Cats, Brains and Computers: A Dangerous Cocktail?

Hoagy: runs hot on kibble.
I read, with great interest (and not a small amount of humor), a recent press release from the University of Michigan, about development of a computer similar to the cat brain. The research published in Nano Letters in April: “Nanoscale Memristor Device as Synapse in Neuromorphic System” (1).

A skeptic, I went immediately to the whys and hows of such an undertaking. Continually amazed at my cats’ bizarre yet occasionally functional behavior (presumably directed by their brains, although one often wonders), the thought of using such a brain to model a computer brought feelings of amazement…and perhaps, concern.

Why a cat brain as a computer? Apparently, speed and energy efficiency have a hand here. Cats can recognize a face faster and more efficiently than a computer—in fact, faster than a supercomputer—who knew? Even a computer with a dedicated power supply and 140,000 CPUs is reported to perform 83 times slower than a cat’s brain, according to the author of this study, University of Michigan computer engineer Wei Lu. (May I point out that kitties require a little dry kibble, or maybe inexpensive canned salmon as their dedicated power supply?)

Lu has been working to develop a computer based on a “completely different paradigm, compared to conventional computers”. While a cat’s brain is complex, it is simpler than a human brain and thus considered a reasonable goal for this new model (2).

In terms of how to accomplish this new computer model, a bit of background (you Neurologists and Computer Engineers can skip this part): The mammalian brain works, in the most basic terms, by chemical and electrical signals moving between neurons via a synapse. Synapses can be thought of as reconfigurable switches that form pathways between thousands of neurons. Synapses actually form a memory of these pathways based on the strength and timing of the electrical signals the neurons generate.

In computers, logic and memory function are located at different places in the circuit and each computing unit is connected to just a handful of others near it in circuit. Thus computers execute code in a linear fashion, line by line.

A mammalian brain, on the other hand, can execute many operations at once, in parallel. For instance, most of us can recognize a face almost instantly, and with far less energy than that used a computer. (How’s that for a psychological boost—you can think faster and more efficiently than a supercomputer!)

Towards this project, computer engineer Wei Lu had previously built a “memristor”, which is a hybrid of a transistor and a biological switch or synapse. The memristor can remember past voltages to which it has been exposed.

In this work, Lu connected two circuits with his memristor, and demonstrated that the system was capable of memory, as well as a form of learning called “spike timing dependent plasticity”. “Plasticity” refers to the ability of connections (the synapses) between neurons to gain strength based on their stimulation and that of neighboring neurons. Spike timing dependent plasticity is thought to be the basis of memory and learning in mammalian systems.

One potential advantage of a cat brain-simulated computer would be a machine with the ability for adaptation. For instance, a conventional computer could chart the path through a living room full of furniture, given the coordinates of the furniture and the door, but if the sofa were moved, that path would need to be altered. A conventional computer could not make such an adjustment, on the fly.

Claims of computers that simulate a cat’s brain are not new. IBM announced such a computer in late 2009 (3). However, strong criticism of IBM’s announcement came from Henry Markham, head of the Blue Brain project at Lusanne, Switzerland-based EPFL, where his team is attempting to reverse engineer the mammalian brain (4).

In an interview last December with’s Greg Fish, Markham noted, in terms of IBM’s and others’ work to simulate the mammalian brain, that they aren’t the kind of simulations that help one understand the brain (4).

I have no intention in joining the debate, but Markham makes an interesting point on the importance neural networks and their potential for solving computing problems that normal artificial intelligence cannot easily solve (4).


  1. Sung Hyun Jo, Ting Chang, Idongesit Ebong, Bhavitavya B. Bhadviya, Pinaki Mazumder and Wei Lu. (2010) Nanoscale Memristor Device as Synapse in Neuromorphic Systems. Nano Lett., 10, p. 1297–1301. PMID: 20192230 Online Publication March 1, 2010
  2. Unconventional Computer Modeled on Cat’s Brain. Scientific Computing. (April 2010).

  3. IBM Moves Closer To Creating Computer Based on Insights From The Brain.
    (Nov. 2009)
  4. IBM Cat Brain Computer Debunked. (Dec. 2009)
The following two tabs change content below.

Kari Kenefick

Kari has been a science writer/editor for Promega since 1996. Prior to that she enjoyed working in veterinary microbiology/immunology, and has an M.S. in Bacteriology, U of WI-Madison. Favorite topics include infectious disease, inflammation, aging, exercise, nutrition and personality traits. When not writing, she enjoys training her dogs in agility and obedience. About the practice of writing, as we say for cell-based assays, "add-mix-measure".

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.