“Given that ChatGPT and other forms of generative AI create their output by synthesizing what they find on the internet”
Wow that is almost, but not completely, the opposite of what it does. And given its current form, it cannot really replace much (and we don’t really know how it will look in the future, or if it will be much better than how it is now).
Looking at self driving cars, how much money was thrown at the problem, and how far are we still from looking at a fully self driving car, I would say AI replacing jobs is 90% marketing and 10% substance (there are fields where it is becoming very effective, like for example machine translation).
I agree with your opinion that all this companies are illegally abusing the content freely available on the internet to train this models.
What annoyed me in the article, is that they are talking about “what jobs AI will steal” and immediately gave it abilities that it does not currently have (and I’ve seen it a lot of this on fear mongering type of articles). It reminded me on all the articles “truck drivers will be out of job in 3 months”.
It does not find content on the Internet (Gemini, copilot, … Is trying to give it that functionality, with hilarious results). It does not act in any way.
If you want to put it in simple terms, the current AI is a tool that reads a lot of content, and when you ask it something it gives you an answer trying to recall something and forming a coherent answer (hence all the hallucinations). It is extremely good at forming coherent answers, quite bad at giving correct answers and absolutely incapable of haveing other types of functionality (though they are trying hard to create new multi-agent tools that on paper are capable of independently searching for information, or using other tools to get the answer it needs, but it still fails a lot of times).
I have no issue with what you said. I am just not sure if you are misunderstanding what they mean with “find on the internet” or if I do. But that is pointless to argue about.
how far are we still from looking at a fully self driving car
While on one hand, pretty far, on the other I don’t see why they’re not common. We’ve had a few geo-fenced pilots of self-driving cars with mixed results, but the money is in trucking. I would have expected there to be trucking pilots, such as on fixed routes, in convoy, or even remotely piloted within distribution centers.
Good points. There have been tests on self driving trucks, but not much more. My opinion is that the tools are not mature enough, and the industry is not willing to risk putting trucks on the road that may get stuck in the middle of the trip, because there is a roadblock and it cannot circumvent it, or that it goes on big detours because it somehow sees non-existing roadblocks.
Also there is still a problem of liability. If a truck fails to give way to an ambulance or a firefighter truck, or if it gets in an accident, who is responsible? The manufacturer in theory, unless they waive responsibility to the owner of the truck, and in that case what company would risk their face and money on a technology that has not proven itself?
All in all, at the moment I see a lot of reasons to doubt the technology, and few reasons to embrace it, unless it becomes trustworthy enough that it is economically viable.
Ps. Putting trucks on a fixed route, in a convoy, feels a lot like re-inventing the train haha
The benefits of a convoy (or road train 😆 ): there’s still a driver to take the fall, but they only need to be in the lead truck. Or if the convoys are longer, take another page from actual trains and also pay for a guy in the caboose to make sure the train stays together
“Given that ChatGPT and other forms of generative AI create their output by synthesizing what they find on the internet”
Wow that is almost, but not completely, the opposite of what it does. And given its current form, it cannot really replace much (and we don’t really know how it will look in the future, or if it will be much better than how it is now).
Looking at self driving cars, how much money was thrown at the problem, and how far are we still from looking at a fully self driving car, I would say AI replacing jobs is 90% marketing and 10% substance (there are fields where it is becoming very effective, like for example machine translation).
I mean, isn’t that kinda what it does?
OpenAI scrape (illegally, in my humble opinion) the internet and trained chatgpt on it. That sounds to me what they are saying. Where is the issue?
I agree with your opinion that all this companies are illegally abusing the content freely available on the internet to train this models.
What annoyed me in the article, is that they are talking about “what jobs AI will steal” and immediately gave it abilities that it does not currently have (and I’ve seen it a lot of this on fear mongering type of articles). It reminded me on all the articles “truck drivers will be out of job in 3 months”.
It does not find content on the Internet (Gemini, copilot, … Is trying to give it that functionality, with hilarious results). It does not act in any way.
If you want to put it in simple terms, the current AI is a tool that reads a lot of content, and when you ask it something it gives you an answer trying to recall something and forming a coherent answer (hence all the hallucinations). It is extremely good at forming coherent answers, quite bad at giving correct answers and absolutely incapable of haveing other types of functionality (though they are trying hard to create new multi-agent tools that on paper are capable of independently searching for information, or using other tools to get the answer it needs, but it still fails a lot of times).
I have no issue with what you said. I am just not sure if you are misunderstanding what they mean with “find on the internet” or if I do. But that is pointless to argue about.
While on one hand, pretty far, on the other I don’t see why they’re not common. We’ve had a few geo-fenced pilots of self-driving cars with mixed results, but the money is in trucking. I would have expected there to be trucking pilots, such as on fixed routes, in convoy, or even remotely piloted within distribution centers.
Good points. There have been tests on self driving trucks, but not much more. My opinion is that the tools are not mature enough, and the industry is not willing to risk putting trucks on the road that may get stuck in the middle of the trip, because there is a roadblock and it cannot circumvent it, or that it goes on big detours because it somehow sees non-existing roadblocks.
Also there is still a problem of liability. If a truck fails to give way to an ambulance or a firefighter truck, or if it gets in an accident, who is responsible? The manufacturer in theory, unless they waive responsibility to the owner of the truck, and in that case what company would risk their face and money on a technology that has not proven itself?
All in all, at the moment I see a lot of reasons to doubt the technology, and few reasons to embrace it, unless it becomes trustworthy enough that it is economically viable.
Ps. Putting trucks on a fixed route, in a convoy, feels a lot like re-inventing the train haha
The benefits of a convoy (or road train 😆 ): there’s still a driver to take the fall, but they only need to be in the lead truck. Or if the convoys are longer, take another page from actual trains and also pay for a guy in the caboose to make sure the train stays together