How would pure ai ever be "much stronger" in this scenario?
That doesn't make any sense to me whatsoever, it can only be "equally strong", making the approach non-viable because they're not providing any value... But the only way for the human in the loop to add an actual demerit, you'd have to include time taken for each move into the final score, which isn't normal in chess.
But I'm not knowledgeable on the topic, I'm just expressing my surprise and inability to contextualize this claim with my minor experience of the game
You can be so far ahead of someone, their input (if you act on it) can only make things worse. That's it. If a human 'teams up' with chess AI today and does anything other than agree with its moves, it will just drag things down.
These human in the loop systems basically lists possible moves with likelihood of winning, no?
So how would the human be a demerit? It'd mean that the human for some reason decided to always use the option that the ai wouldn't take, but how would that make sense? Then the AI would list the "correct" move with a higher likelihood of winning.
The point of this strategy was to mitigate traps, but this would now have to become inverted: the opponent AI would have to be able to gaslight the human into thinking he's stopping his AI from falling into a trap. While that might work in a few cases, the human would quickly learn that his ability to overrule the optimal choice is flawed, thus reverting it back to baseline where the human is essentially a non-factor and not a demerit
>So how would the human be a demerit? It'd mean that the human for some reason decided to always use the option that the ai wouldn't take, but how would that make sense? Then the AI would list the "correct" move with a higher likelihood of winning.
The human will be a demerit any time it's not picking the choice the model would have made.
>While that might work in a few cases, the human would quickly learn that his ability to overrule the optimal choice is flawed, thus reverting it back to baseline where the human is essentially a non-factor and not a demerit
Sure, but it's not a Centaur game if the human is doing literally nothing every time. The only way for a human+ai team to not be outright worse than only ai is for the human to do nothing at all and that's not a team. You've just delayed the response of the computer for no good reason.
If you had a setup where the computer just did its thing and never waited for the human to provide input but the human still had an unused button they could press to get a chance to say something that might technically count as "centaur", but that isn't really what people mean by the term. It's the delay in waiting for human input that's the big disadvantage centaur setups have when the human isn't really providing any value these days.
But why would that be a disadvantage large enough to cause the player to lose, which would be necessary for
> pure AI teams tend to be much stronger.
Maybe each turn has a time limit, and a human would need "n moments" to make the final judgement call whereas the AI could delay the final decision right to the last moment for it's final analysis? So the pure AI player gets an additional 10-30s to simulate the game essentially?
Why? If the human has final say on which play to make I can certainly see them thinking they are proposing a better strategy when they are actually hurting their chances.
They can, but that invites retaliation. If they do it on all non-EU trades that is in effect an attack on investment in every non-EU economy. There are all sorts of consequences.
There are practical difficulties too. What is the investments are made through an off-shore subsidiary or by investing in a fund in the US? It might even encourage the latter (single trade to buy the fund instead of managing US investments) and mean management fees move to the US.
If they impose it on just US trades it might still be subject to the problems above and would be a very serious step and one investors will hate.
I think we might well end up with something like this and a return to more money going to national capital markets rather than global, but its not going to be an easy transition.
I read it, and it's not enough to make a judgement either way. For all we know none of this had anything to do with his ban and he was banned for something he did the day before. There's no way for third parties to be sure of anything in this kind of situation, where one party shares only the information they wish and the other side stays silent as a matter of default corporate policy.
I am not saying that the author was in the wrong and deserved to be banned. I'm saying that neither I nor you can know for sure.
> There's no way for third parties to be sure of anything in this kind of situation,
Not just third parties, but also the first party can't be sure of anything - just as he said. This entire article was speculation because there was no other way to figure out what could've caused the ban.
> where one party shares only the information they wish and the other side stays silent as a matter of default corporate policy.
I don't think that's a fair viewpoint - because it implies that relevant information was omitted on purpose.
From my own experience with anthropic, I believe his story is likely true.
I mean they were terminating sessions left an right all summer/fall because of "violations"... Like literally writing "hello" in a clean project and first prompt and getting the session terminated.
This has since been mostly resolved, but I bet there are still edge cases on their janky "safety" measures. And looking at the linked claude.md, his theory checks out to me. I mean he was essentially doing what was banned in the TOS - iteratively finding ways to lead the model to doing something else them what it initially was going to do.
If his end goal was to write a malware which does, essentially, prompt injection... He'd go at it exactly like this. Hence sure as hell can imagine anthropic writing a prompt to analyze sessions determining bad actors which caught him
Asserting that somebody is "victim blaming" isn't giving somebody the benifit of the doubt, and in the context of a scenario were few if any relevant facts are known reveals a very credulous mindset.
The other day we were discussing a new core architecture for a Microservice we were meant to split out of a "larger" Microservice so that separate teams could maintain each part.
Instead of just discussing it entirely without any basis, I instead made a quick prototype via explicit prompts telling the LLM exactly what to create, where etc.
