bigzyg33k

Surely you see how your anecdotal experience isn't sufficient to understand the false positive rate of these systems, right?

My 11 year old reddit account was banned a couple of hours into my engagement vacation to bhutan. After putting a local sim in my phone, I was first locked out of my account - I then passed their verification and unlocked it, just to wake up the next morning to an account ban. Their anti abuse systems are clearly very poorly calibrated for anybody who even slightly deviates from normal account access patterns.

That counts as 200% on exploitbench to me!

They already have a bright light that when covered, disables the camera

I believe they rewrite the chain of thought to protect their IP, i.e. the chain of thought reveals information about how the model works in a manner that may aid replication

we don't know if that's due to inherent limitations of the tokenisation of audio, or a byproduct of reinforcement learning. In my own usage, I noticed a significant degradation in capabilities over time from when they initially released advanced voice mode. The model used to be able to sing, whisper, imitate sounds and tone just fine, but I imagine this was not intended and has subsequently been stunted via reinforcement learning.

I don't find the articles argument that this is due to tokenisation convincing.

advanced voice mode operates on audio tokens directly, it doesn't transcribe them into "text tokens" as an intermediate step like the original version of voice mode did.

RAG certainly doesn't reduce hallucinations to 0, but using RAG correctly in this instance would have solved the hallucinations they describe.

The purpose of the system described in this post is OCR inaccuracies - it's convenient to use LLMs for OCR of PDFs because PDFs do not have standard layouts - just using the text strings extracted from the PDFs code results in incorrect paragraph/sentence sequencing.

The way they *should* have used RAG is to ensure that subsentence strings extracted via LLM appear in the PDF at all, but it appears they were just trusting the output without automated validation of the OCR.

do you care to elaborate or are we being mysterious today?

No it didn’t. You’re literally just making this up, I worked at RL

Based on your comment it's apparent you neither follow the industry closely nor understand it's dynamics. The vast majority of the billions of dollars are being pumped into R&D, not marketing existing legacy devices.

You also seem to be implying in your comment that the orion glasses displayed at connect last year were a last minute pivot, which is a ludicrous statement

Meta haven't abandoned the metaverse, and made it very clear from the beginning that "the metaverse" was something that does not exist, and will not exist in any form until the end of the decade. They continuously reiterate this during earnings calls, while increasing their capital expenditures on it.

You cannot determine it's a waste if the effort isn't completed, and if you have no insight into their progress.

I completely disagree. This is really just more of the great execution that I've come to expect from Sam Altman.

Core to OpenAI's strategy is that they control not just the models, but also the entrypoints to how these models are used. Don't take it from me, this is explicitly their strategy according to internal documents (https://x.com/TechEmails/status/1923799934492606921).

Some important entrypoints are:

- Entrypoints for layman consumers: They already control this entrypoint due to ChatGPT, the app. They have a limited moat here because they are at the whims of the platform owners, primarily Apple and Google. This is why they are purchasing Ive's startup.

- Entrypoints for developers: They acquired Windsurf, and are actively working on cloud development interfaces such as the new codex product.

- Entrypoints for enterprise: They have the codex products as described above, but also Operator, and are actively working on more cloud based agents.

A rebuttal that I anticipate to the above goes something along the lines of this: "If they have so much capital and dev experience, why are they acquiring these businesses instead of building internal competitors? This is a demonstration of their failure to execute"

The current AI boom is one of the most competitive tech races that has ever occurred. It is because of this, and particularly because they are so well capitalised that it makes sense to acquire instead of build. They simply cannot afford to waste time building these products internally if they can purchase products much further along in their development, and then attach them to their capital and R&D engine

Whatsapp has been audited several times, and to say that an engineer at meta “has been lied to” is laughable given he/she has access to the entire codebase

Source: also worked at meta and had full access to WhatsApps codebase, like most engineers at meta

Do you seriously believe nobody in the non English speaking world can write code?

Here's a (kinda) ELI5: you would use a language model to create "embeddings" of the text, which you can think of as a set of numbers representing the "meaning" of a set of characters.

These numbers can be plotted as points in a space, and embeddings of things with similar meanings are plotted close to each other. So things like "exam preparation" would have embeddings close to things like "top study tips".

Say you have created embeddings for a large corpus of text (in this case all youtube captions) once. If you create embeddings for a user query, you can search for embeddings close to it, and these will be "semantically" similar to the query.

The advantage is that unlike traditional full-text search, the user doesn't need a query that includes words present in the text.

TikTok’s audio is on by default

You’re making an error here with your user estimations - you’re using steam as a way of tracking users, and while the quest 2 is also the most popular headset on steam, most users with a quest do not have a gaming pc, and primarily play games stand-alone

They do, I got my current FAANG role from a reach out a year after I applied, and another FAANG company has been sending me emails for over 2 years now asking if I'd like to interview with them again.

when was this? I'm pretty sure facebook explicitly mention this in their preparation email now.

she's quite popular on tech twitter, and was/is followed by a lot of stripes c suite - musk was an initial investor in stripe

Also seems like a good place to plug two posts from insiders who say "the billion dollar code" was bullshit:

https://avibarzeev.medium.com/was-google-earth-stolen-7d1b82...

https://johnmccrea.medium.com/why-the-billion-dollar-code-is...

You can fix the delay for which key using by modifying the following variable:

(setq which-key-idle-delay 0.1)

But that isn’t how the feature worked

I don’t think you’re shadowbanned - at least, I can see your comment

They mention this in the earnings report

Advertising accounted for $25.4 Billion of their revenue, up from $17.4 Billion in the same quarter last year, an increase of 46%

Revenue from other sources was $732 Million, up from $297 million in the same quarter last year, an increase of 146%

This is incorrect, Americans use MM/DD/YYYY, and from my understanding most of the rest of the world (I think the Netherlands too) use DD/MM/YYYY

I may be wrong, but I think the photographer wrote that description - it's not unusual to see image descriptions like this on stock photography websites as it allows for more precise searches.

As somebody who has microdosed, I'd attribute most "enhancement" to perceived improvement in mood, and an improvement in my ability to look at problems from "outside" the box.

I'm currently leetcoding in preparation for tech interviews, and find it much easier to work out the "trick" to solve problems while microdosing LSD

WhatsApp/Text message is a really common form of marketing for dealers in London

I mean, I reckon he's well aware of this, given he was a tech lead at youtube