voiper1

OK I can't even (easily) find the numbers.

What is an average person's carbon footprint? what is a rich person's carbon footprint? 10x, 100x?

Best I see right now is https://www.visualcapitalist.com/co2-emissions-by-income/ which is from 2019 and doesn't say exactly who's covered - I assume it's a global mix?

Income Group Average income (USD) Share of CO₂ Emissions Top 1% $310K 15% Next 9% $90K 34% Middle 40% $16K 43% Bottom 50% $2K 8%

So the top 10% by income account for 49% of the carbon emissions, and the next 40% for 43%, and the lowest 50% for 8%.

So that would mean if I got this right: Top 1%: 94x / year / person Next 9%: 24x / year / person Middle 40%: 7x / year / person Bottom 50%: 1X/year/person

So the top 1% person uses about 14x more carbon than a middle 40% person. Over one order of magnitude. I actually expected it to be far worse, what with the yachts and private planes.

Any better sources/numbers for this?

I'm still new to LoRA/fine tunes, but: I can't just dump in 1gb of data, correct? I need to structure it in Question/Answer or the like?

So it would seem the cost really becomes converting/curating the data into a usable format first.

Wow, I read your comment, and the response, then read it again, and I still thought you were intending to be offensive that the author was so bad they couldn't possibly collaborate with people.

Somehow I missed the simpler idea of "Hah, in 2024 I have no face to face with colleagues in anymore!"

Sounds just like the kind of long form articles I keep seeing links to before ChatGPT that I decided to stop reading. It's giving back what it's been trained on :shrug:

I just made it an option in the "I don't have a password" form that instead of setting a password it just logs them in. So they don't even see / have a password to remember.

I use a password manager. I don't like flipping back to my email, I already have a secure password.

I've only seen a handful of prompts that were better than just a few words asking for the result. Although, those prompts were more chains of very creatively asking what result you want.

In other words, I highly doubt this is providing much value.

"What is my purpose? To pass the butter" is now even worse, to roleplay as a believable fake character.

The comment you're responding to is in the context of whether a government issued ID should be required _because that would cut down voter fraud_. Additionally, the context is that's it's being pushed by Republicans - that it would prevent Democrat voter fraud.

1 - Texas: Crystal Mason. Her conviction overturned. So, this is a bad case to quote. And, ID was irrelevant here. She wasn't pretending to be someone else.

2- North Carolina: Leslie McCrae Dowless. He's Republican committing voter fraud. Also, Dowless was charged with multiple counts related to illegal ballot handling - seemingly not something Voter ID would have helped with.

3- Pennsylvania: 2020 - via https://www.ydr.com/in-depth/news/politics/elections/2022/08...

a) Bruce Bartman... serving five years on probation after he illegally voted for Trump

b) Melissa Ann Fisher of Bucks County recently was sentenced ... signed a mail ballot for her deceased mother: mail in issue AND public deaths should be remove you from the voting rolls. Not directly a voter ID issue.

c) Ralph Holloway Thurman of Chester County pleaded guilty... is a Republican

d) Robert Richard Lynn of Luzerne County pleaded guilty last year to completing an absentee ballot application and signing his deceased mother's name... Also absentee for deceased.

I'm not going to continue looking them up. Maryland is about voting in two states, so voter ID won't help that either. More about absentee ballots and more about absentee on behalf of deceased individuals. Voter ID on in-person voting is irrelevant to those cases.

It's republicans comitting fraud or cases where Voter ID won't directly help. That's part of that comment's point.

That was fantastic. The voice host really made the vibe!

but worse! there's no tools or RAG/data yet...

>Agents help you create custom behaviour and workflows with a simple set of instructions and examples. So, it's just custom instructions baked in? I hope at least it's harder for them to get overwritten by the user?

>We’re working on connecting Agents to tools and data sources... So tools and RAG for data sources aren't available yet.

Way behind GPTs/assistants. What's the point of this yet?

>We will give a 3-week notice before updating gpt-4o to point to the new snapshot gpt-4o-2024-08-06.

