Computers learn to get personal

Originally published 30 May 1994

A young Russ­ian stu­dent came to my col­lege to study com­put­ers. We became friends. When he returned to Rus­sia we exchanged letters.

He nev­er got used to the fact that my let­ters were typed on a com­put­er and print­ed on a laser. In Rus­sia, he wrote, it is unthink­able that a per­son­al let­ter would not be handwritten.

Do Rus­sians retain civ­i­liz­ing graces from times past? Or is it that they have so few per­son­al computers?

I sus­pect the lat­ter. Yet I con­cede that there is some­thing inher­ent­ly gra­cious about a hand­writ­ten letter.

So I took notice when I heard of a com­pa­ny in Ore­gon called Sig­na­ture Soft­ware that will sell you a key­board type-font for the Mac­in­tosh that looks like your hand­writ­ing. You sim­ply fill in a form by hand, repro­duc­ing a few dozen non­sense words and an alpha­bet, and Sig­na­ture will gen­er­ate a font that you can install on your computer.

To give the print­out the look of authen­tic long­hand, each char­ac­ter is tai­lored by the soft­ware to the char­ac­ters around it. The result is impres­sive. I have received a few let­ters in per­son­al fonts, and if I had­n’t heard about such things, I would have assumed they were handwritten.

OK, so the pho­ny hand­writ­ing is a bit of a cheat, but at the same time it adds a touch of human­i­ty to com­put­er com­mu­ni­ca­tion. If the medi­um is the mes­sage, as Mar­shall McLuhan said, then the mes­sage has just become a tad more personal.

There is progress at the oth­er end, too: Com­put­ers are learn­ing how to read longhand.

Apple’s hand-sized Per­son­al Dig­i­tal Assis­tant (PDA), called New­ton, has no key­board, only a glass screen upon which you write with a plas­tic sty­lus. The hand­writ­ten words dis­ap­pear from the screen, replaced by their type equivalent.

Well, sort of. New­ton’s hand­writ­ing recog­ni­tion skills are not yet per­fect. I wrote, “The quick brown fox jumped over the lazy dog.” The com­put­er inter­pret­ed this as “The well-known cox jumped over the lazy key.” I tried again, tak­ing more care with my script. This time New­ton came up with “The quiche brown fox jumped over the lazy dog.”

Not bad. Even human inter­preters have trou­ble with my scrawl. And the machine must cope with an almost infi­nite vari­ety of hand­writ­ing styles, both cur­sive and block. To be fair, I did not give New­ton time to learn, for learn it can, if you teach it your per­son­al script.

Com­put­er mag­a­zines have panned New­ton’s hand­writ­ing recog­ni­tion abil­i­ties, and those of a sim­i­lar sty­lus-dri­ven device by Tandy/Casio, called Zoomer. But Apple and Tandy/Casio deserve cred­it for pio­neer­ing a tech­nol­o­gy that will be impor­tant in the future.

Machines that respond to voice instruc­tions are also in the pipeline. Sev­er­al com­pa­nies are work­ing on voice-dri­ven per­son­al com­put­ers. With­in five years, many per­son­al com­put­ers will respond to both hand-writ­ten and voice com­mands. Key­boards will become an option­al accessory.

These new input tech­nolo­gies are a step towards giv­ing machines a human face. The big ques­tion is: Can we turn com­put­ers into humans before com­put­ers turn us into machines?

The chal­lenges faced by design­ers of long­hand and voice-dri­ven machines are for­mi­da­ble. Pat­tern recog­ni­tion by com­put­ers takes us to the heart of arti­fi­cial intel­li­gence. Con­sid­er, for exam­ple, the prob­lem of homo­phones, words that sound the same but have dif­fer­ent mean­ings (such as “sum” and “some”). A com­put­er must learn to inter­pret these words in their con­text, as does the human brain.

Apple’s New­ton has some­thing called Intel­li­gent Assis­tance that allows the com­put­er to antic­i­pate what the user wants it to do, anoth­er way of mim­ic­k­ing the human brain. Although the “assist,” as present­ly exe­cut­ed, is some­times less than “intel­li­gent,” the achieve­ment is impres­sive. With New­ton, the engi­neers at Apple are hop­ing to repeat the tri­umph of the Mac­in­tosh, a user-friend­ly machine that went a long way toward mak­ing com­put­ers behave the way we do, rather than the oth­er way around.

It is not hard to imag­ine per­son­al com­put­ers of the future that are as friend­ly and respon­sive as pets, ready at a spo­ken word to do our bid­ding. In many ways, this may seem a fright­en­ing future, but it far less fright­en­ing than a world in which humans become automa­tons at the bid­ding of machines.

Rus­sia is far away and the mails are slow. It can take a cou­ple of weeks for a let­ter to reach my friend. How­ev­er, the day is not far off when we will be con­nect­ed on the infor­ma­tion super­high­way. I’ll jot a greet­ing with a sty­lus on a glass screen, and the mes­sage will be instant­ly print­ed in my per­son­al script in Moscow.

In spite of my friend’s mis­giv­ings about com­put­er-gen­er­at­ed per­son­al mis­sives, we may be head­ing in the right direc­tion after all. We’ll have the best of both worlds: the con­ve­nience and speed of the com­put­er age, and a few civ­i­liz­ing touch­es from yesteryear.