titaniczero

Yeah I can’t fully agree with axioma #2. While people learn when you challenge them, every person reacts differently when they struggle. And there is this thing, anxiety, that makes people insecure. Mathematics anxiety is even a thing [1].

So I think this axioma is based on a previous one: everyone wants to be challenged, which I believe is not true. You have to give people a reason for that motivation first, to be able to challenge them. Otherwise you will make people feel insecure, because they will think that they’re not smart enough, and we are taught that maths = smart. (Maybe we have to tackle this social construct first.)

——-

[1] https://en.m.wikipedia.org/wiki/Mathematical_anxiety

Well, my guess is that updates and inserts are much more frequent than searches in their use case. You're assuming a balanced frequency for these operations and it hardly ever happens.

ipv4 space is very limited and you can easily compute all the hashes. There is salting and combined with rotating salts it could work but no one guarantees that you’re not storing them

But the idea that the brain functions on a basis of complex statistical processes doesn't imply that statistics created the brain itself, it just suggests that the brain's processes can be modeled/understood through the lens of statistical methods. Statistics is just how we describe a tool we invented to analyze observable data, the brain could be a similar tool and it doesn't need to be the same tool.

It's akin to saying that we created the concept of "physics", yes we created it, and physics govern, for example, how a car moves, but it doesn't mean our concept of physics created the car or physics itself, we just use physics to describe and understand the car's movement.

Maybe statistics can't generate anything, but if you imagine everything we can do as very complex, unimaginable multi-dimensional functions that can generate outputs based on inputs, we can use statistics to find functions that fit any real (ground truth) function in the observable universe.

When you fine-tune it, do you train just the head/last few layers or do you also unfreeze the model afterwards and retrain the whole model with a very small LR for a few epochs?

10% of the population in spain was using netflix before this, as a reference 12% in germany, not that much difference. We'll see, but I think it has more to do with the decline in quality of the content and people sending a message than anything else.

My experience is similar but tbh I feel it's mainly due to the fact that oauth is implementing an important security layer (authentication/authorization) and that's hard by definition. There are lots of steps and things you can't afford to overlook or you'll be vulnerable to multiple different attacks

It’s probably provided as system instructions for rejecting things. You can use the API and feed it with different instructions with the system role

Because I was in a hurry and I don't have unlimited time, while providing info about the past that I already have was still valuable for the point

IMO if your app is less secure than your competition by default, the app is less secure, period.

Telegram is said to have been given authorities access to user data [1], despite the fact that they advertise the opposite. I guess that’s what happens when your app is not encrypted E2E by default.

Also, they have used their own encryption algorithm in the past (I don’t know now) instead of the well known and proven algorithms out there. Something highly criticized by experts, back then [2]

-[1] https://www.androidpolice.com/telegram-germany-user-data-sur... -[2] https://eprint.iacr.org/2015/1177

Why? Genuinely asking.

As a software engineer this all is very inspiring. I’m tickled and excited to do things again. I’m also learning deep learning and statistics and having a blast with toy projects.

I’ve noticed that bing chat is better if you instruct it not to search anything, that way it will use the model knowledge. I’ve learned to use the model knowledge or the web search results summary depending on what I want. But ChatGPT is still way better for model knowledge because it has fewer restrictions.

I wish they would make this distinction clearer in the UI. Most of the time it can answer without resorting to search, I think it would be better if the user explicitly specifies that they need web results.

OpenAI should implement an oauth authorization server and allow developers to use "Login with OpenAI account" into their apps.

I would be more concerned if we don't get to the point of decent local models for every task, because of the power these companies would have

> Actually nowadays I bet higher percentage of Swedes know English than know Swedish.

Is that hyperbole? Because that would be very sad if true, culturally speaking. There are 5 different official languages in my country (Spain) and there are several laws to prevent them from becoming extinct for this very reason.

> the EU has proved itself to take consistently anti-big US tech stances

Is that how you call trying to protect consumers over there? Really, there are genuinely good reasons for those "stances".

