Reminder that EVERYTHING Anthropic or OpenAI announce in public is propaganda designed to boost their market cap when they hit the IPO they're aiming for.
It's marketing, folks. There is no intelligence behind the artifice, it's just spicy autocomplete and hucksters in $5000 suits trying to pick your boss's pocket.
mastodon.social/@arstechnica/1…
Ars Technica (@arstechnica@mastodon.social)
Attached: 1 image Researchers question Anthropic claim that AI-assisted attack was 90% autonomous The results of AI-assisted hacking aren't as impressive as many might have us believe. https://arstechnica.Ars Technica (Mastodon)


André Polykanine
in reply to Charlie Stross • • •Charlie Stross
in reply to André Polykanine • • •Pier Hegeman
in reply to André Polykanine • • •G. Clavier
in reply to André Polykanine • • •@menelion "An autocomplete would never suggest new names, new implementation and architectural ideas, and so on, and so forth."
Why? Autocompletes often make changes I do not want and that are out of context with bad parameters.
LLMs are basically large scale autocomplete, running on GPU, with a grain of stochastic variablity put on top.
André Polykanine
in reply to G. Clavier • • •Of course, it hallucinates.
Of course, you don't commit AI-generated code to production, you don't send AI-generated emails without checking, you don't… you get it, of course that is absolutely true.
But I don't get why not to use it as a tool that can make processes far quicker and spare your time from boring stuff.
Dave
in reply to André Polykanine • • •@menelion @Enthalpiste
In the past few months AI has wasted far more of my time than it could ever possibly save.
One of the things I do in my day job is proof-read, re-word and fact-check matters of local history. I spent days trying to unpick a batch of information that simply didn't add up and couldn't be referenced. In the end it became apparent that someone had searched the internet for some background and believed the first things they read on those wretched AI overview search engine results.