How? How in the world you burn 500.000.000$ in a month?? I would LOVE to have a look of the distribution of tokens per employee. It needs to be at least 10 people who did the 95% consumption as soon as they saw no limits. Other theory is that they had Mythos available and prompted "hi"
A year or so ago, I started clicking “new” instead of the logo, which I usually click to get to the front page. I had no idea what was going on and just assumed HN had changed the algorithm and suddenly become much more active.
I think it took me between one and two weeks to realize I wasn’t actually clicking the front page.
That happens in most speech to text systems, even Superwhisper, Monologue and Wispr Flow. I read somewhere it comes from training on YouTube audio and happens when there is silence. I guess it depends on the model but most of them are based on Whisper which has this problem
Ha, I also have this happen all the time in response to mouse clicks. When playing with Apple Foundation Models + Whisper I noticed that it happens so often that I had to explicitly filter this out before acting on transcriptions.
Hackernews people will probably leave Google Search, but the rest of the world will surely stay with it for a few more years until we, the Hackernews people, can assure them that the information that those LLMs give is 100% the same as they would find in Google, but faster and to the point.
I think the next most significant milestone for this will be Siri and Google Assistant using this technology all the time. This, I think, will cause Gen Z to start leaving Google Search, and slowly, older Gens will follow.
Me personally? I can't handle Google Searches anymore. I started using Kagi and ChatGPT with the search module, but for things I need to be 100% sure about, I go to Kagi directly.
A little off-topic, but I love that the first paragraph describes the project. Usually, posts exclude that information, and the landing is not of more help.
My take on LLMs is that it won't scale much of what we already see because this is "just" text prediction on steroids. I'm not an expert by any stretch of the imagination, but that's my opinion, and going through that path, no, I don't see this path as the best path for "autonomous development machines", only powerful autocompletion like we already see today.
I agree. I've been using Copilot for several months now, and the only thing it (almost) consistently helps me with is predicting relatively trivial snippets.
Anecdotally, I've had it mispredict from very simple contexts, such as skipping numbers in series' where the pattern should've been extremely obvious.
I've had it sneak in sublte and obvious bugs on a regular basis, to an extent where I don't have much confidence beyond any code I can grasp in a single look and be confident it's correct. Sorry bros, I'm not on the hype train this time. Feels like crypto all over again.
There are many way how current LLM can be scaled in different dimensions and there are research around it e.g.:
1) Many different AI with different role: business analysis, tester, developer. You as developer are treated as customer and write simple prompt but business analysis AI will make a proper step by step prompt to Developer AI - so that you don't have to very good with prompt engineering
2) bigger context for LLM so you can feed up to date documentation and full repo
3) LLM having access to do RAG on web search to get up to date information
4) LLM having access to terminal and debugger so Tester/Developer AI can automatically see the flow how code is executed and variables states during execution
5) faster and cheaper LLM so that you give a task before you go to sleep and all those AI in a loop try to solve this task trying many different options until passed all tests.
Yes, but that doesn't change how they fundamentally work. You can replicate things fast, with many variations, almost like brute force but more "thoughtful." Don't get me wrong, this opens many possibilities, and I am even thinking about having a local AI machine for my stuff. However, working with multiple layers of knowledge connections is where I don't see LLMs arriving. Maybe some other technology is based on this, but following LLM evolution will be better data and "patches".
+3.5B views in a child-oriented channel. It is high, and even more so near Christmas. I would expect a higher CPM for that target, but maybe the category is saturated.
It makes sense. He loves stats and uses them a lot to improve his videos so with this, he can have a platform that can monitor the entire ecosystem.
Right now it's pretty basic but I would bet it will become more complex and have some paid membership for advanced stats and suggestions for your videos.
Let's please clarify the term "improve" and remove the positive implication -
What he's doing is making his videos "more effective at monetization" and that has fuck all to do with quality.
You may disagree with my tone here but I'm biting my tongue at the Little Bobby Tables view of children's attention and development (qualified Educator so I can talk shit about Mr Beast any day) so I'd like to keep it analytical as possible.
I very much appreciate Ed Bolian at VinWiki for his behind the scenes discussions about monetization. They are healthy! Because of those, I genuinely have tried to follow quality channels and bring them revenue via their presence (Ammo NYC, VinWIKI, TomleyRC) sitting through ads or maybe checking out sponsors. It's not that hard to be adult about it.
One of these days I hope we get to see a breakdown of where all Mr Beast's money is / was used, how it was sheltered, taxed, moved, or otherwise employed because that didn't just sit in the bank. Reference: Taylor Swift. Citation: Scott Swift FINA Report.
I wonder why you are downvoted.
I have no skin in talking about your second part (other channels + breakdown of MrBeast money), but I sympathize with your first part.
Optimizing for monetization seems to degrade quality most of the time.
E.g. click bait video names/thumbnails that do not tell the viewer what they are about to watch. Or reducing the feature set of apps so no casual user is confused, but advanced users are left behind.
Or stretching video lengths to improve ad income without actually adding any valuable content.
...
I suspect others downvoted due to the waffly style of the comment. What was that last Swift thing about? Reads like someone with multiple bees in their bonnet
Of course, if you take "improve" to your terms. For MrBeast, it is to have more views (other than monetization itself); he deliberately works towards it and is public about how he uses YouTube tools to get more views and retention. So, I think my comment is on point.
Other creators will fight for different ways to improve their content, which can technically be 4K HDR or having better guests, etc.
Do youn't like Mr. Beast's "empty" content? Then you don't like society, which is the one taking MrBeast to the top, and I would completely agree with you.
Do you think MrBeast is lying when he says that his videos on his main channel lose money because of the production costs and that he just puts the money he makes back into making more videos?
Yeah right? If people don't want viruses, just don't download them!
My point is that ordinary people don't know that TikTok could be even remotely bad for them, and when they do know, they will think that it is not that dangerous.
And Apple/Google dialogues about data transfer and all that will be dismissed by the user the second its pop-ups.
The user I am describing above is "the crowds". That's why TikTok is so massively used even though their security problems.
Isn't the onus on the user to assume risk? If they wish to have an accurate perspective then they have ever opportunity to do so. Sorry, if the contrary is true, and markets don't result in the best of all possible worlds then we have to conclude that free market capitalism is not going to work. The two are inexorably yoked together.
