Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:5 years? I know some top schools will stay around as novelty but overall seems like collapse is imminent. When graduating classes start experiencing increasing double digit unemployment? Seems like that is just around the corner.

There is zero chance of this. For one thing, AI is no where close to ready for prime time. Second, you still need humans to develop content, review fact and oterhwise be creative. Three AI can do things that humans can in the physical world. The fact is, yes, there are tasks that AI can complete, but only 80% of the way there and it requires humans to get the other 20%, regardless of task.

Yes but when AI is completing 80% of the tasks and only requiring deminimus effort from humans to complete it, the jobs for humans will start drying out. And as humans "fix" or "complete" the task to 100%, AI is learning what the human is doing (when companies start using employees to feed training data to AI as many companies have started to do) and iteratively getting better at completing the task to 100%. And, I don't think you're 5 years ago. There's an exponential growth curve happening in the world of AI. Think about the last year and how much more functionality and utility we got from the top two models out there.

They still mostly suck.

No they actually dont. If you've been using the models over the last two years, you know exactly how good they gotten. They're not perfect (and I'm not sure they can ever reach perfection given what LLMs are); however, to say they "suck," just means you haven't really been using the models other than to fix some emails.

The other problem is that most people don't know how to actually write proper prompts. They get garbled information or the model hallucinates. There's all kinds of issues when your prompting is actually the problem. People think asking Claude questions as if you were just chatting with the bot is the way to get a fully fledged answer. It's not. But no one is actually taking the time to learn how to prompt (especially non-coders).

Seems like it won’t be very disruptive then if the problem is that “most people” simply can’t use it.

And/or that most people simply don't need gen AI enough to learn. Which is the whole problem--we are being told that we have to use it, that we're going to be left behind if we don't...but there is literally nothing in my life that I need it for.

You can't threaten people into using it and then tell them, "oh actually it won't work the way it's hyped to unless you invest lots of time into learning how to use it and also you're stupid if you expect it to be easy to use" and expect them to embrace it.

And also...I know how to use search to find what I need. I like to vet sources myself. I am on top of my inbox. I want to write my own emails; I am not looking for shortcuts to my humanity. Gen AI has no value for me, and then you tell me I have to invest time learning how to use it properly? Please.

5 years? They transfer the DCUM patients to different psych lockdowns every 18 months. Definitely a newer nutjob.

Anonymous wrote:OP is the same poster who asked 5 years ago if crypto would cause all banking to collapse; and if self-driving cars would cause the auto industry to collapse. How's that going?

+1. OP is the same person who asked 5 years ago if the economy and housing in DMV will collapse due to the pandemic.

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:5 years? I know some top schools will stay around as novelty but overall seems like collapse is imminent. When graduating classes start experiencing increasing double digit unemployment? Seems like that is just around the corner.

There is zero chance of this. For one thing, AI is no where close to ready for prime time. Second, you still need humans to develop content, review fact and oterhwise be creative. Three AI can do things that humans can in the physical world. The fact is, yes, there are tasks that AI can complete, but only 80% of the way there and it requires humans to get the other 20%, regardless of task.

Yes but when AI is completing 80% of the tasks and only requiring deminimus effort from humans to complete it, the jobs for humans will start drying out. And as humans "fix" or "complete" the task to 100%, AI is learning what the human is doing (when companies start using employees to feed training data to AI as many companies have started to do) and iteratively getting better at completing the task to 100%. And, I don't think you're 5 years ago. There's an exponential growth curve happening in the world of AI. Think about the last year and how much more functionality and utility we got from the top two models out there.

They still mostly suck.

No they actually dont. If you've been using the models over the last two years, you know exactly how good they gotten. They're not perfect (and I'm not sure they can ever reach perfection given what LLMs are); however, to say they "suck," just means you haven't really been using the models other than to fix some emails.

The other problem is that most people don't know how to actually write proper prompts. They get garbled information or the model hallucinates. There's all kinds of issues when your prompting is actually the problem. People think asking Claude questions as if you were just chatting with the bot is the way to get a fully fledged answer. It's not. But no one is actually taking the time to learn how to prompt (especially non-coders).

Seems like it won’t be very disruptive then if the problem is that “most people” simply can’t use it.

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:Better question seems to be how long until the AI bubble bursts.

+1
I honestly believe AI has been waaaaaaaaay overhyped.

For the people making these kinds of statements, you're really not using AI to its fullest capacity and it would be nice if you could actually understand that. For those who are heavily utilizing tokens and are required to do so by their jobs, no one on that end is saying it's "waaaaaaaaaay overhyped."

Or maybe these people just have broader perspective than their own individual job. Just because it is relevant for you doesn’t mean it is relevant for everyone.

Anonymous wrote:

Anonymous wrote:What kind of country do we want? All uneducated people who don’t go to college, for what? What is even the point of human life if we don’t have work to support ourselves?

Well, right now the uneducated people are going to college and graduating just as uneducated as when they started, having cheated to get over a standards bar that has been lowered to the bottom.

Yes… we are actively becoming an illiterate society where the college-educated can’t read. My own DC, privileged with private education his entire life… will go to some college and earn a diploma with 6th grade reading skills. He was never going to be a scholar but the iPad middle school “education” and laptop high school have truncated his education. Effectively stopped at middle school. Poor thing. The future for the rich and privileged will be no-tech enclaves for their kids to truly be educated through at least 12th grade.

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:

Anonymous wrote:5 years? I know some top schools will stay around as novelty but overall seems like collapse is imminent. When graduating classes start experiencing increasing double digit unemployment? Seems like that is just around the corner.

There is zero chance of this. For one thing, AI is no where close to ready for prime time. Second, you still need humans to develop content, review fact and oterhwise be creative. Three AI can do things that humans can in the physical world. The fact is, yes, there are tasks that AI can complete, but only 80% of the way there and it requires humans to get the other 20%, regardless of task.

Yes but when AI is completing 80% of the tasks and only requiring deminimus effort from humans to complete it, the jobs for humans will start drying out. And as humans "fix" or "complete" the task to 100%, AI is learning what the human is doing (when companies start using employees to feed training data to AI as many companies have started to do) and iteratively getting better at completing the task to 100%. And, I don't think you're 5 years ago. There's an exponential growth curve happening in the world of AI. Think about the last year and how much more functionality and utility we got from the top two models out there.

They still mostly suck.

No they actually dont. If you've been using the models over the last two years, you know exactly how good they gotten. They're not perfect (and I'm not sure they can ever reach perfection given what LLMs are); however, to say they "suck," just means you haven't really been using the models other than to fix some emails.

The other problem is that most people don't know how to actually write proper prompts. They get garbled information or the model hallucinates. There's all kinds of issues when your prompting is actually the problem. People think asking Claude questions as if you were just chatting with the bot is the way to get a fully fledged answer. It's not. But no one is actually taking the time to learn how to prompt (especially non-coders).

Or there are those people who think saying "You are an XXXXXXXX expert..." and go on right to the question think that is a great way to prompt and unfortunately that initial statement does nothing much unless it's followed by lines and lines of specific instructions.