jacksilver
@jacksilver@lemmy.world
- Comment on Nvidia Announces DLSS 5, and it adds... An AI slop filter over your game 2 days ago:
It feels like homogenization of art. The filter pays no attention to the original art style and just conforms it to nividias Ai style.
- Comment on How Much Do LLMs Hallucinate in Document Q&A Scenarios? A 172-Billion-Token Study Across Temperatures, Context Lengths, and Hardware Platforms [TLDR: 25%] 1 week ago:
Thanks for providing the actual numbers.
I think one of the more concerning things is, what if you think the answer is in the documents you provided but they actually aren’t. What you think is a low error rate could actually be a high error rate.
- Comment on How Much Do LLMs Hallucinate in Document Q&A Scenarios? A 172-Billion-Token Study Across Temperatures, Context Lengths, and Hardware Platforms [TLDR: 25%] 1 week ago:
Just for context, this is the error rate when the right answer is provided to the LLM in a document. This means that even when the answer is being handed to the LLM they fail at the rates provided in the article/paper.
Most people interacting with LLMs aren’t asking questions against documents, or the answer can not be directly inferred from the documents (asking the LLM to think about the materials in the documents).
That means in most situations the error rate for the average user will be significantly higher.
- Comment on Senators demand investigation after ninth American killed by Israeli settlers or soldiers in West Bank 1 week ago:
This is talking about American citizens killed by Israelis in Palestinian territory called the West Bank (Gaza is the other half). So isn’t about the war with Iran.
- Comment on ELI5. Limit of current gen AI/LLMs 2 weeks ago:
Yeah, that’s fair. I haven’t jumped into the whole agentic side of things as I find LLMs consistently fail at lower level stuff.
Everyone says it’s great at prototyping or writing documents, etc, but I think that’s just cause people have low standards. When coding I find that it quickly messes things up or lacks good quality control (which you only notice if you’re familiar with the domain). For writing it’s fine, but the tone and language always feels off and certainly doesn’t sound like me.
Either way, I would suggest playing around with them to see how they fit into how you do things. I think we’re starting to see things finally slow down on new implementations, and they aren’t going away, so it may be a good time to see if all the fuss is worth it to you.
- Comment on ELI5. Limit of current gen AI/LLMs 2 weeks ago:
The underlying issues, in my opinion, regarding LLMs is their indeterministic nature. Even zeroing out the temperature (randomness of outputs), you can get significantly different results between two almost identical texts.
However, building out an ecosystem supporting new technology is a fairly common progression. If you compare it to the internet things like browser caches, CDNs (content delivery networks), code minifiers, etc. are all ways to help combat latency (a fundamental problem for the internet).
As for the effectiveness of these solutions, RAGs do help a lot when generating text against a select corpus. Its what allows the linked sources in things like ChatGPT and Googles AI results. It’s also what a lot of companies are using for searching their support pages/etc. It’s maybe not quite as good as speaking to a person, but is faster.
Similarly, the reasoning models and managing the models “context” both have shown demonstrable improvements for models in benchmarking.
I’m not sure I personally believe this makes LLMs a replacement for humans in most situations, but it at least demonstrates forward progress for GenAI.
- Comment on ELI5. Limit of current gen AI/LLMs 2 weeks ago:
I think you may be mixing a couple of things together, but I’ll take a crack at this.
When you get an Ai generated response from a search engine, this is usually a modified RAG (retrieval augmented generation) approach. How this works is that the content from web pages are already pre-processed into embeddings (numerical representations of the text). When you perform a search, your search text is turned into an embedding and compared (numerical similarity) to the websites to get the most related content for your search. That means that the LLM only parses and processes a very small subset of the returned websites to generate its response.
Another element you might be asking about is how can these agentic AI systems handle larger tasks (things like OpenClaw). That is a bit more complicated and dependent on the systems design, but basically boils down to two things. The first is the “reasoning models” first break concepts into smaller tasks meaning the LLM only has to worry about a subset of a larger task. Secondly, a lot of these systems will periodically merge all past context into a compressed state that the LLM can handle (basically summaries of summaries) or add them to a database for future/faster reference.
At the end of the day, your understanding of the limits of LLM are correct, all the progress we’ve really seen with LLMs (over the past couple of years) has been the creation of systems to work around their limitations. The base technology isn’t getting much better, but the support around it is.
- Comment on Datacenters in space are a terrible, horrible, no good idea. 2 weeks ago:
Disabling/destroying a satellite has only been shown to be feasible by a handful of militaries in the world in very controlled situations.
Unless you mean you disable it via commands to the satellite, but that assumes there is a way to disable it and that you know who can disable it and can force them to do so.
- Comment on Datacenters in space are a terrible, horrible, no good idea. 2 weeks ago:
Yeah, that was my point. Like all technology it has potential to liberate communications, but also enable bad actors. However, to me, it’s the biggest reason why this technology would matter at all.
- Comment on Datacenters in space are a terrible, horrible, no good idea. 2 weeks ago:
I feel like on part no one ever mentions on things like this are, how do you enforce any jurisdiction on a satellite and what it’s doing.
The main crazy thing about a satellite data enter is you can’t confiscate it and therefore you can’t control it. Hell once it’s up there the only thing any government might be able to do is find the owner and force them to crash it (if possible).
