It isn’t just you and me. Not even the people who designed them fully understand why they give the responses they give. It’s a well-known problem. Our understanding is definitely improving over time, but we still don’t fully know how they do it.

Here’s the latest exploration of this topic I could find.

LLMs continue to be one of the least understood mass-market technologies ever

Tracing even a single response takes hours and there’s still a lot of figuring out left to do.

I highly doubt that. For so many reasons. Here’s just a few:

• What data would you train it on, the Constitution? The entirely of federal law? How would that work? Knowing how ridiculous textualism is even when done by humans, do you really think a non-thinking algorithm could understand the intention behind the words? Or even what laws, rules, or norms should be respected in each unique situation?
• We don’t know why LLMs return the responses they return. This would be hugely problematic for understanding its directions.
• If an LLM doesn’t know an answer, instead of saying so it will usually just make something up. Plenty of people do this too, but I’m not sure why we should trust an algorithm’s hallucinations over a human’s bullshit.
• How would you ensure the integrity of the prompt engineer’s prompts? Would there be oversight? Could the LLM’s “decisions” be reversed?
• How could you hold an LLM accountable for the inevitable harm it causes? People will undoubtedly die for one reason or another based on the LLM’s “decisions.” Would you delete the model? Retrain it? How would you prevent it from making the same mistake again?

I don’t mean this as an attack on you, but I think you trust the implementation of LLMs way more than they deserve. These are unfinished products. They have some limited potential, but should by no means have any power or control over our lives. Have they really shown you they should be trusted with this kind of power?

It looks like they’re still very much trying to build this stupid city, and are doing everything they can to circumvent local opposition.

About a year ago, some guy knocked on my door as part of a canvas to gauge the support of towns in the surrounding area, but he never once mentioned that this dumb Silicon Valley wet dream was the project he was talking about. He spoke in very vague terms about “building community” and providing “housing and opportunities for the working class,” which sounded great so I said of course I support those things. He smiled, jotted something down and then left. I called after him asking what organization he represented, and he turned around, handed me a postcard with a shitty AI generated image on it, and then he was gone. Turns out it was California Forever, the org behind the dystopian exploitation fantasy that is their juvenile idea for a new city. I felt violated, humiliated, and furious. They’re doing some seriously underhanded bullshit to try to force this down our throats.

Reminder that Ellison is an 80 year old billionaire who wants to surveil all of us to keep us on our “best behavior.” Fuck this old and out of touch asshat.

“Citizens will be on their best behavior because we are constantly recording and reporting everything that’s going on,” Ellison said, describing what he sees as the benefits from automated oversight from AI and automated alerts for when crime takes place. "We’re going to have supervision”

Source

Anyone who thinks this could ever be implemented in a way that respects peoples’ civil rights is delusional and/or lying. Ellison is not someone we should be listening to for advice on how society should function, whether it’s about stealing more intellectual property for useless AI crap or mass surveillance.

AI hilarity aside (they really nailed the article conclusion), the title looks like a kooky supernatural jdrama I’d totally watch…

For sure, that makes sense. This company just seems to be conflating quite a bit. Is it AI? No. Are they using a Large Language Model? Kind of. It seems like they put a whole bunch of chemical, soil, and market/tastes metrics into a GPT model, and ran it to fine-tune their process. So, sure, that’s cool, but this all just seems like a silly gimmick. I’d be shocked if this wine was anything special, but then again I’m incredibly suspicious of anyone who slaps an AI label on their product, so who knows.

Because if there’s one thing we know about algorithms, it’s that they have an amazing sense of taste…

I agree with you and your immediate loss of trust when seeing ai-generated images, especially in news articles. My view is that if you are an established media company, then it’s your responsibility to pay for legal illustrations/images. Whether that be hiring an illustrator to make something from scratch or buying the rights to an existing image, participating in that market keeps creatives employed and keeps the quality of the art high, or a least not filled with seven-fingered people. Once we give up on artists creating new works and rely only on remixing drivel without permission, we’ll lose quite a lot of creativity, and people can obviously tell the difference.

I’m a little more ambivalent about independent journalists or newly spun-up media companies that are still getting their bearings using image generators, but frankly my impression whenever I see one even on a small independent site is still “yuck, no image would be better than this crap.”

