Yep. The best bumper sticker-level idea I see is “First, assume China has stolen everything to date because you’re writing your findings on dry erase boards in open offices all across SV. Second, STOP DOING THAT.”
Nice addendum to your recent Scriptnotes episode. My AI anxiety wheel alternates between misinformation, state surveillance, job losses, and environmental disasters. Happy new year?
My coping mechanism is to find solace in the present, have fun, and focus on what I can control. That being said, what movies TV show would you say have the best take on AI and the future of tech?
Upload is probably the funniest and probably (and sadly) one of the most realistic one imo.
A (characteristically) thoughtful, thorough, and clearly explained take on AI. I know you didn’t make that graph, but I’m struggling some with the y-axis (not a measure of actual progress?). A useful comp here might be predictions about full self-driving cars - progress could plateau … or not.
Self-driving cars comparison feels apt -- that last 10% is so difficult. The chart with the black background is about the arc-agi prize, which focuses on tasks that are pretty easy for humans (like Amazon's Mechanical Turks) and classically difficult for AI. https://arcprize.org/guide
Sorry for being unclear - I meant the chart with the gray background, from Ark Invest (the thumbnail image for the post), though I suppose both support the basic idea of fast progress.
Regardless, thanks for taking the time to share all this with us, and for being an informed advocate for writers.
Oh, the Y axis is how many years respondents think it will be until specific metrics for AGI are met. X axis is time (in six-month increments). Agreed it's not an awesome presentation, but it's the best I could find.
Great post @johnaugust.bsky.social. It does feel like something big is coming, though it might turn out to be generated by flesh-and-blood right-wing politicians and billionaires.
Love this take, absolutely agree. I also wish there was a forum for creatives, including screenwriters, to have open and transparent safe-space discussions about AI, without being shunned by the community.
The prototyping/non-programmer thing is interesting, but then again a major push throughout programing language development has been toward that sort of abstraction: programming complex systems in simple English isn't even new.
It took me back a bit to my HyperCard days, but even that isn’t a great comparison. HyperTalk was still specific and esoteric. This was just English and full of typos. Less programming and more just describing — which is something writers are good at.
True; I also think of my kids using Scratch at, like, 6 and dropping blocks to make things happen pretty quickly. But any fine-tuning would require learning, God help them, to think like a programmer.
What LLMs get you is the benefit of fuzzier parsing — and of course the huge amount of hoovered data to draw from. (Although because of that they're a lot less useful the more fine-tuned you get, as opposed to a rules-based system.)
Comments
https://situational-awareness.ai/lock-down-the-labs/
But also war. It’s not a great game tbh
Upload is probably the funniest and probably (and sadly) one of the most realistic one imo.
Regardless, thanks for taking the time to share all this with us, and for being an informed advocate for writers.
https://www.cnbc.com/2024/12/27/openai-needs-more-capital-than-wed-imagined-moves-to-for-profit.html
https://www.bloomberg.com/graphics/2024-ai-power-home-appliances/
https://www.latitudemedia.com/news/metas-all-of-the-above-strategy-shows-why-powering-ai-is-so-complex/