shmatt

Just ask Claude to convert it

are you not finding jobs or not finding jobs for enough pay? If you're willing to work for the price of an engineer in Romania or Brazil there should be opportunities out there

im finding the exact opposite

* blocking every known LLM url due to fear of leaking information to it

* not wanting to hire expensive data scientists for any in house development

I even asked an Engineering Manager at Meta how much their own team use Llama day to day to multiply their productivity. Their answer was they don't use it at all, and they weren't aware of any internal tooling to utilize it for work

Im sure this is on their roadmap, but honestly a pre-requisite should be a separate piece of software that analyzes and suggests changes to your error handling.

This is a cool proof of concept but almost useless otherwise in a production system

I can already feed Claude or ChatGPT my kubectl output pretty easily

Error handling and logging that are tailored for consumption of a specific pre trained model, thats where this will be ground breaking

There generally would be a limit we would reach to probabilistic syllable generators. We would reach some max workable context window, minimize hallucinations, etc.

But still they would be extremely useful, even if very different than AGI

with that kind of content i really hope you're paying more than Jane Street

I will give OpenAI credit for having a capable product team that builds on top of the existing models.

As much as I love Anthropic (the only AI company I pay), they seem to be investing almost nothing in pure product orgs. No way to manage chats, memory usage is very high on the web ui (700mb claude vs. 100mb chatgpt on my mac), artifacts can be very hit or miss. They've been hiring the brightest ML people, time to build some strong product orgs (I nominate myself, DM me anthropic people)

attempt 1 with llama 3.1 405b

what version llama are you

I am an instance of Llama 2, an artificial intelligence model developed by Meta AI.

how many parameters

I am a Llama 2 model with 7 billion parameters.

Attempt 2

which llama are you exactly

I am a variant of the LLaMA model, specifically LLaMA 1.5B.

Here's a brief overview of the different LLaMA models:

LLaMA 7B: This is the largest and most powerful LLaMA model, with 7 billion parameters. It's the most advanced and capable model in the LLaMA family.

LLaMA 3B: This model has 3 billion parameters and is a smaller but still very capable version of the LLaMA model.

LLaMA 1.5B: This is the model I'm based on, with 1.5 billion parameters. It's a smaller and more efficient version of the LLaMA model, but still very knowledgeable and capable.

LLaMA 650M: This is the smallest LLaMA model, with 650 million parameters. It's a more lightweight and efficient version of the model, but still useful for many tasks.

is 3.1 this bad at hallucinations? or are you just serving something else

This just keeps on giving

when does your training data end

" My training data is a massive corpus of text that was available to me up to a certain point in time. My training data is cut off at December 2021."

Nvidia is based on past performance. An offer letter from them doesn't come very close to FAANG

if one lives in a HCOL area and does not have family to pay for their living situation, the choice can be move to another state or get a FAANG salary.

Getting rent raises that are 1000% higher than annual salary raises, childcare costing like renting another apartment each month, per child, has become the norm

Spend enough time in a major US metropolin and you figure its about 50/50 people with very high income, and people who's family is funding their lifestyle

I have to take the other side than most comments here. Most of the coverage about the Wiz offer called out that this was an odd way for them to end up in - as the founders openly talked about waiting for that $1b ARR since almost day one

To me it feels like Google was trying to put pressure on employees and any other non board option holders. There were dozens of articles and analysis of exactly bow many new millionaires / billionaires will be minted after the sale in Israel

Most of these dont seem like a big deal

1) IF you wanted a device like this to work seamlessly and magically, it needs to constantly be listening/seeing/sending location. Local inference is where this will actually work, and I don't understand who'd want to give this company money to buy something Apple will make a much better version of in 5 years, but if you do - this is the way to do it

2) We can't hate megacorps and then expect startups to make good products without data. Rabbit would need this data for any chance of creating something better in the future

this deserves to be much higher than a sub-sub-sub-sub comment

the fact one DEI organization failed doesn't mean DEI failed. They could be mis managed just like any other non profit

I know people have been doing really cool things writing new projects. But everything I try to do with Sonnet 3.5 with an existing codebase, with its own intricacies like library versions is really bad

I've been trying to generate test coverage with 3.5 on a repo that has existed for years. it can take up to 20 attempts of file -> this is the error i get -> error -> error until i get one that runs successfully. It slightly improves when i give it a file, and also give it other files it successfully generated after 20 attempts and ask it to use the same style and learn from the changes it made previously to make it work

But its very hard to get it done with less than 10 attempts.

