I actually agree, I just feel it’s the same hyperbolic argument every time a new technology makes something else (situationally) obsolete.
Knowing how live off the land was once common. It can still be useful. It is even fun for some (to learn, if not do). Yet… Here we are with a vast majority of our population devoid of those skills. It will keep happening.
This is complicated by the fact that to gain advantage (or just maintain relevance) in a complex and changing society, you need to be learning the new skills that leverage new technology. Yet systems become more complex, specialized roles are needed, and we simply can’t learn it all.
Enter the database of all information and a technology capable of quickly searching and extracting information with relevant context. Learn to use it to your advantage, or don’t…
I think its more, roles that didn’t need specialty skills are being removed by a very dumb tool that dumb people think is amazing. It seems amazing if you dont understand the thing you are asking it to do.
Living off the land is a great example. I grow ~10% of my food in a large garden, mostly tomatos, apples, pears, and peppers. Most people dont know how hard and specialized it is to do this. When we lost power for a week 2 years ago, we were totally fine. Provided food to our neighbors, no issues. Because of our time spent learning how to grow and preserve food.
I’d argue allowing the offloading of the simple specializations AI can handle makes an overall more useless person.
I have a small garden, know how to grow crops, and grow hydroponics as a hobby (I have off grid solar, battery back up, and could upscale operation if needed).
So I do get where you’re coming from.
I also have an offline information repository and off line local LLM that could help someone (with less skills than me) maintain/upscale or build a similar system, as well as provide many other skill sets I don’t have or am less competent in.
So the technology is useful to those who know how to use it.
(I have a strong science and medical skill set, and a hobby level skill set for network/computer science, electrical work, and gardening… So ymmv. Point being, I use all tools and skills available, and if something happens to me, the LLM is better than nothing for those that don’t, and can greatly accelerate learning new skills if used correctly, even a relatively stupid open weights model is like a super search engine that can translate language (and tech speak to normy).
It’s one of those things I know is hard to explain. Similar to the covid pandemic, I kept saying “I want to read the papers 10 years in the future”. I want to know what happens after the hype and actual science of the impact something has becomes available.
I’m extremely skeptical that a tool that offloads basic critical thought and brainstorming will be good for society as a whole. I use it sparingly, but it usually wastes my time. Except for summarizing things it is good at that.
I think it will collapse into normalcy in a few years, and it will be moderately useful. The stealing it does is also a major problem.
As a strange example, I asked Gemini a question similar to my published thesis. It stole directly from my these with no attributions.
How to utilize new tools for Society as a whole vs disciplined and intelligent individuals and groups are unfortunately different things. Also, the tech is very early, a great deal needs to be done to make it usable as “advertised” by the big tech hype bubble. However, the cutting edge of small local models is already where the large models were just a year ago, and getting more efficient on worse hardware. However, that is the wild wild West of LLMs, like the early Internet, it’s dangerous and confusing. It’s a cyber security nightmare, while LLMs are closing old wholes faster than ever, using LLMs is going to creat new risks.
I don’t have the answers to the larger social issues. All I can do is ensure I understand them myself and am prepared to use them.
The copyright and attribution issue is real, but it is a “1st world problem”. The accuracy of information is an engineering and computer science problem that will be solved or mitigated over time. The cyber security problem is actual more concerning to me the more I learn.
COVID is a good example, since I had a closer view of the reality than many others. I was an essential worker who was exposed early due the negligence of others brought about by the misinformation and mistrust spread early on before the vaccines were even an option to argue about. Even those working with the best of intent got things wrong, but most of what was put out by the real experts was correct, if incomplete and if anything overly cautious despite the narrative otherwise. We are effectively doomed if a real bio weapon is used now, the level of misinformation and resistance to reality is too high to stop it.
The regulation of LLMs is going to be a war of special interests. I don’t have much faith that uninformed dinosaurs will get it right with ears full of corporate money. It’s going to be a wild ride.
