kaffekaka

Maybe, but what the hell does "ai native" actually mean?

What other changes have been like that? Tipping points.

The rhetoric has basically been "replace 75% of white collar work in two years" so zero impact after 9 months surely seems.. disappointing?

Well this is an actual effect. It is so easy now to generate presentations etc so as a consequence there are ten times as many that people seem to expect you to read. Productivity is higher but that which is produced sucks and is pointless. Actual value creation is harder than nice looking artifacts.

Math was not "just" proof of concept, it was THE proof of concept, since iterating on combustion engines cannot be done nearly as efficiently.

See sibling comment. 10x measured how? In pure coding tasks under certain conditions the speedup is huge, but seems really unclear how this would transfer to any significant speedup across the whole undertaking (project, delivery etc).

Maybe, even likely. But how do you observe this? We are in this situation at work, it feels like "sure AI is helpful" but do we actually deliver any faster?

I do not mean to be mean, but how had you missed this? The models have been writing python to accomplish tasks forever (meaning a couple of years at least).

The models using code/scripts in the background is not new.

NO, you build a new one every time, or else you get the hose again.

The "I don't care if it ain't true, it sucks anyway."

I like Sweden but fall and winter sure are dark. Can totally understand why it might be a bad fit for many people.

Isn't the problem (if OpenAI and Anthropic crash) that the business model is not working, so it won't be feasible to buy all the compute and carry on?

If they fail, the successor must do something differently. Perhaps by not alluding to killing all humanity, destroying the economy, making humans irrelevant etc.

I don't have a good answer, but a question of my own. I wonder if there are only the two extremes: write everything manually or one-shotting without looking at the code?

One-shotting without review is very quick but has its downsides: burnout, less understanding of the code base, less interacton with your colleagues, you miss the poor choices the LLM made so the code base might rot.

Writing everything manually is much slower and in some circumstances not what you want anyway.

Is there a middleground where you get significant parts of the productivity/speed increase but retain the most important parts of the manual process? I have found I preferred the copilot style of working, but the interface was not really optimized.

Instead of spewing out code at 20x speed can we settle for 5x and remain in control? That would be my ideal and more sustainable, right?

Very well written, thank you.

1. Yes. 2. Maybe

Thank you for your explanation. We have very different view of the future. Generating a correct piece of code can not be compared to taking care of an old, fragile individual.

Even if this robotic technology existed today (which it does not), it would take more than five years even to get it out there in the real world. If you think people in general are negative to AI, wait until they're told "AI will look after your elderly mother".

I want to understand. Do you seriously believe that in five years chefs, old age care, house cleaning etc will be "definitely" automated?

What do you mean by automated? No humans in the loop, or that the human doing the work is using some kind of AI tool to aid them?

I would never hire an AI plumber. I want the guy who came here last time, did a good job, and pay him premium prices.

> They're saying "our position of relative privilege is threatened and we must defend it."

No, they are not.

It remains to be seen if "any problem can be solved before a human can understand it". What if it turns out there are (many) problems too hard for both AI and humans.

It has been proved a hundred years ago that some problems can not be proven. So there are problems AI will not solve.

Both can be true.

Real grief for mathematicians, coders, artists etc (not waved away as "he is just in stage 2 bla bla").

But also: yes of course, the challenge now is to use these supposed super powers for good.

OT but what is is that has enabled these advances? Bigger models, better training, more compute, reasoning, all of these?

Theoretical physics would, from my outsider view nowadays, benefit from an injection of vitality and an ability to tackle new venues. But the question is open whether LLMs can truly break new ground or if they work within the convex hull of current theory. And if that matters, too.

I hear you.

Can your ai summarize that for my ai please?

What is?

Sure, I do not believe warp drive is possible, but it was of course just an example.

I think the thing is if/when #3 turns true society is no more.

Then people respond: "the elite will use their ai armies to kill everyone else". This will not happen. I believe more in the worst case being chaos and rebellions. Not good for anyone, not even the elite.

"An algorithm that can solve any math problem", I am not Alan Turing or Kurt Gödel, but this seems impossible.

Could very well be true.

For example, will anyone take on the Riemann hypothesis now, knowing that AI might solve it any day. But then, if AI can never solve it, we will have lost out on all the human effort that would have given us great new theories and insights along the way.

All the work that would give rise to new discoveries might now never even get started. That is depressing.

I agree with you. If we ever want to invent warp drive, a machine that lets us do it much faster ought to be a god send. If there are scientific solutions to our biggest problems, this should be a blessing.

But the risk is of course that the benefits will not be distributed but concentrated to the elite.