ankit219

24 ms·

271.
Apple allowing 3rd party payments in Netherlands w commission(2022)(daringfireball.net)
1 points by ankit219 3y ago|0 comments
272.
by ankit219 3y ago
I have never fully agreed with this premise. On the face of it, yes apps help people stay on the store. But at the same time, Apple does come up with many good apps. Comparing it to the first iPhone is probably valid. They saw the writing o
273.
by ankit219 3y ago
Articles like these always follow a predictable script: > Compiled using data from the research company Wealth X and Forbes, it says the combined wealth of the top five richest people in the world have increased by $464bn, or 114%. Over
274.
by ankit219 3y ago
Easy - just compare the size of the list from 2020-21 to 2023. > 2,640 billionaires with a total net wealth of $12.2 trillion in 2023 > 2,668 billionaires with a total net wealth of $12.7 trillion in 2022 > 2,755 billionaires with
275.
by ankit219 3y ago
This version of teams does not do that. You can hook it up by creating Custom GPTs and add some amount of docs to a specific GPT for retrieval, but you cannot connect an entire codebase to ChatGPT to get answers. Github[1] had introduced th
276.
Clark Valberg (Ex-CEO, Invision) on why Invision is shutting down(twitter.com)
3 points by ankit219 3y ago|0 comments
277.
by ankit219 3y ago
Not a lawyer, but seems like the law on copyright and fair use is not that clear cut. It comes down to arguments and how well they resonate. The lawsuit here argues all possible sides - producing content verbatim, summarizing an article (wi
278.
by ankit219 3y ago
I stopped my ChatGPT plus subscription at the start of December. I did not use the interface that much so was never maxxing out the limits. I have instead switched to a combination of Perplexity for knowledge based questions (since there is
279.
by ankit219 3y ago
That can be done if we use Imagebind from meta(embeds text, image, video, audio in same vector space). I would want to explore this if possible just for a POC if you are okay with it. Would you be interested?
280.
by ankit219 3y ago
Some caution here. Not everything needs to go into a RAG pipeline (Eg: a database table would not necessarily need to be embedded, but its schema should be.). There would be a lot of repetitions, lots of junk and useless data, and numerical
281.
by ankit219 3y ago
I think the answer depends on how many documents you have. To think in terms of tokens (assuming 750-1000 tokens is a page), if you have a good estimate of number of pages you want to query on, you can decide on the approach. Three popular
282.
by ankit219 3y ago
The entire model was initially opensourced by a 16 year old. https://github.com/JJTech0130/pypush They modified it with their own touch.
283.
by ankit219 3y ago
I can relate. I have had conversations about enterprise search and how it can help them especially when done with the help of embeddings + LLMs, but many do not see it as a problem. It's a classic case of people you would be selling to
284.
by ankit219 3y ago
It's pretty clear why they don't want an android iMessage app. In this case, what beeper enables (if successful) potentially is to use Apple's infra for future communication between android to android phones, or android to iM
285.
by ankit219 3y ago
It was open sourced by a 16 year old apparently. https://github.com/JJTech0130/pypush They used this and added their own changes. From their communication about what they are doing, it's remarkably similar, and i
286.
by ankit219 3y ago
I think they ignored a rarely talked about but important aspect. iMessage is free for Apple users because it comes bundled with all Apple products. The cost to run iMessage and deliver millions of messages daily must be a significant number
287.
by ankit219 3y ago
> Are you using a normal training script i.e. "continued pretraining" on ALL parameters with just document fragments rather than input output pairs? Yes, this one. > do you make a custom dataset that has qa pairs about that
288.
by ankit219 3y ago
Not using RAG, and using Supervised Finetuning post pre-training. It's taking all of the corporate data and pretraining a foundational model further with that extra tokens. Then SFT. Problem with usual finetuning is that it gets the fo
289.
by ankit219 3y ago
You could broadly segregate the market into three groups - general purpose, specialized-instructions, and local tasks. For general purpose, look at the uses of GPT4. Gemini might give them competition lately, and I dont think OSS would in t
290.
by ankit219 3y ago
It's all murky because the reporting is done in a biased manner. Tesla has service centers and point of sales in Sweden. Sweden has no minimum wage because everything is driven by CBAs (they even lobbied to CBA supported unions from mi
291.
by ankit219 3y ago
They can recall if the seller requests a recall for any reason. Not the customer. Typical use cases include wrong product sent, wrong packaging, missed an item etc. It's a rare scenario but it has happened to people I know. From a cust
292.
by ankit219 3y ago
From the trial itself, Google has argued and questioned the expert witness from DOJ if defaults really matter. Whole Microsoft question was raised by them too. Exhibits are good with context, but without supplementing the context they are m
293.
by ankit219 3y ago
It's a widespread assumption that defaults matter. But to what degree? I came across it first when I saw Dan Ariely's talk and his work on defaults (may not be the proponent)[1] but my takeaway is that it matters when you don&#x27
294.
by ankit219 3y ago
Don't PG and Jessica have an investment in OpenAI? I think YC also counts it as one of investee companies. Jessica was one of the founding members. (based on a bit of googling) If openai were a public company, the amount of shilling YC
295.
by ankit219 3y ago
They showed a demo where you could upload a file while creating an agent. As others have answered. I think it's about configuring an agent in a way that you give it material on some specific topic (one file, multiple files) and it uses
296.
by ankit219 3y ago
The author of the thread forgot to mention the deck is from 2008. I was reading one of the tweets in his thread about funnels. Google has argued in court about funnels is outdated, and he read this deck and concluded that Google is lying. W
297.
by ankit219 3y ago
I am with you on that one. You don't have to have an opinion so close to an event unfolding. Ideally, we should be able to wait, let the authorities investigate, and then finally get the findings to be able to make an opinion. But toda
298.
by ankit219 3y ago
I think I probably am giving them an unnecessary pass. There is another case in the below links about FB ads where Facebook banned an advertiser's account because of usage of words "panda" and "python". While that c
299.
by ankit219 3y ago
I guess it's a fine line. If Google starts pre screening ads and then bans somewhat legit ads - which may or may not look as legit - then we will see reports about them. Moderation is a tough problem to solve. We want convenience with
300.
by ankit219 3y ago
The article is full of hyperbole. > it's still too early to tell if U.S. District Judge Amit Mehta will side with the Justice Department and try to handcuff one of the world's most dominant tech companies. Till this point, only

More ›