Hacker Newsnew | past | comments | ask | show | jobs | submit | toinewx's commentslogin

> 1. There is no evidence this is AI generated. The author claims it wasn't, and on the specific issue you cite, he explains why he's struggling with understanding it, even if the answer is "obvious" to most people here.

There is, actually, You may copy the introduction to Pangram and it will say 100% AI generated.


That's not evidence, at least not evidence that would stand up to a peer review if the author were to refute it.


It's not proof but it's definitely evidence.


When I give my own writings to Pangram, it says 100% human.


What’s the evidence that it is human-generated? Oh I see. If it is AI generated then you still have to judge it by its merit, manually. (Or can I get an AI to do it for me?) And if they lied about it being human-authored? Well what if the author refutes that accusation? (Maybe using AI? But why judge them if they use AI to refute the claim? After all we must judge its on its own merit (repeats forever))



It's quite obvious that OpenAI benefited from training on Ghibli material.


That's not relavent. You can't sue someone for benefitting from your work. You can sue someone for intellectual property theft, which is what OpenAI did.


there is little mourning for people who didn't get hired at all in the first place.


is being good at chess a proof of human intelligence?


I think there's some correlation but no, you can be amazingly strong at chess and not super-intelligent by other measures, or be incredibly good at physics or philosophy or poetry or whatever and have no talent for chess.

But, having watched some of Naroditsky's videos, it seems pretty clear that he was in fact very intelligent as well as very good at chess.


Maybe, exactly what are you asking? What even is intelligence - that is a question I've never seen formally answered (and even if someone does, their definition may not match your intuitive feel for what it means and thus it is useless outside of the exact paper they defined it in)

Formal definitions aside, it isn't possible for "stupid" people to be good at chess. There also is no other animal or known alien that is good at chess. Thus being good at chess is a strong sign of an intelligent human.

We can't go the other way. There are plenty of humans generally known to be "intelligent" who are not good at chess. There is a lot more than intelligence needed to be good at chess (practice and study come to mind, there might be more).

While there are no known aliens that are good at chess, that doesn't preclude that we may discover them in the future. (not in your lifetime though - the speed of light is too slow for them to learn the rules of our chess and communicate back proof that they are good, no matter how intelligent they are)


Being good at most things is correlated with intelligence, including activities like music: https://pmc.ncbi.nlm.nih.gov/articles/PMC6682658/


The ability to plan and operate several moves ahead of one’s opponent has always suggested higher intelligence.

When applied to war we celebrate the general’s brilliance. When applied to economics we say they had excellent foresight. When applied to any human endeavor, except chess, the accomplishment is celebrated as a human achievement.

This is due to humans placing great value upon thinking and planning ahead. Only the intelligent exhibit this behavior.


I think Hikaru’s fans made him take an IQ or intelligence test a couple of years ago and it showed he wasn’t exactly a mastermind. He said at the time something along the lines that being good at chess shows you are intelligent only in that one domain of chess-type thinking, not general intelligence.


I remember something related to that, but wasn't he trolling the complete thing?


If you've watched a lot of streams or interviews with Hikaru, he gives little indication of being a big brain. Or even an average brain.


I used to watch a lot of his streams/videos, but I always thought he was just not taking himself seriously, being entertaining, "memeing and trolling". I thought it was his strategy to do unfiltered ADHD thoughts even when they don't make any sense, because that's what brings him viewers.


> being entertaining

Were you entertained when he publicly dissed his pregnant wife for her play in the recent US Championships?

I get it that lots of chess nerds enjoy this kind of thing, but that doesn't make it intelligent.

Or if you like tests ... he himself has mentioned not scoring well on an IQ test.


I suppose he could be charactarized as the xQc of Chess, perhaps?


Intelligence has been defined in many ways: the capacity for abstraction, logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving. It can be described as the ability to perceive or infer information and to retain it as knowledge to be applied to adaptive behaviors within an environment or context.

