Software, ratware, and catware

Originally published 8 December 1997

The inva­sion of the roborats!

From the Cal­i­for­nia Insti­tute of Tech­nol­o­gy in Pasade­na come reports of the first sil­i­con chips that incor­po­rate liv­ing brain cells.

Neu­ro­physi­cist Jerome Pine and his col­leagues have found a way to mesh embry­on­ic rat brain cells with sil­i­con cir­cuits, pos­si­bly antic­i­pat­ing the day when com­put­ers will be fab­ri­cat­ed of a com­bi­na­tion of liv­ing and non-liv­ing mate­ri­als.

Pine’s team etched 16 tiny pits in the sur­face of a sil­i­con chip. Each pit is con­nect­ed to the sur­face by a net­work of tun­nels. A liv­ing rat brain cell, or neu­ron, is placed in each pit. As the cell grows, it sends out den­drites — fin­ger­like ten­drils — that wind their way through the tun­nels and make nor­mal synap­tic con­nec­tions with oth­er cells. The tun­nels hold the neu­rons in place on the chip.

The chip is immersed in a nutri­ent broth that keeps the cells alive.

Once con­tact between cells is estab­lished, elec­tri­cal activ­i­ty begins. The cells “talk” to each oth­er, just as they do in a liv­ing rat’s brain. Elec­tri­cal cir­cuits in the sil­i­con chip stim­u­late the cells and lis­ten in on their com­mu­ni­ca­tion. The researchers hope the net­work of cells can be made to “learn” by repeat­ed stimulation.”

So far, the cells on these “neu­rochips” have been kept alive for only two weeks.

Pre­vi­ous­ly, sci­en­tists have attempt­ed to under­stand neur­al activ­i­ty by plac­ing elec­trodes in a liv­ing brain and watch­ing those huge­ly com­pli­cat­ed cir­cuits at work. As point­ed out in the jour­nal Sci­ence, this is rather like try­ing to learn elec­tron­ics by play­ing around with the inside of a computer.

The approach of the Cal Tech sci­en­tists is more like those teach-your­self-elec­tron­ics kits you can buy for your kids at Radio Shack. Start with 16 rat brain cells, let them estab­lish con­nec­tions, then watch them do their thing.

This new research is inter­est­ing on two fronts: It is an ele­gant new exper­i­men­tal approach to under­stand­ing how brains work; and it sug­gests ways of build­ing a new class of hybrid elec­tron­ic devices, part­ly ani­mate, part­ly inanimate.

Although prac­ti­cal suc­cess is far in the future, it is not hard to imag­ine hybrid devices con­sist­ing of mil­lions, or even bil­lions, of liv­ing rat brain cells, anchored in sil­i­con cir­cuits, and taught by repeat­ed stim­u­la­tion to per­form use­ful tasks.

At which point, the lan­guage of com­put­ers will expand to include hard­ware, soft­ware, and ratware.

Told about these exper­i­ments and spec­u­la­tions, one of my stu­dents said: “All rats can do is run through mazes and find cheese. Who wants a com­put­er that finds cheese?”

Anoth­er quipped: “We’ll keep the cir­cuits alive by slip­ping Kraft sin­gles into the flop­py disk drive.”

As we dis­cussed a neu­rochip future, oth­er wacky sce­nar­ios passed through our minds.

Will hybrid com­put­ers made of liv­ing brains cells be absent-mind­ed? (“Now where did I store that pay­roll data.”)

Will neu­rochips age, with con­se­quent loss of mem­o­ry? (“I remem­ber the graph­ic, but I for­get the filename.”)

Will rat-brain com­put­ers resent being con­trolled by a “mouse?”

Will they mul­ti­ply like rats?

Will Bill Gates chase rat­ware off the mar­ket by intro­duc­ing a line of machines incor­po­rat­ing cat brain cells?

Or will the Microsoft cat­ware machines tend to laze about in the sun and stub­born­ly refuse to fol­low instructions?

And what about the pos­si­bil­i­ty of a more lit­er­al kind of com­put­er virus, some­thing sim­i­lar to the rat-borne bacil­li that caused the Black Death?

If neu­rochips are best suit­ed for com­put­ers with sin­gle, ded­i­cat­ed tasks, then the cells used on the chips should be cho­sen to take advan­tage of evo­lu­tion­ary refine­ment for each spe­cif­ic task.

For exam­ple, dog brain cells might be most effec­tive­ly used in chips for robots that fetch slip­pers. Unfor­tu­nate­ly, the same robots might chew up the morn­ing paper.

Human brain cells would clear­ly be best for chess-play­ing neu­ro­com­put­ers, although the machines might be inclined to throw tantrums and stomp off the stage when defeat­ed by more con­ven­tion­al all-sil­i­con machines.

One thing is cer­tain: As we enter the next cen­tu­ry, the dis­tinc­tion between liv­ing and non-liv­ing will become increas­ing­ly blurred. The response of my stu­dents to the rat­ware exper­i­ments is encour­ag­ing; we will need all the humor we can get if we are to live in har­mo­ny with hybrid com­put­ers that need to be both plugged in and fed.