theptip

33 ms·

541.
by theptip 1y ago
This is a well-debated point; should you test the system on its own, or system+scaffold? How much custom scaffolding is allowed? Both avenues are interesting. For AGI presumably the “general” bit means you don’t get to build task-specific s
542.
by theptip 1y ago
> Max didn't need a request router, he just put his PHP file at the right place on the disk. This also elides a bit of complexity; if I assume I already have the Nginx and gunicorn process then my Python web server isn’t much worse.
543.
by theptip 1y ago
Only kind of, they have their own language (Hack) that descends from PHP. It’s JIT instead of interpreted, and breaks back-compat in a few ways.
544.
by theptip 1y ago
> What does this look like? Like asking children learning to read to demonstrate they can read Shakespeare? A staple of modern education is scaffolding learning We definitely need to rethink some assumptions. I view education as serving
545.
by theptip 1y ago
> LLMs seem to be in that vein My opinion is that even if capabilities halted now, LLMs would be more economically valuable than the internet (compared over the same 50 year trajectory). And I predict that they will not halt any time soo
546.
by theptip 1y ago
> What's the hard solution? Ask students to solve harder problems, assuming they will use AI to learn more effectively. Invert the examination process to include teaching others, which you can’t fake. Or rework it to bring the viva
547.
by theptip 1y ago
Is there an opportunity for reusable OSS rule systems? It seems that every government at every level could use an expressive system for defining and versioning them. I suspect part of the gap here is, as you note, everything technical is ou
548.
by theptip 1y ago
In this scenario the precedent has already been set in the 5th, why would the Supreme Court hear it again? The 9th has nothing to do with an immigration hearing in the 5th, the individual’s personal history is irrelevant here. Matter of Rah
549.
by theptip 1y ago
This is cool, I’ve been wanting this for ages. The “AI powered game” usecase is going to be huge and really requires BYO API key to work. I tried wiring something up and immediately hit a gap where I need to be able to call a tool. I think
550.
by theptip 1y ago
I feel things like shoplifting should actually be automatable, it’s a question of ROI currently. One idea I play with is “police 2.0” where you can dispatch a small fast drone to a crime scene, and follow the perp from a safe distance. A lo
551.
by theptip 1y ago
As a user I’d be happy to pay $5 for a bundle of credits and just top up whenever it runs out. And as you say you don’t want to be in the position where a whale costs you $50 by submitting a crazy number of requests. Maybe these are big-sca
552.
by theptip 1y ago
Right now we are in the “laws are seldom actually enforced” regime. It seems pretty clear that laws will be enforced more in future, the obvious response is to go prune the laws to get rid of the ones that we actually aren’t OK with being e
553.
by theptip 1y ago
I like the general idea, and I’ve been surprised this hasn’t taken off elsewhere, eg citizen videos for traffic violations like blocking intersections, it seems these should be ROI positive for the city to implement (lower enforcement costs
554.
by theptip 1y ago
Nice. Pricing seems a bit steep for occasional use; does iOS make it easy to do micro-transactions with Apple Pay? (I get the dev may be trying to put bread on the table with this, which is also fine…)
555.
by theptip 1y ago
I do agree there is a sense in which this has always been true. But most people’s constructed worlds overlapped quite significantly up until now; it was necessary for survival. I think that condition is weakening. Just look at how wealth af
556.
by theptip 1y ago
I think Eliezer’s take here is extremely bad, ie the AI doesn’t “know it’s making people insane”. But I think the author's point is apt. There are a bunch of social issues that will arise or worsen when people can plug themselves into
557.
by theptip 1y ago
The problem with Evolutionary Psychology is that it’s easy to construct this sort of “just so” story for pretty much any causal link, and they are not falsifiable. IMO the simpler hypothesis here is that this isn’t anything adaptive, it’s j
558.
by theptip 1y ago
Anthropic is the most open of the major labs; RSP, actual model cards, and they publish their red teaming results. C.f. openAI who released o3 and didn’t publish any model card or safety eval at all, their justification being “it’s not GA,
559.
by theptip 1y ago
You can steer an AI without training it. It’s called Prompt Engineering. It doesn’t give you full control, but I believe it gives you enough control to do what I proposed for most consumers.
560.
by theptip 1y ago
I think AI gives us the opportunity to write our own ranking algorithms. Of course most people will pick an off-the-shelf editorial perspective but I expect these to be much more configurable than the current state with a tiny number of alg
561.
by theptip 1y ago
Discovery definitely becomes more important in this world. Fortunately AI helps with that too.
562.
by theptip 1y ago
Fair, I was trying to make a broader point without being clear about it :)
563.
by theptip 1y ago
> People are often curious about how much energy a ChatGPT query uses; the average query uses about 0.34 watt-hours My point was twofold, “the average query” becomes less meaningful as the variance increases. Sure one can _in principle_
564.
by theptip 1y ago
Watt-hours seem good right now, and rebut the current arguments that LLMs are wasteful. They might soon stop being the right lens, if your agents can go off and (in parallel) expend some arbitrary and borderline-unknowable budget of energy
565.
by theptip 1y ago
Right. Jevons Paradox. The lower barrier to entry might mean the average quality of software goes WAY down. But this could be great if it means that there are 10x as many little apps that make one person happy by doing only as much as they
566.
by theptip 1y ago
One lens is the “Thrive/Survive” axis. As more people thrive, they have more space for individualism, compassion for others, etc. This can be used to explain the arc of “moral progress” over time. A follow up would be to observe that i
567.
by theptip 1y ago
Agreed, my hunch is that you might use higher abstraction-level validation tools like acceptance and property tests, or even formal verification, as the relative cost of boilerplate decreases.
568.
by theptip 1y ago
It’s a good observation. Counterpoint - the organization labor is drudge work, maybe with cheap AI labor you can have your KB kept in great shape, and likely this makes it easier for RAG to locate the correct document. Maybe there is room f
569.
by theptip 1y ago
I think the point is that with thin margins and capital costs, 1.6% (compounded over decades) could be a large chunk of your profit. TFA discusses this: > To justify their investment and make a profit, "it's very important for
570.
by theptip 1y ago
Depends, in this case the strategic bombers are worth more.

More ›