Empire of AI: Dreams and Nightmares in Sam Altman’s #OpenAI
The Tech won't save us podcast has an interview with Karen Hao, author of Empires of #AI
It is well worth a listen as gets behind some of the dubious assumptions of the big AI companies and the personality of Sam Altman one of the most influential people in this area.
Okay, time for an outlandish claim:
in a few years we will treat most code like we treat HTML today. It's generated by machines, it's okay, it may be ugly sometimes, but as long as it works we don't care.
Remember: first there were the TeX vs Word battles; beautifully hand-crafted layout vs a document that is ugly to read (if it is non-binary at all) but does the job.
Then there was hand-written HTML vs web site generators.
Both these battles were won by the "ugly but fast and simple" fraction, except for special cases.
Yesterday I had an LLM (Gemini 2.5 pro) generate a larger piece of code for me - roughly 700 lines of python. And I couldn't bother trying to improve it. I glanced over it to check for anything fishy, it seemed to do the job, I tested it, it did the job. It's not as beautiful as hand-crafted code, but, well, it does the job. Yes, a bit like vibe coding.
And we're only at the beginning.
Oh yes, and related to that: we will see a flood of machine-generated apps in the app stores. And they will be fine. Maybe writing apps like todo lists, casual games will be not a thing in a few years.
#arafed #AI #LLLM #halluzination
arafed futures - An Artist Dialogue on Chip Storage and AI Accelerationism
https://fahrplan.events.ccc.de/congress/2024/fahrplan/talk/Z7TFKB/
McDonald's hat die Zusammenarbeit mit IBM zur automatisierten Bestellannahme im Drive-Thru nach zwei Jahren Testphase in über 100 Restaurants beendet. Das Projekt wird nicht erweitert, dennoch bleibt McDonald's zuversichtlich, dass KI-gestützte Drive-Thrus zukünftig relevant sein werden. #ai #ki #mcdonalds #lllm
@LukaszOlejnik LLLMs "Less Large Language Models"?