Source: https://platform.openai.com/docs/models/gpt-4o

Yep, commit your code when it "works". Then I can safely go off on a hair brained experiment, knowing I can easily throw away the changes to get back to what worked.

>I'm also a security researcher. My day-to-day job for nearly the last decade now has been to show all of the ways in which AI models fail spectacularly when confronted with any kind of environment they were not trained to handle.

> ... And yet, here I am, saying that I think current large language models have provided the single largest improvement to my productivity since the internet was created.

>In the same way I wouldn't write off humans as being utterly useless because we can't divide 64 bit integers in our head---a task completely trivial for a computer---I don't think it makes sense to write off LLMs because you can construct a task they can't solve. Obviously that's easy---the question is can you find tasks where they provide value?

If you use it as an intern, as a creative partner, as a rubber-duck-plus, in an iterative fashion, give it all the context you have and your constraints and what you want... it's fantastic. Often I'll take pieces from it, if it's simple enough I can just use it's output.

I know this is a bit heavier since it's a whole program:

https://aider.chat/ generates a repo map, suggests related files, then makes your request and does the editing for you. (with intelligent diff for models that work well.)

They have a leaderboard for models for editing and what mode they use: https://aider.chat/docs/leaderboards/

EDIT: Oh, it looks like you're dumping the entire repo. Well, sure, I guess. Once it seems too big you can select just what you need. Also gui tools like the open source https://www.continue.dev/ lets you pick files, search codebase via RAG, add docs, etc...

groq just released:

llama3-groq-70b-8192-tool-use-preview llama3-groq-8b-8192-tool-use-preview

Sonnet3.5 also has tool usage, and is slightly cheaper than 4o ($3 input vs $5).

If you _only_ need tool usage, then 4o-mini or others will be cheaper...

mistal's Mixtral 8x22B has function calling for $2/$6, which is a bit cheaper than 4o.

Oddly, they are only charging slightly more for their hosted version:

open-mistral-7b is 25c/m tokens open-mistral-nemo-2407 is 30c/m tokens

https://mistral.ai/technology/#pricing

I think unify.ai (like openrouter) does that - it has several paramters you can choose from.

But the underlying "how to choose a model that's smart enough but not too smart" seems difficult to understand.

Would VS Code support a plugin to _display_ python with brackets, but open/save to disk in standard indented format?

Wow, I find this very useful. The eliciting an expert, alternative experts, oppositional viewpoints... fascinating.

We're just getting started on what we can do with LLMs!

So the long game is rival countries are secretly collaborating, so we secretly have world peace!

calling it sudo would create the expectations that all the options, config, and usage is exactly the same.

It would appear that it's _functionally_ the same, but using a different mechanism and with a new name so they don't want to be locked down to all the other stuff.

In amatuer tennis, he would say that indeed you need to learn how to play. But you just play and return the ball. You don't need anything fancy to make you win. You just keep playing until your opponent makes a mistake.

>This observation applies only to amateur tennis. In professional tennis it’s just the opposite: 80% of the rallies are won rather than lost, as unforced errors are infrequent. This is true in chess as well, as high-level players don’t blunder, and thus it really is that litany of other skills that results in high standings.

I think this means at a high level, you're already not blundering. So you need to actually do something MORE in order to win. You need to force your oponent to lose, and not just play "OK" and wait for them to make a mistake.

That's a compelling story, I wish you the best of luck!

EDIT: Explore AI as part of this (e.g. using RAG), many have found it super useful as a tutor for personalized explanations when you don't understand the textbook. If you have it grounded in the textbook or other sources, it should have fairly minimal hallucinations.

Reminds me of: "Louis CK Everything Is Amazing And Nobody Is Happy"

+1 for "Practical Engineering". He has an amazing series on making a new pump station called Practical Construction:

https://www.youtube.com/watch?v=PdcXkmvXXwU&list=PLTZM4MrZKf...

>At least when you share a Google doc or something there's an option that explicitly says "anyone with the URL can access this".

Unfortunately, it's based on the document ID, so you can't re-enable access with a new URL.