Ask HN: How to live life before AGI | Hacker News Reader

Ask HN: How to live life before AGI

I feel like it's pointless to worry about the future if the future is highly speculative, but it seems more and more certain that AI will eclipse humans in every domain outright soonish (1-3 years).

As a consequence, it seems pointless to work at a job. The local optima are either hedonism-maxxing or capital-maxxing, any middle ground seems like a foolish enterprise when all that effort over marginal returns won't matter soon.

NOTE: If you disagree with this premise, please provide actual evidence that the trends in improvement in AI models will not continue, and will peak at Mythos level somehow. Do not just assume that things will remain the same despite that not having been the case for the last few years.

View on HN

I agree with Mo [1] and doubt AGI will ever be a real thing. People will probably market something as "AGI" just like people market big-data-LLM chat bots as "AI".

Unlike Mr. Deutsch I also believe animals are alive and can think. My cat is thinking right now.

There will be plumber robots though. They will break pipes and flood houses for a few iterations. Move fast and break the pipes. That might be a fun poll, how many here would allow a first generation plumber robot in their home and what might you hide from its view?

[Edit] Added a poll [2]

[1] - https://www.youtube.com/watch?v=VucjurQUHO8 [video][6 mins][language]

[2] - https://news.ycombinator.com/item?id=48302523

SO you believe that there will always be things humans can do that AI system or robots can't? What makes you so sure, what is the "critical gap" that no neural net could pass that only a human brain can?

I did not say that. I suggested that the first few generations of robots will have tickets updated with "It was not supposed to do that..." and tens of thousands of dollars of damages and real plumbers testifying in court.

As for AI it's just a chat bot using a language model and cool math, dont get me wrong, very cool predictive statistical math and big data shoved up it's back side. It can barely calculate basic math right now and I would not even trust that. [1]

Amazon let AI run its operations and caused spectacular outages. [2]

Trust it if you want to. I've never gotten a correct straight answer out of it though in fairness I do not throw softballs.

[1] - https://news.ycombinator.com/item?id=48297336

[2] - https://news.ycombinator.com/item?id=47323017

AGI is exceedingly unlikely to happen, so your worry is better spent on other things.

> If you disagree with this premise, please provide actual evidence

That's not the way that works. You're the one making the positive claim, it's on you to provide evidence to support your claim.

Why is it exceedingly unlikely to happen?

It's an extraordinary claim for which there is little, if any, supporting evidence. It is also a claim that is likely to emerge from one of the oldest and most deeply-rooted human cognitive biases, pareidolia.

It weren't exceedingly unlikely, we'd see some reasonably solid evidence of the possibility that is independent of pareidolia.

Look, you are free to worry and plan as you see fit.

But IF you were going to try to do some calculus on how to live in preparation for the future, you might want to consider the unsustainable habitat that climate breakdown is causing in the current now and near future.

Sure anything is possible, I'm not anti ai or llm by any means (just go look at my post history), and marvel at the thought of what agi/asi could look like. Philosophically, it is interesting to me. And I don't use the word "philosophy" to trivialize the concept. It genuinely is interesting to me.

But at the same time you should be aware that AGI is largely a concept that has been deeply inflated by play-acting radical Effective Altruists that have effectively reinvented / modernized Pascals Wager. The wake of generative llms is going to eventually recede (as all new technologies, even transformative ones) and leave us with a lot of extremely useful tools, a couple of bag holders and new and interesting processes for all sorts of automatable tasks.

These tools will not solve the climate catastrophe. The frontier AI labs want you to believe they will. But it's likely (if not certainly) that we cannot just technology our way out of this one.

Consider what life might look like with mass migrations and upward pressure from the global south, mega floods in some regions, mega droughts in others, and persistent wildfires that occupy most of the United States.

You don't have to consider too hard: this is already happening.

The concept of a "job" can encompass many things. Some of those change when machines become more capable, some don't.

For example, one part of a job is the satisfaction, the purpose, the meaning of life derived from being useful. Not having that can be pretty bad for people. So, even if the machines take over a lot of what is done by people today, you'll probably want to find a way of being useful.

Hedonism and wealth are both remarkably bad at delivering the same contentment that being useful does. That said, of course, a bit of hedonism and capital is also good.

> As a consequence, it seems pointless to work at a job.

You mean a programming job? Or all jobs?

People working as house painters, plumbers, farmers, chefs, engineers, architects, cops, soldiers, nurses and doctors aren't going to be irrelevant in 3 years, let alone 30. I think you're projecting your experience on the rest of the world, most of whom will be unaffected by LLMs in the short-term future.

I think most white collar jobs will fall at the rate that programming jobs do.

Regarding all those jobs that require working in the real world, I do imagine there will be a bit of a hold-out, though robotics could come in and cause the same phenomenon.

Reply on news.ycombinator.com