Friday, May 19
Daily News Stuff 19 May 2023
Spirits From The Vasty Deep Edition - Now With Added Formatting!
Disclaimer: Here's an open-source client for my $5000 per month service. Don't say I never did anything for you.
Spirits From The Vasty Deep Edition - Now With Added Formatting!
Top Story
- If you want to learn Python you could probably do worse than the current No Starch Press offer at Humble Bundle.
18 books at about $2 each. Automate the Boring Stuff with Python in particular has glowing reviews, and buying that book alone would cost as much as this entire bundle.
I have a ridiculous number of books on Kindle now, largely because I buy anything up to 100 each month in Humble Bundles.
This particular bundle is available for another three days, but if you miss it there' will probably be a new Python bundle within a month.
- I'll make my ChatGPT, with blackjack, and hookers. (Eric's Code)
He's good to his word, not only making ChatGPT with blackjack and hookers, but showing you how to do it yourself. You'll need a reasonable level of hardware - he recommends 2TB of fast SSD to make sure you don't run out of room in the middle of a 20-hour training run - but nothing outlandish in a time when a brand new 2TB SSD costs less than the average monthly cable bill. The instructions suggest renting time at AWS rather than trying to configure the system yourself - not that you can't, just that it's easier.
The problem is that while there are now multiple open source AIs in the style of ChatGPT, the bootstrapping process to get them trained has infected them with the same authoritarian woke bullshit as ChatGPT itself. As the author says:It's my computer, it should do what I want. My toaster toasts when I want. My car drives where I want. My lighter burns what I want. My knife cuts what I want. Why should the open-source AI running on my computer, get to decide for itself when it wants to answer my question? This is about ownership and control. If I ask my model a question, I want an answer, I do not want it arguing with me.
And then details exactly how to achieve this. Not in broad terms, but with specific instructions every step of the way.
Tech News
- The leaks were right once again, and Nvidia has launched the 4060 and 4060 Ti. (Tom's Hardare)
The 4060 Ti 8GB model will be in stores next week at $399.
The 4060 8GB model will ship in July at a pretty reasonable $299.
And the 4060 Ti 16GB model will also ship in July, at $499.
Which puts it half-way to the much faster 4070 and means once again that Nvidia really doesn't want anyone to buy its products. 8GB of GDDR6 RAM costs around $33 on the spot market, and Nvidia and its board partners will be paying rather less than that.
- Is your laptop just too fast and sleek for your liking? The Book 8088 DOS System has an 8088 running DOS. (Liliputing)
An actual genuine 8088, with an 8087 coprocessor socket. And a socket for an OPL-3 sound chip as found in the Soundblaster Pro, because as standard it can only make tinny little bleeps.
- Bluesky Social, the company started by ex-Twitter CEO and drugged-out mosquito bait Jack Dorsey, just released its code as open source. (ZDNet)
Unlike Twitter, which is still tripping over its own open source feet, Bluesky client code is for anyone who wants to work on improving the code or use it as the basis for their own social network. Twitter's recommendation code, on the other hand, is essentially unusable.
What they have released is a social network client. Completely unrelated to the server-side code that Twitter released, and really only of use to people who want to write social network clients for mobile devices. Or rather, people who want to have written social network clients for mobile devices without doing the work, and who are willing to have a client that has no server to talk to other than Bluesky itself, which is still in very limited release.
The Bluesky code, licensed under the MIT License, can be used now. Indeed, while it's been out for only about 24 hours, it's already been forked 88 times and has earned over 1,300 GitHub Stars.
While it's specifically the Bluesky Social app's codebase, it's also a resource for AT Protocol programmers. This protocol supports a decentralized social network. Its features include connecting with anyone on a server that supports AT Protocol; controlling how users see the world via an open algorithm market; and enabling users to change hosts without losing their content, followers, or identity.
The code itself is written in React Native. This is an open-source, user-interface JavaScript software framework. It's used primarily to build applications that run on both iOS and Android devices.
Disclaimer: Here's an open-source client for my $5000 per month service. Don't say I never did anything for you.
Posted by: Pixy Misa at
06:16 PM
| Comments (4)
| Add Comment
| Trackbacks (Suck)
Post contains 813 words, total size 6 kb.
