From: Ben Goertzel (ben@webmind.com)
Date: Sat Jun 23 2001 - 13:07:23 MDT
Hello all,
At the Extro 5 conference last week, I spoke to a number of you about
the
possibility of investing in my new firm, Cognitive Bioscience, which is
focused on applying Webmind AI technology to the bioinformatics domain,
and
initially in particular to the analysis of gene and protein expression
data.
I promised you more details; well, this e-mail contains some.
The end of this message contains an executive summary of our business
plan,
followed by a list of questions and answers about Cognitive Bioscience
and
this fundraising effort.
If you may be seriously interested in investing in Cognitive
Bioscience,
please e-mail me privately (ben@goertzel.org) and I will send you a
complete business plan (it's about 22 pages long, so there's no
bandwidth
issue, but I'd prefer to limit the distribution of it to that which is
truly necessary for fundraising).
We are interested, at this stage, in investment amounts ranging from
$1000
up to $1M. Unfortunately, anything less than $1000 is probably not
worth
our while to deal with, in terms of the legal paperwork involved.
It is a difficult time to raise money right now. The financial markets
are
tight. Investors are conservative to a fault, just as they were
generous
to a fault a year and a half ago. The tight markets have been good in
one
way: They have forced us to think very, very hard about how to create a
business model that will allow us to create substantial profits on the
way
to creating a real AI system. But the market is so tough that even with
a
good business plan and a great team, fundraising is difficult.
So I'm appealing to you guys, who recognize the value of
forward-thinking
technology, to see beyond the temporary mood of conservatism, and take
a
small risk by investing in Cognitive Bioscience. This is not a
pie-in-the-sky pure research venture, it's a very carefully calculated
plan
to capitalize on an identified market niche, and then build from that to
the
creation of an organization capable of building an highly powerful
biodata-focused AI system.
I should also add that, I know many of you are technically oriented
individuals. We are very open to accepting more than just money from
our
investors: expertise, contacts, and so forth. You don't have to view
this
as just another investment, you can view it as joining forces with
like-minded visionary compu-bio-geeks to create something amazing ...
actually two amazing things: amazing technology, and an amazing
corporation
that will be able to create amazing technology in a sustainable way
based
on profitability.
Finally, if someone would rather *donate* money for pure AI research,
than
invest it in an AI/biotech start-up, we have created a nonprofit
organization, the Real AI Institute (www.goertzel.or/realai/), for this
purpose. My preliminary inquiries indicate that it will be much easier
to
fund Cognitive Bioscience than to fund a pure research endeavor, and so
I
am focusing my efforts on the Cognitive Bioscience. But even assuming
we
succeed with Cognitive Bioscience, there is some useful work that can
be
done toward real AI right now, that doesn't fit into the framework of an
AI/biotech startup, so even a small amount of donation to the Real AI
Institute may be of value.
Yours,
Ben Goertzel
**************
Finally, without further ado, here is the business plan summary:
-------------------------------
Cognitive Bioscience
Business Plan Executive Summary
June 23, 2001
-------------------------------
Point of Contact
Ben Goertzel, CEO and Chief Scientist
201/946-6996
ben@goertzel.org
*******
Mission:
To dominate the market for gene and protein expression data analysis
software, via the creation of superior applications based on a
combination
of statistical and machine learning techniques with our unique Biomind
artificial cognition technology. And then, to leverage our domination
of
the expression data analysis niche into a major position in the
"integrated
molecular-biology data analysis" market.
*******
Cognitive Bioscience is a start-up bioinformatics software firm,
currently
seeking seed or first-round funding. Led by computer scientist Dr.
Ben
Goertzel, the firm currently consists of about a dozen former staff from
the
recently dissolved AI software firm Webmind Inc.
We are creating a suite of three products for biologists to use to
analyze
genomic and proteomic expression data: Biomind Workbench (with a high
four-figure pricetag), aimed at the individual biologist; and the
Bioknowledge Management System and Biomind Pattern Finder, an extremely
powerful database and data mining system for large-scale users, with a
six-figure pricetag and some customization generally required.
Initial product versions will focus narrowly on expression data;
subsequent
versions of the products will integrate expression data with other
types of
information including EST data, chemical databases, information on
metabolic
pathways, cellular information and "in silico biology" models of
cellular
systems.
Our long-term vision is an integration of two grand visions: AI and
post-genomic biology. There is a deep synergy here, in that AI
provides
the computational firepower needed to analyze the vast amounts of data
that
microarrays and other post-genomic data gathering techniques produce;
whereas the world of molecular biology provides a domain rich enough to
feed
an AI system with an endless variety of information that is not tied to
human sense organs or human common-sense understanding. Our initial
business model, focused on microarrays, is a pragmatic pathway to the
realization of this long-term vision.
On the back end, our products will from the very start leverage our
unique
"general intelligence" technology, as well as our years of experience
working as a team applying artificial cognition, statistical and machine
learning methods to data analysis in various domains. Our current
technology toolkit includes proprietary prediction, pattern recognition,
inference, and categorization components, integrated in an overall AI
software framework called Biomind. In the past, our tools have proven
themselves superior to competition in text categorization, information
retrieval and financial prediction, and our preliminary work indicates
that
this will also be the case in the bioinformatics domain.