I fail to see how ChatGPT would help here.

It's only useful when you have some cognitive work that you want to automate so it saves you time and you can do it on a large scale, but I don't think writing massive amounts of viruses is the goal here. The objective is to code very specific and particular viruses or exploits for very specific targets.

Also, writing code is only a small part of hacking, especially when it is against specific targets.

The idea is interesting but I still don’t get how it works.

Also, what are the benefits for me, as a user? I could give you ideas for the challenges but I’m missing a “search” or “explore” feature to find solutions for my problems in an engaging way that would encourage me to contribute my own

Jeez, I don’t think I’ll ever understand the american thing with Apple. It must be a social thing, I’ve seen people humiliate and even shun people who have Android (“the green bubble”) or other non-Apple devices.

I feel that they're always very biased in these comparisons.

Yeah, but to be fair GPT 2 is open source and Whisper (a high-quality speech recognition and multilingual translation model) is also open source. A few years ago I needed a good model for transcription for a project and I couldn’t find anything decent. They really have contributed to the open source community.

If they keep releasing older models and keep their cutting-edge technology for profit I’m fine with it.

They are language models, it's what they do best. They "understand" (give more weight) to the correct relations between words (tokens) and try to predict the next token based on previous tokens. So when you feed it with instructions, for the next tokens the model will give more weight to the tokens related with those instructions.

On the other hand, they can't handle actual logic, reasoning, etc.

Objectively speaking, how valuable would that information be if it were to be leaked? Random weekly aggregated and sampled machine data associated to random IPs that in most cases would no longer be yours (and that’s in the worst case that both servers were compromised or misused by an employee during a specific time window, because IP is not stored in the analytics server but temporally in a proxy to prevent abuse).

I don’t know, I’m all for privacy but people get passionate too easily the moment someone even mentions the word telemetry and I think we should focus on actual privacy issues, like analytics in the context of ads, social networks (tiktok, fb, ig…), etc.

The atmosphere now has borders?

I have thought about this but sometimes I reuse the variables on purpose to reuse the memory allocated, especially in a language like Golang where sometimes it is not that clear whether it will be allocated in the stack or the heap. I guess we should avoid doing this in hot paths, right?

Yeah I agree with that one. They set up a call and he stood by his decision to disclose it on oct 15th. Then 3 days before the disclosure deadline they rewarded him.

Yeah, they tried to set up a call to dissuade him but he stood by his decision. Then only 3 days before the disclosure deadline they decided to pay out.

They ended up rewarding him with $70,000 tho

> This way too I can troll people on the internet when they suspect this is happening and I can say "bUt ThE bAtTeRy LiFe!" to defend Meta: my corporate overlord business daddy.

Please, stop with the sarcasm.

Okey, let’s say they manage to record us without a huge impact on our battery life. Now, how do you send these recordings or even the extracted keywords from a popular app, a client installed on devices controlled by the users and susceptible to reverse engineering and network traffic analysis without anyone noticing it?

It’s just too much risk and they don’t even need it, see my relevant reply here: https://news.ycombinator.com/item?id=32950204#32953216.

Try any voice recording app for a few hours, now use the facebook app for the same number of hours. The impact on battery life of a mic actively recording alone is very noticeable, so noticeable that your phone has a special chip just to recognize patterns similar to “hey siri”.

> Considering how easy it is to implement these things without anyone noticing since it's closed source

You can reverse engineer those things and analyze your network traffic. You can’t have a client in a device controlled by the user, in this case an app, send anything to a server without anyone noticing it.

And frankly, they don’t even need it. Just with your contacts they can link you to your friends and common interests without even you having a facebook account, all you need is friends with a fb/ig account who have linked their accounts to their phones and use whatsapp.

The contacts are known to be sent to the server, they are known to be linked to facebook except in the european union where there is a different app from WhatsApp Ireland and a different privacy policy that specifically states (in the version outside of EU) that it shares your contacts with facebook and they are much more valuable and much less risky than reading your messages.