This was recently posted by Steve Farmer on the Indo-Eurasian Research list (and is reproduced here with his permission).

**********

Smithsonian magazine just published another ludicrous article on
the work of Rao et al., even at this late date not checking with a
single Indus archaeologist, script expert, computational linguist
(Mark Liberman, Fernando Pereira), or statistician/ mathematician
(Cosma Shalizi, etc.) who has spoken out against the article --
nor even apparently reading the Wikipedia article on the
so-called Indus script to see if there were any criticisms of Rao's work.

Popular science writing really stinks -- and it is getting worse
yearly. Is this an unintended consequence of Internet? I remember
from years ago when popular science writing, before Internet, was
much more accurate. (I grew up on Scientific American when
publication there actually meant something. Smithsonian magazine
is of course something else: mainly an avenue to draw donors.)

Here's a link to the article:

[tinyurl.com]

The stupidities in the article -- more polite language isn't appropriate --
begin with the title ("Can computers decipher a 5,000 year-old language?").

Note to the author: (1) Neither scripts nor would-be scripts are
"languages", and languages don't get "deciphered" . (2) No one
seriously thinks that any significant body of Indus symbols go back
5,000 years (we do get things looking like anthropomorphic plants
on earlier pottery sherds from that far back, but it's quite a stretch
to think of those as strings of symbols).

Below is the comment I sent to Smithsonian magazine the second I read
the article, which is based on no research at all -- just a credulous
interview with Rao, apparently.

Steve

*********

Note to Smithsonian:

It is too bad that the author of this article didn't bother to talk
to anyone in Indus studies about the work of Rao et al., whose work
was thoroughly discredited almost immediately by a series of well-
known computational linguists -- who quickly demonstrated that
"conditional entropy" cannot distinguish linguistic from
nonlinguistic symbols -- and the present author and his colleagues,
who presented new evidence that the Indus symbols were not part of a
writing system at a major Indus conference in Kyoto, Japan, in May.
Of the roughly 40 Indus researchers and linguists attending that
conference, Rao and his colleagues did not have a single defender.

For our initial refutation of Rao et al. published within hours of
publication of his paper, see:

[www.safarmer.com] .

For data further showing -- using far more data than Rao used in his
paper -- showing that "conditional entropy" cannot distinguish
linguistic from nonlinguistic signs or even what language families a
language belongs to, see:

[www.safarmer.com]

The latter document also has links to scathing discussions by the
influential computational linguists Mark Liberman and Fernando
Pereira that further undermine the work of Rao et al. You'll also
find there a link to the well-known 2004 paper by me and my
colleagues (Michael Witzel of Harvard and Richard Sproat of Oregon
Health and Science University) that Rao et al. claimed to "refute".

Discussion of the odd handling of the evidence in Rao's work has been
discussed widely in the archaeological and computational community,
and even in public discussions on Wikipedia. The big question is why
Smithsonian magazine would publish this article on Rao et al. without
doing a little fact checking first and without talking to Rao's
critics: that is not responsible scientific reporting and badly
misleads the public.

Steve Farmer, Ph.D.

Palo Alto, California
[www.safarmer.com]

************

Hermione
Director/Moderator - The Hall of Ma'at

Rules and Guidelines

hallofmaatforum@proton.me