I should have stated my feelings a little better. I should have said that comments like those seem a little stale because we have been making them for years and none of us have the time to thoroughly research and review the claims and actual achievements that have been made. I do think that we need to develop more robust learning systems. I am trying to write a short page on my thoughts about something like this. One of the problems is that if we try to partition learning acquisition and retrieval algorithms on a more abstract level the partitions are immediately brittle because the distinctions become intangible at a higher abstract level where it has to be assumed further detail would have to be added to make them feasible. But if we had a higher abstraction to start with they would make testing them out much easier. Jim Bromer
On Thu, Feb 14, 2019 at 11:30 AM Linas Vepstas <[email protected]> wrote: > > > > On Thu, Feb 14, 2019 at 7:47 AM Jim Bromer <[email protected]> wrote: >> >> These comments by Linus seem a little stale. Have you ever made >> inquiries about your recollections about what IBM said Watson could >> do? Have you looked into the current progress of Watson with medical >> knowledge? > > > Indeed perhaps they are stale. I had assumed that if there was progress, I > would have heard about it with the same ferocity as the original round of > advertisements. > > I did not mean to imply that expert systems are wrong, or are bad. They seem > quite viable, now that there is more ram, disk and compute power. Algos can > dig deeper, explore more branches, find and resolve more inconsistencies. The > Jeopardy tournament clearly proved that. Historically, the systems were > incapable of "learning from experience" but tighter coupling to experiential > systems might solve that. Whether or not a suitcase fits in a trophy, or the > other way around depends on one's experience with trophies and suitcases; to > have experience with either, you have to be a human (with a need move items, > and a interest in trophy-granting activities (and a suitcase in a closet, and > experience with places that have closets or attics ...)) Common sense is > experiential; something you develop over decades of interacting with a > reactive environment. > > My critique of Doug Lenat is that perhaps he underestimated the magnitude of > the task,and pinned too much hope on "microtheories" to resolve deductive > inconsistencies. > > There is an open question: is it better to curate an existing knowledge-base, > slowly expanding it, or is it better to create a more robust learning system, > that will be able to rapidly acquire the needed knowledge once it's turned > on? I'm betting on the latter, not the former. > > --linas > > -- > cassette tapes - analog TV - film cameras - you > Artificial General Intelligence List / AGI / see discussions + participants + > delivery options Permalink ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/T9ccd0aac7d42f57b-Mb47b0a3c9e375fef2e198571 Delivery options: https://agi.topicbox.com/groups/agi/subscription
