It needs to be more trustworthy. If I have to double check everything, I still have to figure out how to do whatever it’s doing, then figure out how it’s doing the thing, then verify if it did it right. By then, I could have just done it in step 1.5 probably.

I just think this is patently false. Or at least there are/were orgs where cloud costs so much more than running their own servers that are tended by maybe 1 FTE across a bunch of admins mostly doing other tasks.

Let me just point out one recent comparison - we were considering cloud backup for a couple petabytes of data, with a few hundred GB changing or adding / restoring every week or less. I think the best deal, where we held the software costs equal was $5/TB/Month.

This is catastrophically more expensive over a 10 year lifespan of a server or two and a small/mid sized LTO9 tape library and tapes. For one thing, we’d have paid more than the server etc in about a year. After that, tape prices have always tended down over time, and the storage costs for us for tape is basically $0 once in archive storage. We put it in a cabinet in another building - and you can fit A LOT of data in these tapes in a small room. That’ll cost basically $0 additional for 20 years, forget about 10. So let’s add in electricity etc - I still have doubts those will be over ~$100k over the lifetime of the project. Labor is about a wash cause you still need people to manage the backups to the cloud, and I think actually moving tapes might be ~.05 FTE in our situation. Literally anyone can be taught how to do it once the backup admin puts the tapes in the hopper or tells them which serial # to put in the hopper.

I also think that many companies are finding something similar for straight servers - at least it was in the news quite a bit for a while. Now, if you can be entirely cloud native - maybe it washes out, but for large groups of people that’s still not possible due to controlling hardware (think factory,scientific, etc)or existing desktop software for which the cloud isn’t really a replacement and throughput isn’t great (think Adobe products, video, scientific, financial etc data).

Though it’s more work with current AI at least compared to another team member - the AI cannot have access to a lot of context due to data security rules.

I feel this also misses something rather big. I find there’s a huge negative value of people I have to help through doing a task - I can usually just get it done at least 2x if not 5x or more faster and move on with life. At least with a good intern I can hope they’ll learn and eventually actually be able to be assigned tasks and I can ignore those most of the time. Current AI can’t learn that way for various reasons, some I think technical, some business model driven, whatever. It’s like always having the first day on the job intern to “help”.

The other problem is - unless I have 0 data security rules, there’s just so much the AI cannot know. Like I thought today I’d have Claude 3.7 thinking write me a bash script. I wanted it to query a system group and make sure the members of that group are in the current users .k5login. (Now, part of this is me not knowing how to prompt, but it’s also stuff a decent intern ought to be able to figure out.) One, it’s done a lot of code to work out what the realm is - this is useful generically, but is just code that could contain bugs when we know the realm and there’s only one it’ll ever operate in.

I also had to re-prompt because I realized it misunderstood me the first time, whereas I think an intern would have access to the e-mail context so would have known what I meant.

Though I will say it’s better than most scripters in that it actually does a lot of “safety” stuff we would find tedious and usually have to have something go wrong to add in, so … swings and roundabouts? It did save me time, assuming we all think it’s method is good enough - but this is also such a simple task that I think in some ways it’s barely above filling out a lot of boilerplate. It’s exactly the sort of thing I would have expected to see on stack overflow back in the day.

EDIT: I actually had a task that felt 100% AI could have done… if there was any way for it to know lots and lots of context. I had to basically fill out a long docx file with often AI like text describing local IT security standards, processes, responsibilities and delegations. Probably over 60% I had to “just make up” cause I didn’t have the context - for higher ups to eventually massage into a final form. But I literally cannot even upload the confidential blank form, forget about have some magic way for AI to get a brain dump from me about the last 10ish years of spoken knowledge and restricted wiki pages. Anything it could have made up mostly would have “been done” by the time I made a functional prompt.

I don’t think we solve this till we can run frontier models locally at prices less than a human salary, with integrations into everything a human in that position could access.

For home use (and small uses at work) I’ve found cyberpower to be cheaper than APC and yet work as well. You’d likely need to get a model with a network card option, and that’ll cost more I think. I’m not in EU though, so IDK what model would meet your needs and price point (which seems pretty low to me for a network enabled UPS).

Where is the actual link for the article? Is it a video?

I think it’s more the cloud being the issue here. Such an obvious and large and valuable target. Of course Microsoft also isn’t that secure historically.

Probably just so you don’t accidentally waste time unknowingly rereading a book.