I actually agree, I just feel it’s the same hyperbolic argument every time a new technology makes something else (situationally) obsolete.
Knowing how live off the land was once common. It can still be useful. It is even fun for some (to learn, if not do). Yet… Here we are with a vast majority of our population devoid of those skills. It will keep happening.
This is complicated by the fact that to gain advantage (or just maintain relevance) in a complex and changing society, you need to be learning the new skills that leverage new technology. Yet systems become more complex, specialized roles are needed, and we simply can’t learn it all.
Enter the database of all information and a technology capable of quickly searching and extracting information with relevant context. Learn to use it to your advantage, or don’t…
I think its more, roles that didn’t need specialty skills are being removed by a very dumb tool that dumb people think is amazing. It seems amazing if you dont understand the thing you are asking it to do.
Living off the land is a great example. I grow ~10% of my food in a large garden, mostly tomatos, apples, pears, and peppers. Most people dont know how hard and specialized it is to do this. When we lost power for a week 2 years ago, we were totally fine. Provided food to our neighbors, no issues. Because of our time spent learning how to grow and preserve food.
I’d argue allowing the offloading of the simple specializations AI can handle makes an overall more useless person.
I have a small garden, know how to grow crops, and grow hydroponics as a hobby (I have off grid solar, battery back up, and could upscale operation if needed).
So I do get where you’re coming from.
I also have an offline information repository and off line local LLM that could help someone (with less skills than me) maintain/upscale or build a similar system, as well as provide many other skill sets I don’t have or am less competent in.
So the technology is useful to those who know how to use it.
(I have a strong science and medical skill set, and a hobby level skill set for network/computer science, electrical work, and gardening… So ymmv. Point being, I use all tools and skills available, and if something happens to me, the LLM is better than nothing for those that don’t, and can greatly accelerate learning new skills if used correctly, even a relatively stupid open weights model is like a super search engine that can translate language (and tech speak to normy).
It’s one of those things I know is hard to explain. Similar to the covid pandemic, I kept saying “I want to read the papers 10 years in the future”. I want to know what happens after the hype and actual science of the impact something has becomes available.
I’m extremely skeptical that a tool that offloads basic critical thought and brainstorming will be good for society as a whole. I use it sparingly, but it usually wastes my time. Except for summarizing things it is good at that.
I think it will collapse into normalcy in a few years, and it will be moderately useful. The stealing it does is also a major problem.
As a strange example, I asked Gemini a question similar to my published thesis. It stole directly from my these with no attributions.
Understandable.
How to utilize new tools for Society as a whole vs disciplined and intelligent individuals and groups are unfortunately different things. Also, the tech is very early, a great deal needs to be done to make it usable as “advertised” by the big tech hype bubble. However, the cutting edge of small local models is already where the large models were just a year ago, and getting more efficient on worse hardware. However, that is the wild wild West of LLMs, like the early Internet, it’s dangerous and confusing. It’s a cyber security nightmare, while LLMs are closing old wholes faster than ever, using LLMs is going to creat new risks.
I don’t have the answers to the larger social issues. All I can do is ensure I understand them myself and am prepared to use them.
The copyright and attribution issue is real, but it is a “1st world problem”. The accuracy of information is an engineering and computer science problem that will be solved or mitigated over time. The cyber security problem is actual more concerning to me the more I learn.
COVID is a good example, since I had a closer view of the reality than many others. I was an essential worker who was exposed early due the negligence of others brought about by the misinformation and mistrust spread early on before the vaccines were even an option to argue about. Even those working with the best of intent got things wrong, but most of what was put out by the real experts was correct, if incomplete and if anything overly cautious despite the narrative otherwise. We are effectively doomed if a real bio weapon is used now, the level of misinformation and resistance to reality is too high to stop it.
The regulation of LLMs is going to be a war of special interests. I don’t have much faith that uninformed dinosaurs will get it right with ears full of corporate money. It’s going to be a wild ride.