I don't think we've come anywhere close to realizing the benefits of machine learning (including LLMs, etc.) but we may be reaching some saturation as to the marketing of "AI" into everything. I'm not sure you can buy a low end laptop without it claiming to have AI baked in.

And with the adoption of increasing the underlying models that form the basis for the GPT-type algorithms comes the obvious loss of privacy. Every platform, every computer, every app that wants to suck your private information in to improve their models are, without your explicit opt-in, violating your privacy.

Which is just for language modelling, and which happens to have become complex enough now that it looks smart because it can communicate close to human level by imitating human-produced language and images (which is another form of communication if you think about it). But that's all it does. All the predictions that AI is now about to take over anything that doesn't have to do with language is pure speculation, and if people are investing in it just because GPT and AI generated images look so cool now then they are about to lose their money.

Reasoning at all.

We have been given no idea if you agree with the article or not.

Nor have you stated your own outlook.

It is simple to run down assertions when you don't explain what your objections are, what your own views are, what you think are perhaps different facts, and no explanation of what you might think flaws in the reasoning are.

So simple even a 1990s AI program could do it.

I have occasionally posted about my skepticism about LLM and where it will/ will not go. I would like to note that you have not indicated whether you have glanced at, or read the article.

I tend to make my points without beating around the bush. It is rare that you will see me type 2 actual graphs here, lol.

AI LLM technology has known issues, like the hallucination issue, and other anomalies. And based on expert opinions I have read, these issues can not be resolved using current practices for optimizing training. Another "breakthrough" is going to be required. I think you acknowledge that.

Here's the thing, the people behind the "AI Hype", are avoiding the subject, or glossing over it. It is my opinion they are leading us into an ugly place. A place where power usage in this country increases drastically and cancels out the advances we have made with clean energy. (Have you read about Altman saying we need a power technology breakthough too?) And where misinformation is greatly amplified. Other issues, like accuracy, will keep the technology from accomplishing many of the stated goals of the AI community.

I don't think the current AI LLM tech is going to prove very useful, certainly not where accuracy is required. I have a fair understanding of general computer programming, and some of its challenges. That includes the use of databases, and statistics. I am a self taught developer for small to medium size db applications. If we actually analyze what an LLM is doing, versus, what it would need to do for a breakthrough, it does seem, that it would be a different beast. So maybe like you say, the next AI spring. Who knows, maybe never.

So if it is never, or if it is 10 years, don't you see that we are going to waste natural resources, and possibly create more problems than we solve with this generation of technology?

I think they will be downsized at the same time as smart phones and cars are upsizing in compute power. If we are directing the device to search for a movie that we remember a scene from we don't expect it to read our minds. Rather, like a human assistant or a friend, we will have a back and forth with it. It will be perhaps a little more prone to mistake words and have to ask for clarification more often compared to a server base LLM.

Another class of applications of ANN tech are one's that learn simple patterns where perfection is not expected. If you are driving your car (not self-driving because you don't trust that, say) and you make a turn to go home, the car might pipe up and say "This is Thursday, don't you usually pick up something at the supermarket for, like, date night with your wife?" It may have picked up clues from the pattern of many Thursdays. Maybe the third Thursdays are something else and the car hasn't noticed that pattern yet. No big deal between you and the car.

It seems that the energy usage is of some concern, but not as much as feared. I think the blockchain systems, especially for crypto mining, are a much bigger concern. https://www.theverge.com/2023/10/10/23911059/ai-climate-impact-google-openai-chatgpt-energy

The statements Altman made, I reference one about needing a power breakthrough, are more recent, as are many other articles I have seen. Don't ask me to hunt them down rt now.