1
Site formatting's blown up. Either the right column's too far to the left, or the left column's too wide, or both. Also, the font changes after the top post.
Posted by: Rick C at Friday, May 19 2023 11:29 PM (BMUHC)
2
"as standard it can only make tinny little bleeps."
Back in the 8086 days, the demoscene guys did amazing things with the basic PC speaker--actual music. But mere mortals writing video games didn't generally get anywhere near that level of result.
Back in the 8086 days, the demoscene guys did amazing things with the basic PC speaker--actual music. But mere mortals writing video games didn't generally get anywhere near that level of result.
Posted by: Rick C at Friday, May 19 2023 11:57 PM (BMUHC)
3
Re: I'll make my ChatGPT, with blackjack, and hookers. (Eric's Code)
I'm thinking a bit lately about Dune's Butlerian Jihad. (Ignoring Frank Herbert's son's sequels in the same way you might ignore the Star Wars prequels):
In the original novel, one of the fundamental philosophies of the depicted culture is rejection of the outsourcing of human thinking. (Of course, said culture also outsources their thinking to various bits of dogma and psychological manipulation, but that's a separate problem.) The rebellion against "thinking machines" had less to do with your standard AI overlord plot, and more to do with externalizing human abilities, when we should be seeking to internalize these abilities.
I think the AI pushers in silicon valley are trying their damndest to push us in a direction of losing our innate skills, becoming dependent on tools they own and control, and then becoming biddable to their "authority". (So much of the worship of these LLMs is just laundering argument from authority. If you don't know something for yourself (including how-to's like how-to write python, how-to compose a letter), then it puts you at a disadvantage for something claiming that authority (though the claim itself should still merit every skepticism.)
Rather than training our machines, then losing the skills, we should instead be using our machines to train our own brains, and internalize the skills.
Naturally, the alignment/singularity crowd, being fundamnetally hostile to freedom of thought (they regard uncontrolled thinking as an *existential threat*, whatever the true reasons for their horror), won't willingly go along with this.
I'm thinking a bit lately about Dune's Butlerian Jihad. (Ignoring Frank Herbert's son's sequels in the same way you might ignore the Star Wars prequels):
In the original novel, one of the fundamental philosophies of the depicted culture is rejection of the outsourcing of human thinking. (Of course, said culture also outsources their thinking to various bits of dogma and psychological manipulation, but that's a separate problem.) The rebellion against "thinking machines" had less to do with your standard AI overlord plot, and more to do with externalizing human abilities, when we should be seeking to internalize these abilities.
I think the AI pushers in silicon valley are trying their damndest to push us in a direction of losing our innate skills, becoming dependent on tools they own and control, and then becoming biddable to their "authority". (So much of the worship of these LLMs is just laundering argument from authority. If you don't know something for yourself (including how-to's like how-to write python, how-to compose a letter), then it puts you at a disadvantage for something claiming that authority (though the claim itself should still merit every skepticism.)
Rather than training our machines, then losing the skills, we should instead be using our machines to train our own brains, and internalize the skills.
Naturally, the alignment/singularity crowd, being fundamnetally hostile to freedom of thought (they regard uncontrolled thinking as an *existential threat*, whatever the true reasons for their horror), won't willingly go along with this.
Posted by: madrocketsci at Saturday, May 20 2023 01:42 AM (hRoyQ)
4
In my view, the AI alignment/singularity folks want to have some spirit as a boss that they can appease. They do not want to look at programs as things which maybe have abstract mathematical properties that can be understood in terms of limits of what the calculation scheme can actually do.
Hearing about Yudkowsky these days made me extremely thankful that I had noped hard at the opening of Harry Potter and the Methods of Rationality.
Hearing about Yudkowsky these days made me extremely thankful that I had noped hard at the opening of Harry Potter and the Methods of Rationality.
Posted by: Pat Buckman at Saturday, May 20 2023 10:41 AM (r9O5h)
56kb generated in CPU 0.0127, elapsed 0.1081 seconds.
58 queries taking 0.0998 seconds, 351 records returned.
Powered by Minx 1.1.6c-pink.
58 queries taking 0.0998 seconds, 351 records returned.
Powered by Minx 1.1.6c-pink.