Thursday, January 5, 2023

Research and learning

TL;DR --  A machine might learn, can it research? People do both. Comparisons of this nature will be even more commonplace as we see ML hit new milestones, so to speak. The question is still open depending upon many factors, of which one is knowing what is going on. That is one thing to address; bring the information out so that general discussion can ensue. On the other hand, people forget more than does the machine? Not really. But, we look at an important figure as an example of our future work. Charles Thomson was the Secretary of the Continental Congress for years. Knew all of the players. He was childless, so forgotten in some senses. Who was his family? We ask since collateral relationships have included new entrants to the country over time, ceaselessly. 


It's a strange time that we live in. There is an article in the latest ACM Communications (The End of Programming) by someone pushing ML which is now considered the basis for AI which I see as AIn't for good reason. In fact, this current thrust of technology needs attention with more urgency going forward than many might imagine. Hence our interest arose, somewhat motivated by an earlier example (Gairdner Foundation). 

The gist of the ACM article is that we see the progress. Say, in games from stumbling at chess a few decades ago to the recent collective of wins (and draws). Somewhat, there were examples of brute force bullying. Too, there was the conquering of Go by not having any limits such as things thought of as normal which were learned from human. In other words, the machine made choices that people might have ignored because they were considered silly. But, realize this, folks, please: the thing did this with huge iterations where the algorithm played against itself in some type of metrical scheme that would be equivalent to 100s of human lives (or literally, so many games to go over a lifetime; we can get specific). Too, in doing so, the system pulled loads and loads of electric power off of the grid while running through an expensive collection of computers (networked, of course) almost non-stop for a period of time. 

It's like the chess, Jeopardy and other wins. What we saw was one human with talent against a collective of hardware/software plus the handlers keeping the thing going. So, fairness needs to be discussed, always. Too, fragility is latent far beyond what humans might have. One hope is to get the thingees to be able to move the use of their learning from one situation to another. This is difficult due to the close mapping of learning to the associated data used for training. There are lots of things to discuss. 

Yes, we can be prepared for this. Can it actually be a non-profit's goal to help hone the dynamics or even to help handle truth issues? Why not? 

Now, another thing is that the computer is being used  by humans in unexpected amounts (scale is a watchword) and ways (it's more than ethics; humans do have rights; being creative in reasonable senses is one of them - AIn't has not seen anything yet). So, take looking at families and history. Lots of confusing information is around and about. One would hope that not having provenance shown with regard to sources would inhibit reproduction of some bit of information or that we would discuss the issues and work to hone that process. Are those matters getting worse?

At the same time, we see that genealogy is for those with offspring. Whereas, we have used Joseph and Ann Gardner as an example of people to remember. They had no offspring. This has been a common thing all the way through our history. We touch upon one guy below. Not long ago, we did a post on Lorenzo and Peggy. And, have mentioned others in posts here and there (gather these). So, one might expect that the 400th will be like the 300th and motivate people to look at their past. Too, one would expect lots of this research to be presented for our use, say as hints to foster more work. 

There is another thing to mention out of many left on the table. When we were looking at the Gardiner barque, Bostonian, that wrecked after an interesting bit of voyaging, we were interested in finding out the New England connections. But, in doing so, we found conflicting stories. Which to believe? We have on our plate to go back and look. 

As I work, I always check even if I don't list references. We will get to that. Right now, consider all of this material as a large table of content with varying amounts of comments to guide future work. In some cases, we use pointers to other material in our posts but do need to improve our bibliography. The one that Wikipedia did for Charles Sanders Peirce is a great example. BTW, CSP was childless and on our list. Back to Gardiner OR. Gosh, it was four years ago that we pulled all sorts of information into a post (The Gardiner that was) with the purpose of clarifying some of the issues. Tracking down the information was facilitated by the technology of today. By now, we would need to do a sweep again to gather later material. 

Oh yes, back to the first paragraph. The approach is to use data and model so as to find associations using mathematical means. One way to look at it is that the system (to be defined) maps connections which can be a fairly large set. To get an idea, connections relate to what we might think of parameters (knobs) which for humans usually are small. Of course, a TV remote with its partner that handles other media can get tricky to handle. Wrong push and the TV won't go on, sort of thing. Well, hold on, as now, one approach that has digested all of digital stuff (almost) - or ate a library - came up with 175M parameters (something like that, again - we're being purposely fuzzy here as that whole bit of notions will come back into favor - Zadeh, if you must know). And, one thingee (not naming names) can write text that looks good. The old thing of mismatched English (same for other languages) is gone. But, to me, the stuff reads like the double-talk of political realms. Which is fine. 

The issue is that anything that is artificially enhanced ought to fall under truth in advertising or under the guise of responsible press. Like, at one time it was considered okay for some simulated (pre-recorded) bit of TV to be announced as 'live' which was unsettling. That sort of thing improved; now, we're back to the days of not knowing (and being snowed by falsity - to be discussed). 


That was a brief preliminary statement. To the purpose of the post. This was seen in an article in an Iowa paper (on-line, of course). It was about someone related (supposedly) to our subject that had the thought to help bring him to the attention of interested parties during these time of remembering the events of 250 years ago.  

Charles Thomson (Wikipedia, WikiTree) came here with his father and brothers from Ireland where he had been born. His father died before (actually, almost right on) the arrival. At Baltimore, Charles was taken in and worked. His brothers were taken elsewhere. Eventually, he got through law school. And, when the turmoil of the split with the Brits struck, he got involved. This Harvard article does the best job that I have seen of looking at his work: Course of Human Events blog - November Highlight: Charles Thomson. The brothers? We're looking for them; at least one has multiple claimants for membership to D.A.R. However, even for those with progeny, we have see differences of opinion about who begot whom. 

Too, Charles has many who claim to be related to him, as well. 

This cursory introduction touched briefly upon a subject that will come up again and again while we cover the many aspects of intelligence and its future. Back to beginning of this post and the ACM article? The author had been an instructor of Harvard. Last year, we looked at their summary position on AIn't. It is nice to see the positions being taken and described in order to have a more full debate.  

BTW, that lament in the quote reminds me of the Old Planters Society's attitude from 120+ years ago about the injustice involved. 

Remarks: Modified: 01/08/2023

01/06/2023 -- Clean up and add pointers. Do a post for Charles

01/08/2023 -- We had another icon (Charles Thomson) come up for our 250th (1774, Continental Congress) and 400th (Cape Ann as one of the first of the capitalization attempts of history). 

No comments:

Post a Comment