If youtu.be premium is actually ad free (and I can still run sponsor block) I can see paying for it. I get a lot of value out of it and can’t really justify doing a lot of work to avoid paying.

It’s fascinating how normies will figure out a VPN to use shitty social media but not figure out more open systems or alternative sites.

It’s also the anti commodity stuff IP has been allowing. If Hershey makes crap chocolate, there is little stopping you from buying Lidnt say. But if Microsoft makes a bad OS, there’s a lot stopping you from using Linux or whatever.

What’s worse is stuff like DRM and computers getting into equipment that otherwise you could use any of a bevy of products for. Think ink cartridges.

Then there’s the secret formulas like for transmission fluid now where say Honda says in the manual you have to get Honda fluid for it to keep working. Idk if it’s actually true, but I l’m loathe to do the 8k USD experiment with my transmission.

You’d think the government could mandate standards but we don’t have stuff like that.

Testing the Jellyfin photos thing out now. I don’t know if it’s working right, but when I first looked at it the issue was I thought it seemed very video focused. I guess otherwise I’m learning docker after all.

Fair enough, last time I tried docker, which was a long time ago, I had all sorts of issues with permissions and persistence. I guess it’s probably better now.

I don’t want a research project. I just was hoping there was an easy to use program to make the viewing better than samba shares. Maybe I just need a set of programs that will display thumbnails over samba.

That seems a strange definition to me - so if a boxer gets back up after losing the match, well his opponent didn’t beat him in that fight?

I hear this a lot, but what would beating the Taliban involve? While the US was there, the Taliban was at best in hiding, it was not holding territory. If you mean removing the very idea of the Taliban from the world? That is both hard to do and arguably also a genocide, at least a cultural one. The US has been good at that, but it’s also frowned on in the current world - see Gaza headlines.

This is also why I’d suggest it’s kind of impossible to both not be the worst of the colonialist systems and stop terrorism (and it’s kind of unclear that even the colonial cultural suppression / conversion / excesses / crimes actually would stop terrorism).

Yes definitely. Many of my fellow NLP researchers would disagree with those researchers and philosophers (not sure why we should care about the latter’s opinions on LLMs).

I’m not sure what you’re saying here - do you mean you do or don’t think LLMs are “stochastic parrot”s?

In any case, the reason I would care about philosophers opinions on LLMs is mostly because LLMs are already making “the masses” think they’re potentially sentient, and or would deserve personhood. What’s more concerning is that the academics that sort of define what thinking even is seem confused by LLMs if you take the “stochastic parrot” POV. This eventually has real world affects - it might take a decade or two, but these things spread.

I think this is a crazy idea right now, but I also think that going into the future eventually we’ll need to have something like a TNG “Measure of a Man” trial about some AI, and I’d want to get that sort of thing right.

Yea, that was a bad way to phrase it - I just meant that from what I’ve heard tokens are very much not word by word. And sometimes it’s a couple words, but maybe that was misinformation. And I was trying (and failing) to make an analogy for a human - a concept is a compression of what otherwise would be a bunch of words, though I kind of meant more like a reference I guess.

I think it’s very clear that this “stochastic parrot” idea is less and less accepted by researchers and philosophers, maybe only in the podcasts I listen to…

It’s not capable of knowledge in the sense that humans are. All it does is probabilistically predict which sequence of words might best respond to a prompt

I think we need to be careful thinking we understand what human knowledge is and our understanding of the connotations if the word “sense” there. If you mean GPT4 doesn’t have knowledge like humans have like a car doesn’t have motion like a human does then I think we agree. But if you mean that GPT4 cannot reason and access and present information - that’s just false on the face of just using the tool IMO.

It’s also untrue that it’s predicting words, it’s using tokens, which are more like concepts than words, so I’d argue already closer to humans. To the extent it is just predicting stuff, it really calls into question the value of most of the school essays it writes so well now…

Well, LLMs can and do provide feedback about confidence intervals in colloquial terms. I would think one thing we could do is have some idea of how good the training data is in a given situation - LLMs already seem to know they aren’t up to date and only know stuff to a certain date. I don’t see why this could not be expanded so they’d say something much like many humans would - i.e. I think bla bla but I only know very little about this topic. Or I haven’t actually heard about this topic, my hunch would be bla bla.

Presumably like it was said, other models with different data might have a stronger sense of certainty if their data covers the topic better, and the multi cycle would be useful there.