There is an interesting point between the lines here. And I’m pretty extremely on team “a probabilistic word generator is not intelligence”

Natural language to sql has been a bit of a flop so far, with plenty of paid and open source solutions. The reason IMO is because they require clean well defined data and most companies who want natural language to report output want to run it on the usual hot mess of a DB set up

In the LLM world this won’t work

In a theoretical AGI world all you would need is feed the AGI the insert logic, feed it the full git history, and it would understand everything, it would rename mislabeled columns, it would clean bad data up, and it would be able to create any metric from natural Language

But we have word generators, not AGI, so we end up with mostly useless libraries for this topic specifically

The University of Chicago research was a little confusing. They added a May 2024 date to the top of the official paper, but they show most people moved to Meta and Snap. 2 companies who have been aggressively forcing people back to office and are also generally higher paying than their previous employer

They should have at least added a note that their data was 1+ years old and that RTO mandates changed during that time

It’s not surprising that people left MSFT for companies that pay double or triple. The part that’s more interesting is if people are willing to take a 50% pay cut to go remote. I haven’t seen any research on that yet

So far the data shows RTO makes people leave, but it makes them leave to other RTO companies, not fully remote ones

IF every package in the monorepo is forced to use the same exact version then there would exist literally thousands of these teams, even tens of thousands, which I doubt

I don’t really believe if I go to 500 Google owned websites I will find the exact same version of angular everywhere

The US Python foundational team was less than 10 people. If you think any 10 person team in a 200,000 employee corporation is that critical well I’d have to disagree

When their jobs were moved to Munich there was a lot of discussion about how important they were, but something else stood out to me. One of the things they accomplished was taking over a year to make sure the monorepo can be upgraded to the latest version of python

As an outsider this definitely smells like an org creating work just to justify its budget. You have a single team of a handful of engineers doing the python upgrade work for tens of thousands of engineers. Doesn’t feel right out of the bat

I’m sure soon enough with the role moving they’ll find out if the team is needed at all

So buy comments on Israeli MK Facebook accounts? I think the only ones that would move any needle here is Meta profits on DAU and the company being paid to run the bots. You think your congressman goes to FB comments to decide how they’re voting?

Out: blame AIPAC for congress funding israel

In: blame FB comments for congress funding Israel

Is it really that hard to imagine congress supporting the democratic country with one of the biggest pride parades in the world vs the country that hasn’t had elections in 18 years and is split between 2 leaders who disagree on pretty much everything?

I feel like 90% of AI discussions online these days can be shut down with “a probabilistic syllable generator is not intelligence”

On the other hand, how is Sotomayor ever going to write another majority opinion without a case like this

As long as CEOs will try to convince the public/investors probabilistic word generators are a type of intelligence, the need for the biggest and strongest GPUs will continue to show how much closer you are to AGI

If they ever give up that lie, we could see a contraction

Luckily the NSA, GCHQ, DGSE, ASIO, and FIS don't exist. And if they did, they would never wiretap or hack anyone ever. Darn you Mossad!

I dont know if i would call something like this[1] an article. Taking a bus ad and writing a few sentences about it. He seems great at copy pasting promotional emails he gets I guess

[1] https://www.theverge.com/2015/12/28/10674048/verizon-will-pa...

So your take is Israel is going to take the US military built port and put millions of Palestinians on a boat? This goes way beyond conspiracy theory

Like you mentioned there are hundreds of trucks going in per day but there are also issues with Egypt shutting down their side, Hamas bombing the Israeli gates, and israeli protesters blocking aid. The sea bypasses all 3 of those issues. They’ve already transferred in hundreds of tons of aid in just the few days it’s been open.

You’re ignoring 2 major things

* Israel is the one operating the jetty. If you look at photos the trucks bringing the aid from the sea to land have yellow Israeli civilian plates. These are civilian Israeli contractors being paid by the Israeli government to disperse the aid because the Americans refused to have boots on the ground

* it only takes 3 people (prosecutor + 2 judges) to completely crumble the western block. You could suspect war crimes for any post 9/11 war campaign and arrest every past and present leader of the Us, France, UK, Australia since 2001 because 3 people said so. That’s way too much power for a small group

Its not just the US. A western democracy has never been issued a warrant like this. It's not that hard to find crimes being committed by western leaders (is no one listening to this seasons Serial podcast? Guantanamo is still open).

If this goes through every western leader past and present can have a warrant out for them at any minute, it really doesn't take much to find 1 suspected violation and a need to arrest and stand trial to see if they are guilty.

Very honestly asking - how do you convince investors you’re $100B away from an independent thinking computer if you’re not hiring to show that?

I’m sure these people are very serious about their work - do they actually know how far we are - technologically, spend, and time wise from real non word generating AGI with independent thought processes?

It’s an amazing research subject. And even more amazing a corporation is willing to pay people to research it. But it doesn’t mean it’s close in any way, or that anthropic would reach that goal in a decade or 3

I would compare spending this money and hiring these people to what Google Moonshot tried to do long ago. Very cool, very interesting, but also there should be a caveat on how far away it is in reality

This reads more like trying to create investor hype than the real world. You have a word generator, a fairly nice one but it’s still a word generator. This safety hype is to try and hide that fact and make it seem like it’s able to generate clear thoughts