existent integral of which is (c-1)/((a-c+1)(b-c
> > > +1)) (1-x)^(a+b-c+1) Hypergeometric2F1[a,b,c-1,x]. Try to guess when
> > > will they will have put this one into the reach of Mathematica's
> > > lookup table.
>
> > > Martin.
>
> > I am a bit puzzled. I tried this integral in Mathematica, the
> > Integrator, and Wolfram|Alpha. The first gave the below result within
> > a second.
>
> > -((Gamma[c]*(-1 + Gamma[-1 + c]*HypergeometricPFQRegularized[
> > {-1 - a + c, -1 - b + c}, {-1 + c}, x]))/
> > ((1 + a - c)*(-1 - b + c)*Gamma[-1 + c]))
>
> > The Integrator responded with a box:
> > "You need to use Mathematica to evaluate this expression. Download a
> > trial version of Mathematica [here]."
> > I need to look into that, as I was not aware it might do such a thing.
>
> > W|A gave the same result as Mathematica.
>
> I tried this integral a few weeks back, in the context of the thread
> "hypergeometric to elliptic and back". I must have entered it as
>
> (1-x)^(a+b-c) HypergeometricPFQ[{a,b}, {c}, x]
>
> which produces the "most likely no formula exists" message now too.
Mathematica returns it unevaluated. I don't know why, given it handles
the other form you had indicated.
> Another example is:
>
> (1-x)^(b-2) HypergeometricPFQ[{a,b}, {c}, x]
>
> Now, does this one exist or not?
Again, Mathematica returns unevaluated when using the PFQ form for
input, and gives a lengthy result when using 2F1.
>
> Sorry for the bobble,
>
> Martin.
Daniel Lichtblau
Wolfram Research
clicl...@freenet.de
Daniel Lichtblau schrieb:
> On Jun 13, 7:24 pm, cliclic...@freenet.de wrote:
> > Daniel Lichtblau schrieb:
> > > On Jun 13, 5:17 am, cliclic...@freenet.de wrote:
> >
> > > > 1. (Oleg Marichev) on the Wolfram Functions site: "the formulae were
> > > > checked from beginning to end - with Mathematica", or words to this
> > > > effect.
> >
> >
> > Is WRI afraid of the thousands of inconsistencies this would uncover
> > between the numerics of Mathematica and the symbolics of the formula
> > collection? Why isn't your QA department doing this anyway?
>
> For that first, it would simply mean removing symbolic results from
> functions.wolfram.com that were regarded as incorrect. I think we all
> agree that removal of incorrect formulas would be a good thing.
>
> For the second, so far as I am aware the QA people are not asked to
> look into work on functions.wolfram.com. That work which gets
> incorporated in some way into Mathematica is a different matter. Given
> that f.w.c is pretty much pure philanthropy, this makes sense to me:
> we all suffer from finite resources.
>
> You can send feedback for the site to comm...@functions.wolfram.com
> Alternatively, you can send bug or other feedback via
> http://functions.wolfram.com/signin/
> Either of these should suffice for reporting bugs, I think.
>
I have no doubt that this is why I continue to stumble over bugs in
the functions site that could easily have been removed by the kind of
numerical checking I outlined. The only problem is that Marichev's
attempt to make readers believe the formula collection was
systematically checked with Mathematica (see <http://blog.wolfram.com/
2008/05/06/two-hundred-thousand-new-formulas-on-the-web/>) thus
remains ridiculous hype. And that's what this thread is about ...
>
> > > > 2. Wolfram Integrator response: "Mathematica could not find a formula
> > > > for your integral. Most likely this means that no formula exists".
> >
> > [...] I must have entered it as
> >
> > (1-x)^(a+b-c) HypergeometricPFQ[{a,b}, {c}, x]
> >
> > which produces the "most likely no formula exists" message now too.
>
> Mathematica returns it unevaluated. I don't know why, given it handles
> the other form you had indicated.
>
> > Another example is:
> >
> > (1-x)^(b-2) HypergeometricPFQ[{a,b}, {c}, x]
> >
> > Now, does this one exist or not?
>
> Again, Mathematica returns unevaluated when using the PFQ form for
> input, and gives a lengthy result when using 2F1.
>
Actually the antiderivative can be expressed quite compactly as
(c-1)/((a-c+1)(b-1)) (1-x)^(b-1) Hypergeometric2F1[a,b-1,c-1,x].
So the Integrator (as well as Mathematica) is missing an entire class
of PFQ integrals, making your "Most likely this means that no formula
exists" message look laughable too. Perhaps you should be more careful
when selling failures as features ...
But strictly speaking, you are on the safe side here, I think: Even an
"Integrator" returning this message for any well-formed input would
not be lying, because the set of well-formed inputs dwarfes that of
integrable ones. Whence the message seems to belong to the realm of
meaningless marketing blurb anyway ...
Martin.
Vladimir Bondarenko
> Perhaps you [Wolfram Research - VB] should be
> more careful when selling failures as features ...
:)