(from Wikipedia)

Intelligence is multifactoral. Being good at chess was one aspect of intelligence in the complexity of Daniel's life, and in anyone's life.


Intelligence presents in different way, but they’re all correlated with each other. The idea that people can be intelligent in different things is mostly a myth: https://researched.org.uk/2018/09/26/myth-busting-gardners-m...


What makes this guy's blog post a more authoritative reference than 1000 other sites, as well as everyone's personal experience of feeling naturally "good" at some tasks and "bad" at others?


From that artcile:

"'Even in our book, we don’t want to call this theory a complete myth, but instead label it as ‘nuanced’."


But read the next two sentences:

> Why? Well, the basic idea behind this theory is that people are different, and maybe you’ve noticed – they really are. People have different interests, different abilities, different moods, etc.

The author isn’t saying that the multiple-intelligence theory is itself valid. Rather, in an educational context, there is a kernel of value in the idea that different students are different. That’s entirely consistent with intelligence being a single thing.


Not proof (if there is any such thing for intelligence) but a very strong indicator, yes.


It definitely is, considering "good" to be a top 10 blitz player like Naroditsky was.


Not that long ago I would have said no, but I increasingly think that intelligence is mostly about the ability to learn. And chess, at a high level, requires a mixture of achieving an extremely high degree of unconscious competence working right alongside a high degree of conscious competence for things like opening prep. And most, if not all, high level chess players that have tried, in earnest, to do things outside of chess have excelled across a wide variety of fields.

But I think where people get confused is in the inverse. If you take a very smart person and he dedicates two years of his life to chess, all alongside training from some of the world's best, then he's still going to be, at best, a strong amateur at the end. In fact I know at least one instance where this exact experiment was tried. This is generally unlike other fields where such an effort would generally put you well into the realm of mastery.

But that's the unconscious competence part - chess takes many years of very serious training to even start to have it become 'natural' for you, and it's that point that your training journey begins all over again because suddenly things like opening preparation starts to become critical. So it can give the appearance that since seemingly smart people don't do particularly well at chess, while people like Magnus who has/had (daddyhood changes a lot...) a complete 'bro' personality, is arguably the strongest player of all time, it gives the impression that being smart must not be a prerequisite for success at chess.


The education system bored me a lot and made an effort to portray me as some kind of mentally disabled retard. It was rather interesting to me that successful career grown ups couldn't win a single game.

I wasn't interested in chess but I could see their entire plan unfold on the board. Unless they were actually good I didn't even try to win, in stead I let them unfold their plan into their own demise.

My winning streak ended when I got to play against the best kid from a different school. His was the biggest brain I have ever seen from the inside. He pretty much violated basic principles of the game in a way that still bothers me 35 years later.

The game was much to open to really look far ahead. The way one would play against a computer. His actual goal was to trade his knights and bishops for two pawns each!?!?! He pulled off 3 such trades. He carefully set up the trades and it made no fkn sense.

Then came a really long and slow pawn push that didn't allow me to trade my knights and bishops for more than a single pawn.

It took so many moves that he pretty much convinced me that 2 bishops and a knight are worth less than 5 points. I haven't seen a second game like it but I'm still 100% convinced.


If we define it as such (which it seems we have in some way) then yes.


nothing is, we're all stupid, thanks for the reminder


I like turtles


yes, it's a proof of human intelligence at chess. nothing more.


I'd say the opposite. Every time a guy crushes me, I flip the board over and call him a fuckwit.


the name is obviously a pun/joke, and the website is also trying to be humorous.

but I don't think it helps your product by now. It looks you built something that is actually useful and the pun/joke is not needed anymore. same for your website, I think the humorous part is distracting.

I think you should pick a unique name, the joke is no longer needed.


I don't disagree necessarily. I also love the name with all my heart


You should keep it if you love it, but understand that the name is limiting the project's reach. Not all projects are aiming for reach, but this appears to be a serious useful tool that you're charging for, not some toy. The name is stopping people from understanding the project and is limiting your income.