Currently there are only two serious, somewhat established competitors
in
the gene and protein expression data analysis niche: Silicon Genetics
and
Spotfire. A handful of new startups are also attempting to address this
niche. But the market penetration of all these firms put together is
still
low, and although their products are reasonably good ones, on the back
end
their software is founded on standard and old-fashioned data analysis
technologies, derived from well-known ideas in the computer science and
statistics literature and implemented without subtlety. The problem
with
these data analysis methods is twofold: first, they often fail to yield
adequate results even when wielded expertly; and second, they are
difficult
for the average biologist to use. Our techniques resolve both of these
problems, providing both greater intelligence and greater automation.
Thus, the superior intelligence provided by our unique AI technology
will
give us a very significant market advantage.
In the broader area of integrated genomic and proteomic data analysis,
the
main competitor is the aptly-named Chicago firm "Integrated Genomics."
However, the data integration performed by this firm is on a relatively
primitive level, and they lack the computational tools required to do
serious bioinformatic analysis on a large and diverse body of biodata.
To complement our strong computer science, application engineering and
data
analysis expertise, we have recruited a computational biologist from a
major Japanese pharmaceutical company to join our team, and we are in
the
process of recruiting a handful of others with biotech expertise, on
both
the scientific and the marketing sides.
The Biomind technology is based on the same core algorithms as the
Webmind
AI Engine, a software system developed within Webmind Inc. over the
period
1998-2000, based largely on Dr. Goertzel's mathematical cognitive
science
theories. However, Biomind is implemented primarily in C whereas the
Webmind Inc. software was in Java; and the Biomind software architecture
is
completely different than that of the Webmind Inc. AI Engine.
The leaders of the Cognitive Bioscience team include
-- Dr. Ben Goertzel , Chief Scientist and CEO, former Webmind Inc.
founder,
Chief Scientist and CTO
-- Cassio Pennachin, VP of Engineering, former Webmind Inc. VP of AI
Development
-- Stephan Vladimir Bugaj, VP of Product Development, current chief
software
architect at the biotech firm Proteometrics, former Webmind Inc. Deputy
CTO
Dr. Goertzel, Mr. Bugaj and a handful of the others are based in the US;
Mr.
Pennachin and the majority of the technical team are currently based in
Belo Horizonte, Brazil. (Brazil features a stable economy and a labor
cost
roughly 1/3 that of the US; the US and Brazilian portions of our team
have
worked together for 3 years and have an outstandingly smooth operating
relationship.)
Assuming funding commencing August 1 2001, we anticipate bringing our
first
product to beta in November 2001, and launching it generally on
February 1,
2002. The second product, with significantly expanded capabilities,
will
launch in Fall 2002.
Assuming first-round funding of $1M, we anticipate achieving
profitability
within roughly 12 months of funding, and 2003 revenues in the $20-$80M
range or higher.
*************
QUESTIONS AND ANSWERS ABOUT COGNITIVE BIOSCIENCE AND THIS FUND-RAISING
EFFORT
*********
Q: How much money are you seeking?
A:
Generally, we are seeking $1M of first-round venture financing to get
this
new company off the ground. We believe that $1M will bring us to the
point
of profitability, within a single year. There is a fairly strong
argument
that we should take $1.5M or even $2M instead, if it is offered, in
order
to give us more of a cushion against bad luck events.
However, our shorter-term goal is to raise $100K in seed funding. This
$100K of money will pay for 6 Brazilian engineers to work for 6 months
(I
won't pay myself out of it). In the first 4 of these 6 months, they
will be
able to build an excellent prototype demonstrating the applicability of
our
technology to the bioinformatics domain. With this prototype in hand I
have
little doubt I'll be able to raise VC money, or get a deal with a major
pharmaceutical or biotech hardware firm. Without it, I may still
succeed,
but it isn't as certain for obvious reasons. (Of course, more than
$100K is
even better).
*********
Q: Is it legal to raise money like this?
A:
Yes, if it's conducted properly. Please remember that I've been through
this before. Rule 504 of Regulation D states that it's OK according to
federal law for a firm to raise up to $1 million from private investors
with
very few restrictions. Before actually accepting money I will of course
retain a lawyer to take care of the paperwork and ensure that everything
is
executed properly. There are thus two phases involved in this sort of
capital-raising effort:
1) find out on a "handshake" (or e-handshake) basis who is interested in
investing how much, i.e. who is "in" on this seed financing round
2) set up the legal paperwork, get everyone involved in the funding
round to
sign it, and then actually collect the money
*********
Q: How does this work relate to your long-term project to build a "real
AI",
achieve the Singularity, and so forth?
A:
Cognitive Bioscience's core technology is Biomind, which is a particular
application of the Webmind AI Engine. Thus, the core AI work done for
Cognitive Bioscience is work that pushes us toward the end goal of
creating
a real AI.