It in a sense sounds a bit like the wild west of the original internet. Admittedly Musk being at the forefront of it all sounds terrible, but I think there is something fascinating about an information hub that could be completely independent of any country.
- Comment on $1,000 car loan payments are on the rise, stressing household budgets 2 weeks ago:
It was $51k for a used car.
If it was new I could understand, but that much for a used car is crazy. I’m guessing she had to buy when used car prices were almost the same as new, but I feel like that’s still a lot.
- Comment on Anthropic says it ‘cannot in good conscience’ allow Pentagon to remove AI checks 2 weeks ago:
What a company says and what a company actually does are not the same thing.
- Comment on Day 5 of posting an indie game I found that I think looks cool - Demon Tides 3 weeks ago:
I played the prequel (at least it looks like it’s tied to this game) called Demon Turf. The animation was really interesting and the world building was cool, but the platforming itself was only okay and some of the boss battles just felt broken. Hopefully they learned from that game to make this one even better!
- Comment on Xbox’s leadership shift proves it: the gamer era is over, AI runs the show now 3 weeks ago:
To be honest Phil’s just done a bad job at the helm or maybe wasn’t empowered to do the things that needed to be done. Everything since the xbox360 has been weak from Microsoft, although Sony has really only held on due to some banger first party games.
Nintendo continues to understand that you need to offer something different in the console space to remain relevant. While the steam deck (and other handheld) are coming for the switch, they at least offer something unique.
- Comment on Video games are losing the "attention war" to gambling, porn, and crypto, according to industry report 3 weeks ago:
I mean, at that point a lot of games are also gambling.
- Comment on Quantum teleportation demonstrated over existing fiber networks — Deutsche Telekom’s T‑Labs used commercially available Qunnect hardware for the demo, claims 90% average accuracy 3 weeks ago:
This is about “teleporting” information not physical material (if my understanding is correct)
- Comment on Favorite Mario Party... 3 weeks ago:
Yeah, 8 was the last one I remember enjoying. I think I tried one of the switch versions, but the boards were really boring. The fun things I liked in 8 was that the board events changed the map/pathways a fair amount.
- Comment on Man jailed for a year after endorsing neo-Nazi views and making antisemitic speech at Sydney rally on Australia Day 4 weeks ago:
I mean, I’d hope you’d have to prove intent to commit a crime, which would be the crime. Saying the words alone shouldn’t be a crime.
- Comment on Mewgenics becomes the most-played roguelite ever on Steam 4 weeks ago:
Let’s be honest with ourselves combocore is really just another phrase for rhythm games, which let’s you aggregate things like guitar hero (rythmncombos) and DMC (actioncombos), as it’s all about hitting buttons at the right time interval. Surprisingly many darksouls also fall into the combocore/rythmn category - sekiro and Lies of P (with their parrying/dodging timing) which we’ll now call soulcombos.
Now we just need to start promoting this through the steam tags.
- Comment on Mewgenics becomes the most-played roguelite ever on Steam 4 weeks ago:
And yet I tend to like most games that fall into the “rouge lite” category. It feels too broad and yet also seems to work to classify games.
I think it may just be a bit like “RPG” or “Action” that are actually very wide categories that now have a lot of subcategories to help better explain them.
- Comment on 4 weeks ago:
I tried Calibre web and Kogma.
Calibre is just bad software at this point, it’s clunky and not really designed as a server.
Kogma was fine, but a web only interface made it hit or miss. The big selling point for me with audio bookshelf was the ability to download local copies.
- Comment on 4 weeks ago:
Oh yeah, the multiple libraries is a good point!
- Comment on 4 weeks ago:
You can read using the web client or dedicated apps (android and ios). I feel like the clients work just as good if not better than similar software.
I haven’t tested how it handles two versions (audio/ebook) of the same book, but I have ebooks and audio books and it works well for me.
- Comment on 4 weeks ago:
The only real downside I’ve run into is it’s very opinionated about folder structures around authors.
- Comment on 4 weeks ago:
The one I’ve enjoyed the most is www.audiobookshelf.org, it may be “focused” on audio books, but works really well for everything. It also supports offline mode (meaning downloading local copies in the app).
- Comment on YSK that murder is now legal in California. Fully legal. Slaughter a family of four by driving over the speed limit. Go home. 4 weeks ago:
Yeah, everyone seems to be forgetting the secret to getting out of jail in the US - be rich.
- Comment on USA Used Claude AI Model to Capture Nicolás Maduro 4 weeks ago:
This seems like a lot of fluff to make anthropic/Claude sound impressive. They don’t state what it was used for, but I’d imagine it was just general purpose text processing.
- Comment on Trump nominates hospitality executive to lead National Park Service 4 weeks ago:
I’m sure that the guy is probably terrible, but at the surface level this isn’t the worst pick. Hospitality isn’t all the national parks are, but it is a significant part of the public facing side.
- Comment on 4 weeks ago:
Ah, if your preference is pixel graphics then Blasphemous 1 & 2 would probably be your pace if you haven’t played them.
To me, the GoW gameplay looked a little to heavy and stiff (especially for it being GoW).
- Comment on 4 weeks ago:
If you think this looks good, then the newly announced castlevania game will blow you away - www.konami.com/games/castlevania/…/en-us/