Finally, I asked it to go through the implementation and create a wiki page, concatting the code and outlining in 1-4 sentences above each "file" excerpt what the goal for the file is.
In the end, I went through it to double-check if it held up from my intentions - which it did and thus didn't change anything
Now we could all discuss the pros and cons of that architecture while going through it, and the intro sentence gave enough context to each code excerpt to improve understanding/reduce mental load as necessary context was added to each segment.
I would not have been able to allot that time to do all this without an LLM - especially the summarization to 1-3 sentences, so I'll have to disagree when you state this generally.
Though I definitely agree that a blog article like this isn't worth reading if the author couldn't even be arsed to write it themselves.
Agreed, there is always this little story to remind us of our unity - at least from the perspective of the draftees/workers back in ww1, where everyone was basically forced to fight each other by the elites
1. It's not unique to the play store, as a matter of fact, this started in the iOS app store and was "adopted" by Google. It could definitely be improved though, i.e. if all potential in-app purchases were listable via the store page, like on steam for example
2. The prices were mentioned in the comment you're responding to.
Where? Just checked on every app I've purchased in app unlocks and none of them have any indicator for these unlocks (or others that are still available) on their app store page.
The only way to see them - from my experience which I just verified - is to go into the app and go into the relevant menu's of the apps.
Please explain where you're able to see this information on the app store on iOS or iPadOS
Is this maybe only available for some regions or opt-in for the developer? I this UX doesn't exist on my devices running on 26.2 in the apps I checked. I just verified again but no luck
/Edit: found it! that is way too hidden - Would never have found that without your explicit mention and gif link!
After exploring some more on the play store too, There is actually a similar UI in the app details there too, it doesn't list all items but the price range (cheapest item to most expensive item). Definitely worse then having all items listed, but both could be improved imo by listing them as repeatable purchases, temporary licenses, forever unlocked etc) for informed consent before install. I'd never install any app which has repeatable transactions for example
Those scripts have logs, right? Log a hostname and path when they run. If no one thinks to look at logs, then there's a bigger problem going on than a one-off script.
That becomes a problem if you let the shell script mutate into an "everything" script that's solving tons of business problems. Or if you're reinventing kubernetes with shell scripts. There's still a place for simple solutions to simple problems.
You can literally have a 20 line Python script on cron that verifies if everything ran properly and fires off a PagerDuty if it didn't. And it looks like PagerDuty even supports heartbeat so that means even if your Python script failed, you could get alerted.
Which is why you take the time to put usage docs in the repo README, make sure the script is packaged and deployed via the same methods that the rest of the company uses, and ensure that it logs success/failure conditions. That's been pretty standard at every organization I've been at my entire professional career. Anyone who can't manage that is going to create worse problems when designing/building/maintaining a more complex system.
Yah. A lot of the complexity in data movement or processing is unneeded. But decent standardized orchestration, documentation, and change management isn't optional even for the 20 line shell script. Thankfully, that stuff is a lot easier for the 20 line standard shell script.
Or python. The python3 standard library is pretty capable, and it's ubiquitous. You can do a lot in 50-100 lines (counting documentation) with no dependencies. In turn it's easy to plug into the other stuff.
Debatable I would argue. It's definitely not 'just a statistical model's and I would argue that the compression into this space fixes potential issues differently than just statistics.
But I'm not a mathematics expert if this is the real official definition I'm fine with it. But are you though?
its a statistical term, a latent variable is one that is either known to exist, or believed to exist, and then estimated.
consider estimating the position of an object from noisy readings. One presumes that position to exist in some sense, and then one can estimate it by combining multiple measurements, increasing positioning resolution.
its any variable that is postulated or known to exist, and for which you run some fitting procedure
I'm disappointed that you had to add the 'metamagical' to your question tbh
It doesn't matter if ai is in a hype cycle or not it doesn't change how a technology works.
Check out the yt videos from 1blue3brown he explains LLMs quite well.
.your first step is the word embedding this vector space represents the relationship between words. Father - grandfather. The vector which makes a father a grandfather is the same vector as mother to grandmother.
You the use these word vectors in the attention layer to create a n dimensional space aka latent space which basically reflects a 'world' the LLM walks through. This makes the 'magic' of LLMs.
Basically a form of compression by having higher dimensions reflecting kind a meaning.
Your brain does the same thing. It can't store pixels so when you go back to some childhood environment like your old room, you remember it in some efficient (brain efficient) way. Like the 'feeling' of it.
That's also the reason why an LLM is not just some statistical parrot.
You'll be surprised to know that there are still some mice that don't support that. Admittedly, I've only had that happen once in the last 15 yrs in a budget "gamer" mouse I instantly returned and replaced with a Logitech g903 at the time (though I've switched mice twice since, and both supported it)
> No, it never was intended like that.
It was certainly marketed as that though...
reply