My understanding is that this isn't a consequence of a flawed training algorithm or process; it's instead a consequence of the limitations of LLMs, emergent from their training materials. It closely parallels another example I've seen around the net, that of asking an LLM about getting a car to the mechanic, noting it's a sunny day and the mechanic is just a block away, and having the LLM suggest walking... which is a consequence of the bias in training materials toward walking because lots of people make visible posts about their having done so (because it's looked on favorably), whereas people who drive short distances (of which there are many, probably outnumbering walkers) don't trumpet having done so online, leading LLMs to emit advice about walking when possible (and in the case of the mechanic example, having a lack of comprehension of the pivotal aspect of having the car make it with you to the mechanic's shop).

I read about this phenomenon in the young adult novel "The Secret of Terror Castle", published in 1964. Three teenage detectives investigated a mysterious castle where nobody could spend the night without running away in terror. The culprit? Someone playing the lowest notes on a pipe organ to produce the jitters. This book was the first in a much beloved series called "Alfred Hitchcock and the Three Investigators".

Vaughan-Nichols is right about the problem and wrong about the solution. Voluntary pledge funds and tip jars have existed for a decade and the 60% unpaid figure hasn't moved. Sentry is admirable precisely because it's an exception, the model doesn't scale because it depends on individual corporate virtue, which is in shorter supply than VC funding.

Perens' Post Open licensing approach is interesting but creates a two-tier ecosystem: "free for individuals, pay for commercial use" sounds clean until you realize it breaks the fundamental property that made open source eat the world. The moment a license is commercial-use-restricted, it's not open source, it's source-available, and enterprises will treat it accordingly (avoid it, fork it before the license change, or just use the last MIT-licensed version forever).

What's actually needed: mandatory contribution structured as a fee, not a license restriction. Here's one way to do it. Small flat fee on all US commercial revenue above $5M (the entire world runs on OSS, everyone pays to maintain it), larger marginal fee on companies whose products directly incorporate OSS. Fees flow into a scoring-based royalty pool: your project's share is proportional to how much commercial revenue depends on it, revenue-weighted so a hedge fund running its entire risk engine on a niche numerical library counts for more than fifty startups using the same package for weekend side projects. Maintainers register and claim their allocation like music royalties, no government agency decides who gets hired, just checks cut proportional to actual commercial stakes.

The core insight: you can't solve a collective action problem with voluntary action. You solve it by making the externality visible in the price.

AI learned to code by reading human code examples. Where will the training examples come from if AI codes directly to a human-unreadable language?

So, companies that violate this act are... debased?

I suggest "Layoff Phone".

To be clear, this was true over 20 years ago. (In light of which, the word "now" probably doesn't belong in the above phrase, since it implies recentness.)

Thanks for your questions, Freenet caches data but it isn’t meant to be a long-term storage network. It’s better to think of it as a communication system. Data persists as long as at least one node remains subscribed to it. If nobody subscribes (including the author), it will eventually disappear from the network. So yes, if only your node subscribes then the data will only exist there and won’t be available when your machine is offline. But if other nodes subscribe it will be replicated automatically and remain available even if your node goes offline.

Not from 2023, the linked video is from last month. https://www.youtube.com/watch?...

Reminds me of an episode of Futurama where Fry runs into the room and breathlessly exclaims, "I got here as quickly as I could once I found out what happened a thousand years ago!"

... was pretending that Windows has value.

They were first to market, but I don't think of them as having invented the product.. The emergence of chatbots seems inevitable once the paper in 2017 was authored by several google engineers (titled "Attention is all you need")... it was just a question of exactly who and when. If OpenAI hadn't gone first, someone would have shortly after.

And, in a lot of ways even that google paper's "breakthrough" wasn't so much the tech (neural nets) but the precise adaptation of it that made it highly parallelizable.

And a necessary ingredient was tons of data, and processing power. So this couldn't have happened in a garage operation like the innovators of yore. And the biz models they're all coming up with are all cloud based -- not that I don't see the profit motivation, but so utterly to the exclusion of any offering that could guarantee privacy; all we "know" about chatbot conversation privacy is what each vendor claims at the moment,, which isn't much, wouldn't be verifiable if it was, and could change on a whim tomorrow.

For these reasons, I don't attach much "early innovator" romanticism to the players here.

:) I'm sure he puts on his pedants one leg at a time just like we do.

Answer: diversify.
40% domestic stock, 20% international stock, 40% fixed income is an example mix for non-cash investments for an adult.
Avoid precious metals, crypto, etc. Too unpredictable.
You might not get rich, but you may have a decent life.