honzabe

Imagine that I am building a translation app: one way to do it would be simply a dictionary database where the Czech word "pták" is related to the English word "bird"- we certainly agree that this app does not understand what a bird is at all.

But when I say I do understand what a bird is, what does that mean? The way I imagine it, my brain is wired in such a way that when the word "bird" occurs, a certain pattern of neurons that encode stuff like "living creature", "flying", "feathers", "eggs", and "beak" also fire (each of those is itself a pattern of other associations).

In this sense, I think LLMs can "understand": connections between words "bird" and "eggs" or "flying" have weights closer to 1 while connections to words like "iron" or "multiplication" have weights closer to 0. A mental model of a bird can be represented as a set of vectors in multi-dimensional space.

Maybe I am missing something but I see no fundamental reason why we would not be able to encode complex mental models like chess or liberalism or flat earth theory as sets of vectors (and maybe even determine the set of operations that can move someone from flat earth model to globe model in a similar manner as those examples with 'King – Man + Woman = Queen').

Again, I am not saying I know how the human brain works. All I am saying is that I think we can plausibly explain a lot of it in terms of just networks of units that have some weights attached to them - it is a very powerful concept. That is why I am asking when people seem to be assuming that there is more to "legitimate intelligence".

Also - and I say this with a high level of insecurity because I am no expert - I do not think that LLMs actually play chess the way you describe it. If I even understand correctly what you mean - are you saying that LLMs are only able to play moves they have "seen" in their training sets? If that is the case, I would disagree - I think there is evidence that LLMs have some ability to generalize, which means the ability to work with abstract patterns (which I would be tempted to call mental models).

Out of curiosity, how do you think human intelligence works? I am no expert and I don't claim to know how the human brain works but my layman's mental model of thinking would be basically that: biological neural networks that "do" the thinking are token prediction machines, except "tokens" are not words that appear on the screen but thoughts that appear in consciousness... while the underlying machinery is in both cases a network of units that fire (or not) based on the connections between them.

Sure, human experience is a lot more than just intelligence (I have no mental model of how consciousness or qualia might work) but it is surprising to me that so many people keep repeating arguments similar to yours - that LLMs do not really think, they are "just" doing [X] (where X usually describes how I imagine human intelligence works) - if there is more to human intelligence, what do you think it is?

Can you please provide some sources backing up your claim? What metric did you use?

The metrics that I would use, like the Index of Economic Freedom by The Heritage Foundation, seem to contradict your claim.

The fact that the state owns some companies in key industrial sectors and has stakes in others does not mean that the country's economy is primarily in the hands of the state.

The world would look a lot different if this were something people do "automatically". No, it is laborious and emotionally difficult, and even smart and intellectually honest people who try hard fail at it often.

> The whole point of discussion and debate should be to decide which position to pick (because it's the factually correct one).

Something in this formulation reminded me of those times at university - I was going through some sort of crisis during the first year. In my country at the time, elementary school and even high school approached things kind of like "there are things you don't know yet and once you learn the facts, you will know the truth". University destroyed this for me. I started to realize how flawed our brains are and how complex reality is. There are facts that are and always will be beyond our grasp and we can only have approximations of approximations, simplified models that hopefully work well enough to be useful. It was surprisingly emotionally difficult for me to accept that fundamental insecurity and un-discoverability of some ultimate "truth".

I was not sure how this worked. I knew that the US does not have universal health care (not having it is one of those famous American things) but I assumed that it works on a private basis - you chose an insurance company if you want (and can afford) to be insured, you pay them $X per year and you are covered. Having this tied to your employment gives your employer terrible leverage over you. That would certainly stressed me out.

I live in a country where the unemployment rate is pretty low (2.5%) and the situation is kind of the reverse of what you are describing. Companies are trying to entice potential employees, not the other way around. Give me 6 weeks of paid vacations and do not even think about over-times; you look at me funny and I am gone... there are 10 other companies where I can start tomorrow.

I am now trying to relocate to Spain and it is not like that there - but I get why; because their unemployment rate is really high so employers have the power over employees. But in the US? I am sorry if that is a naive question - I've not been to the US since 2002. A lot has changed, apparently.