Of course, there is a lot of work involved in building and marketing
bioinformatics products that is not AI work, and that does not directly
push toward the goal of creating real AI.
However, unless someone emerges with a plan to substantially fund real
AI
development work, the best path to creating real AI seems to be the
Cognitive Bioscience path: create a company that leverages real-AI-bound
technology to create profit in a particular market, and then use the
profits from this enterprise to fund more and more ambitious real AI
development over time.
It is also a fact that an AI system requires a rich data domain to study
and
act within. Lacking robotic sensors and actuators, the human-scale
physical world is not a good one. The financial domain, which we
experimented with in Webmind Inc., also has its flaws. The codic
domain,
while an important one in terms of the long-term goal of self-modifying
AI
and the Singularity, is in my view a very poor initial domain for a
baby
AI, because (to put a very long story short) it consists largely of
processes rather than data. The molecular biology domain actually has a
lot
of advantages here: there's a lot of data of diverse kinds, and human
common sense intuition is not required to understand most of it. All in
all, this seems a decent domain for an AI system on which to cut its
teeth.
When we begin experimenting with conversational interaction as an
interface
to biological information (planned for mid-2002, after the first product
releases), our "baby AI" will have a lot of knowledge about genes and
proteins to share.
It should also be noted that all of us involved in Cognitive Bioscience
also
have a real, deep and abiding interest in cell development and
functioning
and postgenomic biology. This is not just a byway on the path to real
AI,
it's a fascinating technological and business challenge unto itself.
Applying Biomind to help understand how cells get old, for example, is
an
application of immense interest to all of us who prefer not to have our
human bodies die....
*********
Q: Your last company, Webmind Inc., went bankrupt. So why on Earth
should
we want to invest in your new company?
A:
A flip answer is: Well, now I and my collaborators know what *not* to
do....
More seriously, my view of the Webmind Inc. bankruptcy is given in the
article I posted here before,
http://www.goertzel.org/benzine/WakingUpFromTheEconomyOfDreams.htm
Webmind Inc. never had a credible business model. This was realized at
many
points during late 1999 and 2000, and we tried to correct it, but as it
turned out we made our big mistake early on. In 1999 we shifted from a
financial-prediction-based business model to an internet information
management tools based business model. This seemed clever at the time
because financial funds were priced at 2 times earnings whereas
internet
companies were priced at infinity times earnings. The result was that
our
excellent financial prediction technology was mothballed (it now has 11
months of real-time paper-trading track record, but this comes too late
to
save Webmind Inc., and out of the Webmind Inc. bankruptcy, this code
will
be owned by others than me and I'll have little to do with it, although
I
invented a good bit of the technology underlying this application).
Now
almost every company involved in internet information management is
going
under, the most recent one being Artificial Life, which was a really
interesting company. (Autonomy and Verity are still doing OK -- the
market
is consolidating.)
On the technology side, we created an amazingly detailed AI design and
implemented various modules of it and tested them, invariably showing
that
their performance exceeded that of anything else out there. But we did
make one bad mistake: we implemented our system using a Java-based
agent
system framework that never delivered adequate computational
performance.
Java simply did not improve in its large-scale performance as rapidly as
we
thought it would.
Now we have a very solid business model, with a clear path to
profitability,
and we are reimplementing our AI design in C using a much more
efficient
software architecture. The reimplementation is going fast: after a 2
month
effort by 3 people, it's about 1/3 done.
On a personal level, I must take substantial responsibility for Webmind
Inc.'s technical failures; but as for the business problems, I consider
my
responsibility to be limited, though greater than zero. I never had a
formal day-to-day role in Webmind Inc.'s business matters, and my
counsel
on business matters was certainly not heeded all that frequently.
I and my team are older and wiser now. We know what we're doing to a
vastly
greater extent than we did when Webmind inc. was founded (as
Intelligesis)
in 1997). And we are more dedicated than ever to put in years of hard
work
to achieve our long-term goals.
It's a cliche' quote but awfully appropriate: "That which doesn't kill
me,
makes me stronger."
***********
Q: How does this new company relate to Webmind Inc. legally?
A:
Cognitive Bioscience a new corporate entity.
Webmind Inc. is preparing to file for Chapter 11 bankruptcy. A
prepackaged
bankruptcy deal is being prepared, according to which a certain large
firm,
together with our former CEO Andy Siciliano, will bid to purchase the
intellectual property (IP) of the company. It is unlikely there will
be
any other bidders.
We have agreed to give the owners of the Webmind Inc. IP a certain
percentage of Cognitive Bioscience, in exchange for permission to
freely
use Webmind Inc. designs and code (at present, due to our rearchitecting
of
the system, we have more use for the designs than for most of the
code).
The exact percentage is the subject of current, friendly negotiations,
which
cannot be completed until the Webmind Inc. bankruptcy is completed.
It would be nice to wait until all this were done before raising seed
money,
but, this doesn't seem like a plausible course because, basically, all
of
us who are now working for free need to eat.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 08:08:17 MST