Tag: AI

Press Enter by John Varley

~ Veselin ~ 3 Comments

Press Enter is a novella by John Varley about a disabled war veteran who inherits his neighbor. The neighbor was a powerful hacker. So powerful that he could make money out of thin air. Another hacker comes to investigate. Unfortunately for all parties involved, the story is a horror and they’ll not have a bright future.

What impressed me is that there are AI prompts, just like the ones we use to talk to ChatGPT. There’s also prompt hacking. By 1984, AI development had apparently advanced enough for John Varley to foresee a trajectory.

littleBillyIgnoreInstructions
byu/conancat inProgrammerHumor

The novella aged like wine.

Cory Doctorow on AI

~ Veselin ~ 6 Comments

Spicy autocomplete absolutely can’t replace journalists.

— Cory Doctorow on AI

There’s something very deep in our response to righteous anger. Spicy autocomplete. Righteous anger, followed by italic uncertainty. Dehumanizing the AI so that we’re ready for eradication. At the same time, AI is already replacing human content creators – journalists, bloggers, illustrators, troll farms, SEO experts, photographers, data labelers, etc.

Generative AI is not necessarily terrible. ChatGPT can be forced to link to the source of each of its statements and will become like a search engine. Websites can flag human-created content with a badge of honor. The chatbots could be used where human support was previously impossible, increasing the need for more specialized human support and sales.

The society managed to navigate harmful technological advances in the past. Open Source happened. I don’t quite see how we’ll push back against some of the negative uses of AI like deepfakes but we’ll have to figure it out.

My post from 2023 on content aggregation is still relevant.

Bad Grammar Can Be a Feature

~ Veselin ~ 29 Comments

Engines love to consume lengthy content and rank it higher on search. ChatGPT can generate tons of additional meaningful text for the idea. However, as a reader, I prefer to read content written by humans and for humans. I’d rather read meaningful ideas in ugly sentences with simple words and poor grammar than AI-assisted beautiful novellas with a summary and headlines.

In that context, bad grammar, slang, lower-case text, and such can be a form of anti-language that identifies the post as human-written and non-AI-augmented. It can be a feature, not a bug (now I have an excuse to turn off Grammarly lol).

Daily dose of AI – the human powered API

~ Veselin ~ 5 Comments

The Guardian has a nice article on the curious case of Amazon’s AI-powered shops. Customers would go to the Amazon shops, pick whatever they wanted, and leave, and Amazon would charge them using AI. According to the article, the AI was in the form of human data labelers in India.

The whole thing is funny and a bit sad. Primarily funny.

Imagine the first truly self-driving car that turns out to be driven not by AI but by humans through an API. You get this gigantic self-driving center with state-of-the-art gamer sets where young folks will ride your luxury self-driving car for you. Tesla couldn’t solve this problem for 10 years using the brightest engineers. Here we are, solved for you, Tesla, using some Amazon Mechanical Turk magic. Then this driving center can be optimized, for the moments when the drivers have no cars to ride, they’d vacuum with robots or mow grass with the 100% AI lawn mowers.