So yeah, there are definitely more important things to be upset about when it comes to AI being shoved down our throats, but this also bothers me a lot.

I think that actually sums up one of the administration’s plans in general, to take hold of lawless industries and use them to get rich at everyone else’s expense. Their ridiculous plans for crypto have pretty much the same motive.

So AI practitioners would also be held to the same standards and be subject to the same consequences as human doctors then, right? Obviously not. So this means a few lines of code will get all the benefits of being a practitioner and bear none of the responsibilities. What could possibly go wrong? Oh right, tons of people will die.

I don’t really understand the downvotes, this is a perfectly valid shower thought as far as I’m concerned.

There’s something to be said for simplifying life to reach some kind of lost happiness. But practically speaking, living as an early human, hunter gatherer, or even as an agricultural pioneer, was incredibly hard work. It’s not like there was necessarily more time to relax. You ever tried going into the forest and gathering enough food to keep yourself fed for more than a few minutes? It’s really hard. On top of that, most of humanity’s technological and medical advancements far outweigh the stresses of modern life, in my opinion.

That being said, some point to the fact that since life was hard, less comfortable, more risky, and mostly occupied by essential tasks, that it might counterintuitively lead to a more fulfilling life. I’m not sure I agree with that characterization, but sometimes hard manual labor that yields results feels good, so I get it.

Regardless, solid shower thought, I often think about similar things.

Thank you!

That definitely sounds concerning, do you have a link or anything? Any company would have to comply with a legal court order regardless of their preference or ideology, so how they comply comes down to the amount of information they store and for how long. Just saying, the fact that they helped the cops (as much as I hate the cops in general) doesn’t necessarily mean they handed everything over or volunteered anything they weren’t legally obligated to provide. But I don’t know the specifics of the case you mentioned.

Most if not all non-paid options will have privacy concerns similar to gmail, so you’ll have to pay. I pay for Proton, and though I’m not thrilled about some of the political bullshit their CEO has been up to recently, I think I’ll stay with them for now since they’re still good for privacy, and their other services are solid. They’re also very upfront about what they charge and why, and I think they still plan to transition to a non-profit.

Yup! I use Mobius on iOS to sync everything I want, including random one off file transfers, but mostly to keep my Keepass database and Obsidian vault in sync across devices. It works really well, despite apple making it super difficult to run consistently in the background. As long as you open the app every so often, it’s seamless and reliable.

Syncthing.

I’m sure you know, but you’re probably going to get a lot of grief for this. I’m deeply suspicious of any new AI tool, especially one that tries to get in between me and my news (looking at you Feedly), and I’m sure I’m not the only one. So if you’re not already, I’d prepare yourself for a lot of strong emotions, and probably not in a good way.

If you wanted to get ahead of that kind of thing, you might want to explain what kinds of safeties you’re building into it. For example, on your roadmap you say want it to “Generate argument of for and against perspective then summarise the result of the 2 arguments.” This kind of thing in particular is quite risky. Any time you try to introduce value statements into an LLM summary, you’re in the danger zone. Even if you’re just trying to summarize the actual perspective of the piece, you’re basically just begging the LLM to hallucinate. But asking it to summarize hypothetical opposing arguments is just asking for trouble.

I could go on, but I don’t want to start a pile on. I appreciate when folks try to build cool stuff, you’ve just waded into some choppy waters…

People carry debt from their Christmas gifts for the rest of the year, you’re right that it’s totally not harmless. I’m shocked how much some people spend on holiday gifts.

My family, granted we’re Jewish, basically stopped all gift-giving, except for when there’s a new kid in the extended family, which is rare. I wholly agree that Christmas is the best example of conspicuous consumption/consumerism and extreme capitalism. Frankly, Christmas season makes me kind of uncomfortable because of all the social expectations to give gifts. Fortunately I can usually say ‘“nah I’m Jewish, I don’t do Christmas,” but you’d be surprised how many folks are totally confused by that. So yup, spoil your kids if you want to, but don’t expect me to send you a holiday gift if you’re a functional adult.

Totally understandable. I was hesitant about it too, but to me it’s sometimes worth a premium to get something that just works, and Infuse works better than any player I’ve used in the past. To each their own, good luck!