Thread 82165645

I am trying to get through my Computer Science degree and if that explodes in my face, trucking is the way to go.

It's starting to feel like ChatGPT is the real student here. I'm a good codemonkey but by the time I graduate, AI will be more economically feasible than offshoring work.

It'll be many years until trucking starts to feel the pain of AI replacing jobs. Plus, I like driving, listening to music, podcasts, audiobooks, and I like feeling like a nomad without a home.

>>82165645 (OP)

how many years do you think until ai takes trucking jobs?

my guess is like 4 years

but my guess about agi is like 10 years anyway so who cares, what about you?

>>82165698
It'll take 15 years minimum for the trucking industry to start to feel the real effects of AI.
There's a big difference between self driving non-commercial cars and self driving trucks. At least here in the USA, trucks are the veins of the economy. All goods transported throughout the states are by truck.
That is to say, the importance of the trucking industry cannot be understated. AI would have to be really good at driving for the corporations and federal government to place the nations economic lifeline on it.

And here's the thing. AGI is only defined as having an AI that outpreforms humans at most economically valuable work. The bar is relatively low in retrospect. The most employed occupations in the USA are cashiers, secretaries, office clerks, retail sales people, etc. I envision funding for AI corporations to plummet as soon as they produce an AGI by their standards.

REAL AGI will be produced when we are in WW3. Wars always accelerate techonological advancement, and they are almost always applicable in the domestic, civilian context.

>>82165864

i mean it depends on your definition but in my opinion, agi will be achieved when the LLMs get better at programming than the best humans... i mean even just being close is good enough to self iterate and improve models. We don't know all the little tricks openai and competitors are using to improve at the benchmarks but i imagine a lot more can be achieved without relying on increasing compute like the improvements for the past few years

>>82165645 (OP)
>AI will be more economically feasible than offshoring work.
not really gpus are really expensive and most were disappointed by the release of GPT-5, there will be some improvements but not like they were between GPT-2/GPT-3 or even GPT-3/GPT-4

>>82166104

AGI is only defined as having an AI that outpreforms humans at most economically valuable work
/=

>AI will be more economically feasible than offshoring work.

>>82166123
even optimized ASICS are more expensive than some guy from belarus you could hire to do your dev work

>>82165645 (OP)
>Computer Science
You got SOME time AI is kind shit right now. But I'd hurry it up tho, you might have 5 years at most to make some cash money.
>>82166104
Jeets are paid 1$ an hour to code which means 'computer science' jobs are toast by the end of the decade.
>>82165864
>Trucker
Is getting taken over corpos based in Georgia that pay immigrants $5.15 an hour.
Conclusion:
You're pretty much fucked, coding is the new factory job its going extinct. Your ONLY options are gay prostitute or military. Enjoy AIDS &/or dying for Israel.

>>82166057
You're right in saying compute /=/ better AI, inherently. But compute power is the contsraints of AI to achieve AGI. WHEN AGI is achieved, then maybe compute power can scale back. But in the mean time, I don't see a world where we can just find the most efficient method with the least compute power that can achieve AGI. Maybe I'm wrong, all it can take is a supergenius being born if you look back in history, but I just don't see it.

AGI to us is a lot different than what AI companies (and through extension the federal government) view it as. This is the biggest division that sets the barebone to when AI will stop being packed in the ass with funding. There's no such thing as a product, innovation, or industry that can sustain significant funding, like what we see in AI, forever.

When AI replaces the aforementioned jobs and industries, and meets the definition set by their makers and government, it will be set back a ton. A lot less money, compute power, and time spent for R&D and it'll transition into sustaining it and expanding what they have on a larger scale.

I personally think the catalyst for true AGI will be when WW3 happens. But that's just me.

>>82166104
I still think conceptually my reasoning stands. But GPT 5 is AMAZING at programming and coding, in contrast with GPT 4.

>>82166123
>>82166133
I don't mean that when the definition of AGI is set it will automatically become more economically feasible, but it will be a lot more profitable once sustainability of the AGI is established and it becomes too big to collapse. On the individual businessman or whathaveyou, it will be more economically feasible. On the AI companies, all they have to do is to sustain it.

>>82166133

sorry i guess that wasn't clear

>AGI is only defined as having an AI that outpreforms humans at most economically valuable work

!=

>AI will be more economically feasible than offshoring work.

>>82166216
>Jeets are paid 1$ an hour to code
Does it matter when the avg IQ in india is about 75?
>which means 'computer science' jobs are toast by the end of the decade
There will be plenty if jobs to fix the retarded shit they did, it won't be pretty but it will pay well

>>82166216
Immigrants have always been teeming in the trucking industry this isn't something new. The salaries are still amazing.

>>82166227
>in contrast with GPT 4.
By what? 10%?
>>82166231
I still don't get your point

>>82166280
A 10% increase in accuracy is still pretty good. But no, it's a bit more than 10%

>>82166339
>A 10% increase in accuracy is still pretty good
And now compare that to the other jobs where one went from schizobabble to normal conversations (2 to 3) and then form normal conversation to something that gives good answers and has limited reasoning (3 to 4). The most important thing is that the curve has flattened and people the corps/research teams have started to chase small percentage increases. Also benchmarks are useless when everyone optimizes for them but that's another problem