Why do you love the name?


Because computering is missing out on whimsical fun in every way nowadays!


Can somebody explain the joke to me? Isn't notepad.exe just the file name of the notepad executable on Windows?


Yeah I just found that title confusing


The element that provokes the most laughter is the punch hole in the MacBook screen.


with such a high price, isn't it better to use the chatGPT plus membership instead?

but this is assuming they won't nerf the chatGPT version


For a regular consumer? Definitely. But it's an API - different use case.


yes this is great but I'd like to pick a different voice. the current one feels too robotic


Same, it was using the high quality openai voice until my account ran out of funds.. Now it's using edge-tts which is free. So far it seems like the best option in terms of price/performance, but I'm happy to switch it up if something better comes along.


The gemini example was a wonderful summary of the comments, but audio is not very practical for something that long.

What about putting the text version that's used to make the audio somewhere on the page? (or better, on a subpage where there's no audio playback)


I'll look into it for the next iteration! I could just take the transcript that's already on the page and put it somewhere separate from the audio.

But thinking about it a little more, what would the use case for a text version actually look like? I feel like if you're already on HN, navigating somewhere else to get a TLDR would be too much friction. Or are we talking RSS/blog type delivery?


would you be able to run Sonnet 3.7 on a consumer computer though?


You can't with DeepSeek either but it has aided open source models significantly by being open source.


The new M3 Ultra Mac Studio (512GB version) seems to be capable of running DeepSeek R1 in Q4


can you expand a bit?


The model performs very poorly in practice, while in the benchmark it is shown to be DeepSeek V3 level. It's not terrible but it's at another level compared to the models it is very close to (a bit better / a bit worse) in the benchmarks.


I’d recommend trying it on Google AI Studio (aistudio.google.com). I am getting exceptional results on a handful of novel problems that require deep domain knowledge and structured reasoning. I’m not able to replicate this performance with Ollama, so I suspect something is a bit off.


Same experience here: On AI Studio, this is easily one of the strongest models I have used, including when compared to proprietary LLMs.

But ollama and openwebui performance is very bad, even when running the FP16 version. I also tried to mirror some of AI studio settings (temp 1 and top p 0.95) but couldn't get it to produce anything useful.

I suspect there's some bug in the ollama releases (possibly wrong conversation delimiters?). If this is fixed, I will definitely start using Gemma 3 27b as my main model.


Update: Unsloth is recommending a temperature of 0.1, not 1.0, if using Ollama. I don’t know why Ollama would require a 10x lower value, but it definitely helped. I also read some speculation that there might be an issue with the tokenizer.


Maybe model is sensitive to quantization, by default ollama quantize it significantly.


I tried ollama fp16 and it had the same issues.


Hey, Gemma engineer here. Can you please share reports on the type of prompts and the implementation you used?


Hello! I tried to show it Redis code yet not released (llama.cpp 4 bit quants and the official web interface) and V3 can reason about the design tradeoffs, but (very understandably) Gemma 3 can't. I also tried to make it write a simple tic tac toe Montecarlo program, and it didn't account for ties, while SOTA models consistently do.


Can you share the all the recommended settings to run this LLM? It is clear that the performance is very good when running on AI studio. If possible, I'd like to use the all the same settings (temp, top-k, top-p, etc) on Ollama. AI studio only shows Temperature, top-p and output length.


vibe testing, vibe model engineering...


I really respect the work that you've done, but I am always very surprised when people just speak anecdotally as though it is truth with regards to AI models these days. It's as if everyone believes they are an expert now, but have nothing of substance to provide but their gut feelings.

It's as if people don't realize that these models are used for many different purposes, and subjectively one person could think one model is amazing and another person think it's awful. I just would hope that we could at least back up statements like "The model performs very poorly in practice" with actual data or at least some explanation of how it performed poorly.


In my experience, Gemma models were always bad at coding (but good at other tasks).


Prompt adherence is pretty bad from what I can tell.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: