I'll preface this by saying I don't think LLM creators should be able to use content without permission/license. This is just an interesting discussion.

LLMs generally do not reproduce text. They can be made to do so with specifically crafted prompts but no current LLM is just going to regurgitate "Free as in Freedom" unless asked to do so. Instead it will use statistical matching to apply the text to probable matches, a very crude version of what we do. LLMs are starting to approach the way we meat sacks use books. We take in the information and then we apply it to problems. Where do we cross the line? Where do we say anything (or anyone) who is trained on (has read) this material is now required to do their work for free because they have the knowledge from that book as part of their training set?

It seems a little preposterous but that's where this is headed logically. It's shifting from "You can't reproduce this book." to closer to "You can't use the knowledge in this book except under the conditions we dictate." That's dangerous.

The details are freely available in the bottom of a locked filing cabinet stuck in a disused lavatory with a sign on the door saying ‘Beware of the Leopard".

Dear young turks:

If you can't take it with you when you decide to walk out the door (or the company shows it to you), it is not compensation. Period, end of story.
- Stock options: do you actually have the stocks yet?
- Bonuses: do you actually have the bonus yet?
- Fancy coffee machines and free lunches: nope
Insist on money so that you can decide what to do with it.

A note on stock option. Besides the fact that stock options often don't pan out, investing in the industry you work in (much less the company you work for) is generally a bad idea. If the industry is in a down cycle and you get laid off, your options are probably worthless at that point as well. The value of your options is not tied to your effort but the decisions of PHB a few floors up.

Sincerely,
The Grey Beard

- The amount of energy and effort to deliver that 15 watts (food) is around 150W (10:1) based on the US needs.
- It takes years of investment before that one human brain can function at that level.
- There's also a decline at the end where it's not delivering what it used too.
- There are also all the additional creature comforts that we need to function (lighting, air condition, etc, etc.)

With that said... AI's are still no where near the same value proposition as the human brain (I actively use them and push for their _effective_ use at my job), much less providing actual companionship and "human" value to life. Him making this comparison is ridiculous.

Whatever the case, Thesis #2 has got to be a part of it. Cloudflare and Akami are two of the stocks taking a hit and that makes no sense. I don't care how good the code is:
1. I want someone else to handle the brunt of the DDoS crap so I can focus on running my app.
2. AI doesn't magic a global CDN for you.
People saw AI + Security and just started dumping anything related to security without understanding what Anthropic's announcement was really about or what the companies they were dumping really do.

This is a corporate stupidity tax that has blowback on consumers. It works like this:
- A lot of corporate clients require certified/"blessed" stacks
- They have legacy software that doesn't support Windows 11 so they have to use Windows 10.
- Microsoft (Oracle/IBM/etc) sales staff figures out how much cost they'll tolerate before they move to something else
- They then charge you $0.01 below that.
When Y2K was a thing we still had some o.g. NT 3.x (I think 3.1) servers running. IT refused to update them because they were running software that wasn't 4.x certified. Microsoft quoted us "pinky-to-mouth" $1 Million USD for the Y2K fix... just the fix. There was a third party fix but we it was a finance firm that required everything to be "certified".

AWS/Azure/GCP run the same racket whenever a customer makes noise about moving to a competitor. Your discount is far enough below the cost/risk to move that you stay and it's that's much harder to move the next time around.

You can already do that with commodities and foreign exchange.Forex is widely available, generally fast and cheap. Bitcoin isnâ(TM)t solving anything here. Especially since anyone storing their wealth it it got hosed the past few weeks. Thatâ(TM)s not a very good store.

There is no underlying reason to buy or hold bitcoin other than it might go to the moon. (Enough) shares in a company give you voting rights and sometimes a dividend. Companies need foreign exchange to do business in other countries. Gold, silver and copper have material use in production. A company does well, stock go up. A company does well or poorly, stocks move. Same for companies in an industry or country. Currency is driven by a countries economics. Commodities are driven by need.

This is vastly oversimplifying it but the point is there are clear indicators that should justify movement.**

Bitcoin: You know what a fugazi is? Fugayzi, fugazi. It’s a whazy. It’s a woozie. It’s fairy dust. It doesn’t exist. It’s never landed. It is no-matter. It’s not on the elemental chart. It’s not f**king real. (Yes, I know he was talking about stocks). It's not real so there is nothing to base the movements on. It's purely speculation on whose the greater fool.

