Unfortunately, it doesn't look like this is sufficient.
While I had great success with GrapheneOS in the past, bank apps in Brazil have started blocking it, even when the profile you run it under has Google services installed. So GrapheneOS (again, even with all Google Play Services and all other dependencies installed in a given profile) is still not completely transparent to apps.
This may be a coincidence (as I don't use it every day), but I noticed blocking started just as the recent Felca Law (which introduced mandatory age verification for every software, app and OS in Brazil) came into effect.
How about this for an evaluation: Have this (trained-on-older-corpus) LLM propose experiments. We "play the role of nature" and inform it of the results of the experiments. It can then try to deduce the natural laws.
If we did this (to a good enough level of detail), would it be able to derive relativity? How large of an AI model would it have to be to successfully derive relativity (if it only had access to everything published up to 1904)?
Maybe not you in particular, but I expect people to be more forthcoming in their writing towards LLMs vs a raw google search.
For example, a search of "nice places to live in" vs "I'm considering moving from my current country because I think I'm being politically harassed and I want to find nice places to live that align with my ideology of X, Y, Z".
I do agree that, after collecting enough search datapoints, one could piece together the second sentence from the first, and that this is more akin to a new instance of an already existing issue.
It's just that, by default I expect more information to be obtainable, more easily, from what people write to an LLM vs a search box.
Local governments in BR have already made ads using generative AI that were shown during prime time TV hours[1].
You can argue that is a bad thing (local designers/content producers/actors/etc lost revenue, while the money was sent to $BigTech) or that this was a good thing (lower cost to make ad means taxpayer money saved, paying $BigTech has lower chance of corruption vs hiring local marketing firm - which is very common here).
> The whole bio-fuel industry is a very complex mix of economics (often requires subsidies to make sense), geopolitical (less imported oil), environmental concerns (mass scale farming soil degradation and CO2 emissions derived from it) and logistical (completely different transportation and refining process).
"has existed long before humanity" isn't relevant for my argument.
"Will exist long after humanity" -> maybe, maybe not. If we're smart, capable and humble enough, we could, in principle, intentionally outlast them.
By "intentionally" I mean: we can design our future lightcone such that, by whichever measure you care to choose, there are still humans around. Yes, bacteria could be still around, but it won't be because they _chose_ to be around, it will be because it just so happened that the universe arranged itself in a way that they are still around.
By "in principle" I mean: if we spent enough resources, energy and smarts and built a civilization around this goal, we could plausibly (given the known laws of physics) do this. Whether we _will_ do it or destroy ourselves first any of the possible various means, is an open question.
Lineages of bacteria that exist today, here, will only keep existing in the _far_ future (billions of years from now, after the sun chars Earth and then spends its energy budget) if it just so happens that a panspermia event kicked some off our solar system and then they just so happen to find a suitable solar system to keep existing.
And indeed, accounting for externalities (unmeasured or unmeasurable) is a tough economic proposition. If it weren't hard to account for every single variable, creating a planned economy would be easier (ish).
FWIW, there's a whole sub-field just dedicated to determining the value of life for various purposes (a starting link: https://en.wikipedia.org/wiki/Value_of_life). You may disagree with any specific assessment, but then you have to argue how that value should be calculated differently.
The counterpoint is that not doing so (implying some sort of infinite monetary loss if the entire human species is wiped out) would mean you want to spend every single unit of monetary value of the entire global economy to preventing this (which is also obviously nonsense - people have to eat after all).
So you have to put the monetary value somewhere (although you're completely within your right to question this specific amount).
> I don’t think they can yet self improve exponentially without human intuition yet
I agree: if they could, they would be doing it already.
Case in point: one of the first things done once ChatGPT started getting popular was "auto-gpt"; roughly, let it loose and see what happens.
The same thing will happen to any accessible model in the future. Someone, somewhere will ask it to self-improve/make as much money as possible, with as little leashes as possible. Maybe even the labs themselves do that, as part of their post-training ops for new models.
Therefore, we can assume that if the existing models _could_ be doing that, they _would_ be doing that.
That doesn't say anything about new models released 6 months or 2 years from now.
Is it though? It is my understanding that the quantum fluctuations that give rise to BBs will still exist, even after (and specially after) the evaporation of black holes (perhaps assuming no Big Rip).
Attempting to summarize your argument (please let me know if I succeeded):
Because we can't compare human and LLM architectural substrates, LLMs will never surpass human-level performance on _all_ tasks that require applying intelligence?
If my summary is correct, then is there any hypothetical replacement for LLM (for example, LLM+robotics, LLMs with CoT, multi-modal LLMs, multi-modal generative AI systems, etc) which would cause you to then consider this argument invalid (i.e. for the replacement, it could, sometime replace humans for all tasks)?
Interesting, I have the exact polar opposite perception.
I accessed this through a Qubes AppVM (no GPU, limited memory and CPU budget) and the presence of videos makes this a very slow scrolling experience for me.
In general, anything that involves JS/CSS animation/blur/effects makes sites pretty slow (up to unusable for me). The unlogged homepage for github.com for example, spins my cpu at 100%.
If a generic human glances at an unfamiliar screen/wall/room, can they accurately, pixel-perfectly reconstruct every single element of it? Can they do it for every single screen they have seen in their entire lives?
While I had great success with GrapheneOS in the past, bank apps in Brazil have started blocking it, even when the profile you run it under has Google services installed. So GrapheneOS (again, even with all Google Play Services and all other dependencies installed in a given profile) is still not completely transparent to apps.
This may be a coincidence (as I don't use it every day), but I noticed blocking started just as the recent Felca Law (which introduced mandatory age verification for every software, app and OS in Brazil) came into effect.