The term "hallucinations" is now commonly used for instances of AI making stuff up - like when I asked ChatGPT (before I had paid account) to recommend 5 books about a certain topic and two of the recommended books looked totally plausible, but when I tried to find them, I discovered there are no such books. This is where I see a big difference between GPT-3.5 and GPT-4.

>> I get a response that isn't grammatically correct/doesn't include punctuation

What punctuation? If you mean stuff like commas separating complex sentences, my English is definitely not good enough to spot that. But your mention of punctuation reminded me of problems that ChatGPT has with my native language... any chance you are using ChatGPT in a language other than English?

My experience - the free version made up a lot of things but still felt very useful - enough to want to upgrade to the paid version. With the paid version, I notice very rarely that it hallucinates. It does make errors but it can correct them when I provide feedback. It is possible that I just do not notice the errors you would notice, it is also possible that we use it differently. I would like to know.

After all, if I am to believe what Mercier and Sperber are saying in "The Enigma of Reason", reasoning evolved not as a tool to find the truth and make good decisions but as a social tool - to convince others to cooperate with you, to justify your actions in front of the others, to secure your social status. If your cultish following showers you with cheers no matter what you say or do (especially if you label your critics by that label that your followers hate - woke, ideological, racist, censors, or whatever), how can you not lose contact with reality?

That is the simplified version of my explanation for the phenomenon of previously really successful people saying and doing borderline insane stuff. I could name 10 famous people in the last 3 years that IMO fit that pattern.

If that is the case, what is the pleasure they get out of it? Porn triggers the pleasure centers that sex triggers... but astrology, what does that trigger? Sense of security that you "know" your future after reading your horoscope? Or what? Did the guy have any theory about that?

I have been watching Grand Tours since I was a kid; watching together with my dad was our family ritual and I am emotionally attached to it. However, this year I was a bit busy and I did not watch Giro d'Italia. But I found time to watch content made by NorCal Cycling and The Vegan Cyclist.

I know live-streaming a huge event like Tour de France is not comparable to making videos about a personal experience and publishing them after the race. But GoPros and drones made possible videos like [1] - as a cycling fan, I find that really exciting. A talented filmmaker and storyteller (like The Vegan Cyclist, in my opinion) can make a lot of bang for the buck with them. And what is especially enticing about this kind of content is the experience of "being inside" - personal attachment to the protagonist and a lot of footage from within the peloton. And pro cycling is not very good at it yet.

Maybe that was the hole the Netflix documentary was trying to patch? I have not seen it yet. I have seen the first two seasons of "The Least Expected Day: Inside the Movistar Team" and I think that was a good idea, although they should hire someone like The Vegan Cyclist to make it interesting (BTW, I have no affiliation to the guy).

It might seem crazy that I see events like Tour de France competing with "amateurs" but those amateurs might be able to attract a lot of attention for a dollar and I already see some money going in - like The Vegan Cyclist being sponsored by Canyon and a bunch of other companies.

[1] https://youtu.be/lRwGtPek1FQ

That claim does not ring true to me. Maybe a few years ago, but now? I don't use whatsapp and I don't know anyone who uses it, except for a group of expats from South America who probably use it because it's still widespread in South America and their friends use it, maybe?

It was recently discussed here at HN: https://news.ycombinator.com/item?id=36456233

TL;DR: it might be the 'Embrace, Extend, Extinguish' strategy.

I used to be a voracious reader when I was a kid. That book I mentioned - the Royal Road to Romance - I remember how smitten I was by that book when I was 14. Now it felt really slow (compared to YouTube shorts) and I had trouble getting into it. But once I did, the sparks of the old excitement appeared again. Completely different feeling than after 2 hours of YouTube shorts.

Books are great but I am sure that there is a lot of content on the internet written because someone genuinely wanted to say something (and BTW, not that books written as products are that rare either). It's just harder and harder to find it.

And I am only partially saying that because I have a new account here.

I was re-reading The Royal Road to Romance by Richard Halliburton recently. I love that book and if he lived today, he would be a travel blogger... so I tried to find a blog that would feel like that. And google search gives me more travel blogs than I can absorb, but they all feel like products.

I haven't had the time to go through blogs by HNers carefully yet but I hope there might be some gems in that pile. HN attracts a certain kind of people and if blogs written by them differ from the rest of the internet the same way HN itself, that would be great.