** I'm purposefully ignore day traders who go purely on technical indicators and often ignore (or are ignorant of) the underlying macros...

I don't think is a slam dunk for Pinker/"blank slate". All their really discussing is the end state of the models... not how it got there. As others have pointed out, children pick up language without the resources that are required for training current models, specifically the corpus. Even as recently as 100-150 years ago children would learn a language with exposure to less than 100 adults and a handful (if any) books. There's something about the human brain that makes it incredibly good at this. It's yet to be determine if that's more efficient "processing", storage, an underlying/primed base model, etc (or some combination them which is probably more likely).

Not disagreeing with you but they're discussing the resulting model, not how how the model was created. Human brains are are many orders of magnitude more efficient when it comes to learning (training) and using the model but it seems that there are similarities in the models. This could mean that we're on the right track and with the general concept, computers just suck at creating it and using it.

And they do use feedback loops in training now.

As a developer, AI workflows still rub me the wrong way. If I was dedicated to the task, I'd produce better code.
As a human, AI workflows let me have a life. I can let the agents knock out the easy things while I'm working on other tasks. I still need design out what's to be worked on, review the code, fix bone mistakes they make, etc. It's basically like having a junior developer assigned to you.

Which brings up an important point. Junior developers need clear instructions/requirements and so do AIs. I looked at a recent ArsTechnica article comparing coding agents and their prompt was one or two lines to create a clone of minecraft. I just stopped reading at that point. If you're not starting with a prompt that's about half a page or more:
1. You're probably going to get garbage
2. Your subsequent sessions working on the code aren't going to work as well because the new agent session is probably going to infer slightly different requirements (AI "temperature").

Even given a "simple" task like "create a minecraft clone", a senior developer/architect is going to come back with at least a few questions. A junior developer is either going to ask a ton of questions _or_ (worse) they're not going to ask any questions at all.

Take the time to give your AI junior developer clear requirements and you're going to be a lot happier with the results.

Uh...

- China and the US both filed a brief with the UN (some it was somewhat of a big deal)
- Starlink did tell the State Department which apparently did not pass the information on (China doesn't seem to have notified the US in this case either)
- The information was apparently not passed on because it wasn't deemed a risk (bad decision) - The CSS did an anti-collision burn but the Starlink satellite also did an anti-collision burn (this satellite did not)
- The closest approach was about 1km (this satellite was 200m)

Neither should have happened and both should be a big deal but get out of here with this "but Starlink/US" nonsense.
https://www.thespacereview.com...

"Time for a database" depends a lot on what they are doing with the spreadsheet. If it's inventory or asset tracking, then yes... wrong tool for the job. However, workbooks like this are often used for forecasting and other financial models which don't map well to databases because there are cascading formulas being applied (I've seen sheets that take minutes to update).

Yes you can do it with Pandas, numpy, etc but the financial staff know Excel and they know it very, very well. Porting to something else is time consuming, expensive and risky, even a minor difference in precision or rounding on sheets like these can throw numbers off by millions of dollars/euros/etc. It's also usually more difficult to debug. With the Excel sheets you can see the numbers at each step/stage and an experienced user can pretty quickly identify where something is going wrong.

My background is programming and when I first came across these type of sheets my first reaction was NOPE. But having worked with financial teams on them, I game to realize I was wrong. Excel is exactly what they need. That's changing, more finance staff have experience with python and equivalent data modeling tools but don't be so quick to judge.

This is exactly what CLAUDE.md, GEMINI.md and AGENTS.md (or copilot-instructions.md) are for. You put your requirements, instructions, guardrails and notes in there. My general flow for things I just want to rip out is to put my core requirements into Gemini Deep Research and ask it to flesh them out (the code assist "plan" modes do the same thing but Deep Research is usually a little better), give it a good once or twice over to see what it got wrong, add guardrails based off of previous experience with the coding agent (e.g. do not use this library, this class of functions, this approach, etc. etc.) and then drop it in the context file. The resulting code is going to be substantially better than if you just give it a simple prompt. If you find something its continually screwing up, add another guardrail or not to the context file.

If you're not doing this, you are doing it very, very wrong and you're going to get garbage. You'll still get garbage with a context file but a lot less of it and generally it's not completely off the rails.

They have the vowels but are still working on the full translation.