Re: Smart Interfaces?

Dr. Ernest N. Prabhakar (ernest@drernie.scruz.net)
Sat, 14 Feb 98 13:48:30 -0800


Hi David,

Nice to meet you.

> One of the main
> things I find backwards about computers is that software is generally
> tool or process-centered, rather than goal-centered. I've been thinking
> about how that might be changed, and it would involve some AI planning
> techniques.
> I'm wondering whether Apple has
> any interest in this kind of thing when they have so many more pressing
> issues at the moment. And that's why I wanted to ask your advice.
> Maybe you could suggest other places where I could get paid for working
> out these ideas?

To be honest, I do fear Apple is focused on more incremental
improvements rather than cutting-edge change of the sort you're
interested in. I must admit I am not very in touch with the broader
research community. I will, however, forward your message to my comrade
Rohit Khare, via the FoRK virtual community. He or one of his CS friends
may have some better suggetions.

Good luck, and God bless,

-- Ernie P.

You wrote:
> Dr. Ernie! I enjoyed your home page and your 'Good News for Modern
> Nerds.' I've read your name in the next newsgroups, and I know you're
> a fellow Christian, so I thought maybe you would be a good person to
> ask for some honest advice.
>
> I'm finishing a Ph.D. in AI, but not sure where that's going to take me.
> But I've become quite opinionated about what's wrong with computers, and
> I'm wondering if I ought to apply for a job at Apple. One of the main
> things I find backwards about computers is that software is generally
> tool or process-centered, rather than goal-centered. I've been thinking
> about how that might be changed, and it would involve some AI planning
> techniques. This is not connected or even related to my thesis work,
> but I still am really gung-ho about it. I'm wondering whether Apple has
> any interest in this kind of thing when they have so many more pressing
> issues at the moment. And that's why I wanted to ask your advice.
> Maybe you could suggest other places where I could get paid for working
> out these ideas?
>
> Bill Gates thinks that future computer software should have 3D graphics
> and speech recognition. Others seem to think that we have to create
> interface tools which are more similar to objects we use in the
> real world. But I think these approaches miss the real problem with
> computers, which is this: If they're so smart, why do *we* have to
> understand *them*?
>
> For example, when I used Photoshop last year to try to create a jpg
> file for my home page, I had to decide what tools to use, what settings
> to select, what kinds of compression to use, what memory management.
> I was pretty clueless. Not only did I not have access to a manual (it
> was a college computer lab), but I felt I shouldn't have to study to do
> something simple like this. After all, I'm getting a Ph.D. in computer
> science, and am an expert in UNIX, very comfortable with Macs, and these
> days I play with NEXTSTEP/OpenStep every day. I ended up playing with
> the system for 2 hours, getting help from the lab consultant, -- and
> getting a jpg file which covered the whole screen when I opened it up
> on my NeXTstation at home.
>
> Instead of specifying the process, I want to just specify the result,
> and have the computer work it out. For example, I'd say that I want
> a .jpg file. The computer would give me one! I could inspect it,
> and say, yes, but make it so the size is only about 50k. And reduce
> the size to 3 x 5. Probably, when I use the word "size" the computer
> would ask me if I mean the size of the image or the size of the file.
> And I would just type in such information as it needed to disambiguate
> my request. Basically, the software would know what operators it has,
> what their inputs are, and what their effects or results are. It could
> use something like Newell and Simon's GPS (means-ends analysis) to chain
> together a path between the inputs and desired outputs. This might be
> feasible because the number of things one can do with a collection of
> computer programs circomscribes a narrow knowledge domain, I think.
>
> My feeling is that the computer should be a lot like a butler. I think
> the ultimate computer help is having another person who knows how to do
> what I want to do. I just say, "Frank, print out the animal file."
> Frank says, "Oh, the one you were just editing?" I say, "Yeah."
> Franks says, "Do you want 1 column or 2?" And I say, "1." And Frank does
> what has to be done. I think this is the way computers should operate.
> (Or maybe I just watch too much Star Trek. . .) But I wonder if this is
> a marketable direction, or if Joe computer user would care.
>
> Got any advice?
>
> Cheers,
>
> David Finton
> finton@cs.wisc.edu
> <http://www.cs.wisc.edu/~finton