I like this idea very much. Thank you, the author of that original thread, and thank you, the creator of https://dm.hn

I started recently after a little cancer scare shook me up. It is about trying new things, re-starting my life a bit, I think. I plan to write for fun about things that I find interesting and that give me joy, like walking multiple paths of Camino de Santiago or my attempts to return to competitive cycling at 47.

It also exists in Spanish and Czech (I have international friends as a result of living in multiple countries): - https://crisis40.com/ - https://crisis40.com/cesky/

No ads or pop-ups or anything, very simple design, powered by Hugo.

But every time I read something like this, I am puzzled how they can have so much information and yet be so dumb. One example (but I could give you many): I like cycling and I subscribe to some cycling-related YouTube channels (Lanterne Rouge, Dylan Johnson - normal stuff, informative, bits of light humor now and then). YouTube keeps recommending me one cycling channel I absolutely hate - it is as if it was designed to piss me off. Sensationalist, negative, clickbaity titles, made up doping allegations every other post, throwing up emoticons in titles, creepy thumbnails with syringes and distorted faces. I never watch those videos, I clicked "do not recommend this channel" multiple times, now I just leave YouTube immediately every time this appears. And somehow, with all the data they have on me and all that sophisticated AI, they still have not figured it out. How? Any explanation?

I think it depends on the type of work, personality of workers and maybe other factors. For example, I used to work in a team where we were each doing separate part of the thing and it was not too beneficial to talk about work. Sure, we talked in the office, but mostly about movies, personal stories and stuff like that. I did a lot more work from home because the office environment was distractive and I get distracted easily. I can imagine that remote would be far less effective if I worked in a different department.

I think that is commonly shared definition. We distinguish lying from just being wrong. Saying untrue things in good faith, as Gwyneth Paltrow in your example, is not lying.

https://www.datawrapper.de/maps

2. Cycling - I love road cycling and there is a strong sense of community around cycling. If you are training alone and a bunch of faster cyclists catches you, you can draft behind them for a while and start chatting with them when you catch your breath; if you hit it off, you exchange contacts and start riding together. There are people that I have spent years and thousands of km with. The advantage of cycling is that there is a lot of endurance rides where you can easily talk and you get to know people pretty well.

First of all - good luck, I hope traveling will help you clear your head and find that change you are looking for. It worked in my case. Also, you reminded me of two other things journaling can help with - as you point out, it can lead to insight. Trying to formulate your thoughts clearly in writing can be very potent. And journaling can also work as a storage of good feelings - for example, I walked multiple paths of Camino de Santiago [1], each of them was a few weeks of great times... and I was journaling them. Every now and then, when I feel down after especially long and dark winter, I re-read those journals and they make me feel good. Áh, nostalgia strikes... I am gonna read them right now.

[1] https://en.wikipedia.org/wiki/Camino_de_Santiago

1. I write a training journal that consist of .fit files with extensive notes in .txt files - how I felt during and after the training (including things like "I discovered a new road next to a river, it was beautiful" or "I kept thinking about X and I was pissed), what I ate and how I slept that day, that I was observing swallows flying and that it made me feel good for some reason... writing a journal like this forces you to pay attention to little things. You notice patterns, maybe little signs of overtraining, signals that it is going well, connections that you might otherwise miss, like "I have elevated heart rate every time I eat poppy seed cookies"...

2. when I have something on my mind, I write it as an email for "someone close". This started during a difficult but interesting period in my life when I was actually sending emails to my close friend (we were on different continents and it was before Skype, mobile phones etc. - emails were the only connection we could have at the time). Nowadays I sometimes just write what is happening in my life without sending it to anyone but I usually still think about it as if I am writing to a friend. What is great about it is that this somehow re-frames heavy stuff into a funny story - it works even at the very moment when you are living through something tough; instead of just being depressed or scared you are already formulating in your head how you will write that email later and somehow your perspective changes into something a lot more positive - at least in my case it works like that.

Here is what GPT-4 says about this thread: Person 2 seems to have misunderstood Person 1's statement. Person 1 was suggesting that those who support Elon Musk and Peter Thiel, specifically the people who have totalitarian tendencies, would consider Peter Thiel as a "degenerate subhuman" due to his sexual orientation. Person 1 was not expressing their own opinion about gay people...

Interesting, right? One of my dreams is that I will soon have an AI detector for flawed reasoning (of others as well as my own).