top of page
Search
Writer's pictureDoug Weiss

Research

Artificial Intelligence, and in particular, ChatGPT, have been the subject of considerable attention and some consternation over the past year. We've all likely read or heard about AI to the point it is no longer novel or perhaps even that concerning despite frequent articles fretting about the imminent takeover of humanity by sentient software. But I am especially amused by the counter arguments, those that allege AI is nothing more than a card trick, the dog dancing on its hind legs, if you will. In support of this thesis, critics point out the flagrant errors and limitations of the current iterations.


Large Language Models, such as ChatGPT must be taught and the grist for that mill, it turns out, are enormous caches of content, and because of copyright and other legal limitations a significant amount of what 's freely available is public domain content--especially social media. It comes as no surprise then that along with a lot of useful and informed opinion, commentary and factual information ChatGPT and its cohorts are being fed more than their fair share of gossip, innuendo, misinformation, disinformation and plain old nonsense. We have met the enemy and he is us. No wonder AI gets it wrong all too often and betrays every bias, divergent opinion and hypocrisy known to the human race. It is just a mirror reflecting back what we've taught it.


So, whenever I read one of those dueling Twitter or Facebook feuds that asserts an absurd point of view, something that can and should be be easily refuted with a little effort I understand why it isn't. Someone did the research. You know you've heard it a hundred times as if the person claiming expertise to back up their decidedly ridiculous theory or claim honestly believes that a Google search is research, and a selective meander through the back alleys of the Internet is equivalent to a doctorate in the subject at hand.


One sterling example of this occurred recently when a scientist, Yann LeCun--the head of AI at Meta (parent of Facebook, Instagram etc.) a man who many regard as a signal contributor to the subject of machine learning was called out first by Elon Musk and subsequently by a certain headline worshipping politico for dismissing a development Musk claimed as science. LeCun's grave error it seems was a simple statement that if someone invents, creates or does something they claim as science it must be published. LeCun himself has published hundreds of peer reviewed scientific papers in his career, and that in part is why he is largely regarded as a highly accomplished scientist.


I don't want you to think this is just a silly online spat, it goes to the heart of the matter so kindly bear with me. In response to LeCun's admonition that if something isn't published it isn't science, Musk retorted that the statement was one of the dumbest things he had ever heard. Surely a well considered reply.


LeCun's response, which I am reproducing below, responded to the matter at hand so clearly that I want to share it with you. He said: " To qualify as Science, a piece of research must be correct and reproducible". He then went on to say that to be correct and reproducible it must be described in sufficient detail in a publication--which publication is published so that it can be checked by reviewers and the Science reproduced so that anyone can access that information. What he was describing in brief is the process by which almost every significant development in physics, medicine, mathematics and in short, Science, has been validated for centuries. Theories are just that--a belief, perhaps even an informed guess or opinion but they aren't Science unless they can be independently verified and reproduced.


Musk, who is famous for firing off terse dismissals of those with whom he disagrees,

was quickly echoed by a politico whose grasp of the scientific method elicited this gem on the subject of Climate Science: “You’re going to tell me that back in the ice age, how much taxes did people pay, and how many changes did governments make to melt the ice? The climate is going to continue to change.”


You see, that politician reflected a view held by many that science is just a bunch of made up stuff; attempts to explain something in terms that anyone with a sixth grade education knows cannot be true. Who needs facts, proofs, reproducibility? And therein is the challenge. Research isn't just something you do in between your morning cup of coffee and brushing your teeth. Research must be correct and reproducible.


The Internet has done a lot to help legitimate researchers, by providing a way to access what might be difficult to obtain information, by helping to track down sources and by allowing researchers to use computational tools to make certain assessments and aggregations of data among other things. But the Internet as a whole and very specifically Social Media are not research. So, the next time you see or read that someone did their research on a subject exercise the appropriate degree of caution. Roll your eyes, count to three and and give it the credence it is due.

8 views0 comments

Recent Posts

See All

The Remembrance of Things Past

In Proust's iconic novel, A la Recherche Du Temps Perdu , the taste of a madeleine is the triggering memory that recalls a childhood the...

Relics

On a recent trip to Europe, I was reminded again of the power and pervasive presence of the Roman Empire. For more than 1,000 years the...

Free

Whenever I see an offer that begins with the claim that something is free, my instant reaction is, sad to say, a cynical one. What's the...

Comments


Subscribe and we'll send you new posts every week

  • Facebook Social Icon
bottom of page