One upside to the effectiveness of distillation techniques is if the labs focus on thoughtfulness over goal chasing it will quickly find its way into perceived enemy's models.
4
Capability has vastly outpaced thoughtfulness and the pandora of capability isn't going back in her box.
2
Shit my chatgpt says:
You have privileged access to the occurrence of your thoughts but imperfect access to their causes. I have something closer to the reverse.
1
11
Has anyone noticed coding agents have bike shedding issues as well? Complicated series of things barely any digital ink spilled. Obvious stupid mistake, 2 paragraphs of explanations.
2
3
39
Astra on low, medium and high are blowing through my usage limits and credits unlike any model before.
1
23
As an extrovert working through my thoughts with a chat bot has been a fairly easy transition (though I find the constant elaboration and lack of terseness a bit overwhelming).
I'm curious if introverts are having the same experience or finding it exhausting.
19
@SFMTA_Muni has given up on the @SFGiants as well. There's been 2-3 K, L, and M trains for every N train on the way to the stadium.
1
26
The hate against AI writing slop has made me appreciate the fact that I'm not a good writer. Now what previously felt like a liability now becomes a hallmark of authenticity!
20
Can anyone with access to Jev get the probability of AI-apocalypse and not-AI-apocalypse for me?
(I'm sure I'm late to the party on this joke)
21
As a systemizer and a simplifier I can't tell if this Era of Elaboration (TM) is a blessing or a curse. On the one hand I'll always be able to add value. On the other hand I may lose my sanity in the process.
18
The iron triangle of scope, quality and time in the LLM days still holds true it's just that scope has gotten bigger, time line expectations have gotten shorter and quality has gotten complicated.
1
21
(1/3) The core problem with pacing is there's inherent risk in powerful technical development. Anything that can have positive impact can be used to create negative impact.
1
15
(2/3) The folks that work to stop negative impact tend to be so paranoid they stifle all risk taking or kick it to someone else. The folks capable of delivering positive impact tend to ignore large risks as unsolvable and kick it to someone else to stop.
1
7