Beyond the Hype (again)
by Pat Rooney
While in San Francisco recently, I noticed billboards advertising generative-AI products. And while that information itself is not a bad thing, I couldn’t help but think back to the crypto craze and all the advertising schemes surrounding that nascent product. Generative-AI is indeed the next big thing. As The Economist pointed out recently, ChatGPT, an app launched early in November of 2022, reached 100 million users faster than ever. As The Economist pointed out recently, “Rivals jumped into gear (Google launched its own chatbot) and venture capitalists suddenly found a new darling, pouring over 40 million into the product in the first half of 2023.”
School administration and education officials, at once both fearful and curious, are starting to embrace generative-AI as a means of learning. The fearful fret about the potential damage done to critical and logical thinking, while the curious want to see what generative-AI can really do. The Los Angeles Times reported that one teacher uses chatboxes to show that “anything a chatbox does they, the students, can do better.” Kind of a keep up with the Jones’ sort of thing.
All of this generative-AI development is still in its very early stages and everyone is trying to find applications that not only work but are affordable, as well. Like any new technology, once the hype dies down the trick is to develop the profitability of these products. As the Wall Street Journal notes, “generative artificial intelligence tools are unproved and expensive to operate, requiring muscular servers, expensive chips, and consuming lots of power.” Chris Young, head of Microsoft’s Corporate Strategy, states that “we are clearly in a place where we now have to translate the excitement into true adoption.” As the hype fades and the costs escalate, companies will be working diligently to determine what can be done profitability and what cannot.
Generative-AI is here to stay, clearly, but how and the best way to apply it is still being sorted out. When queried about the billboards and the promotion of generative-AI, the investment firm representatives smirked a bit and looked at each other. The hype is over indeed.
Good article, Pat. I agree that some people with deep pockets and many people of modest means may jump in and get over their head sooner or later.
I see tremendous value in working with empirical data, as in medicine, where zillions of pieces of data can be compared among millions of patients to discover unsuspected causes of disease. For example, maybe people of a specific ethnicity who live in a specific altitude with a specific temperature range and who follow a specific diet are 95% certain to experience severe arthritis between the age of 55 and 60 – – improbable, but possible, but I think AI might be able to figure out this analogous to this.
Another good example: Take the yield of every wheat crop, corn crop, rice crop, etc. and the associated soil data as well as the cost of every grain shipment by truck, barge, and ship, etc. to figure out to most efficiently feed the world population while providing profit to all involved in the supply chain.
But, when working with other types of “data” AI will be working with opinions, and since some opinions are dead wrong, although frequently stated as fact, it would seem AI might have a problem. Here’s one example:
I was able to interact with Thor Hyerdahl at a conference in Hawaii around 1969, when he was well know for his book “Kon-Tiki”. Many miss the point of his book about sailing across the pacific in a balsa-wood raft. They think he was trying to prove that Polynesians were descended from indigenous people in South America. He was not. What he was trying to do was determine the validity of academia’s basis for the belief that such a premise would be impossible. In academia, it was a well-known fact that SA natives had no watercraft capable of making such a trip and, therefore, “The South Americans could not have migrated to the South Pacific.”
Turns out the academics were all quoting each other in support of this belief, and the belief became “fact”. It further turns out that if anyone bothered to read the original source for this belief, they would have found it was from the notes of a long-ago explorer who asked the captain of a boat crossing a lake, “What are those rafts floating half-submerged?” The captain responded “Those a local native rafts made of balsa. They sink in about 3 days.”
The Kon-Tiki is, I believe, still floating and on display in Norway.
This is what frightens me about AI. I’m guessing that AI may not be given access to all the research it needs. My 21 years in t world of computers with IBM taught me “garbage in-garbage out”. If some of the “facts” the computer is fed are “garbage”, the result will be “garbage”.
Without doubt, people far more knowledgeable than me are looking at this problem, but I just don’t believe AI will be given access to the cobweb covered documents hidden in vaults in the Vatican, the top secret files of every government or personal letters written from witnesses of events. I fear AI will be treated to a heavy dose of what is printed by journalists who, in my experience are not always accurate. Think of the amount of data residing in the brains of the billions of people on this planet. Will that data ever be tapped? – not likely unless our entire world is ruled by a dictator and the technology to decode the human mind is available to “them”.
Looking back at what I’m writing, it may sound like I am referring to political information and disinformation, but that’s just a tiny piece of the problem, I think.
I’m not investing in AI unless I understand precisely what the investee is up to.