this post was submitted on 24 Sep 2024
95 points (98.0% liked)
TechTakes
1494 readers
51 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
This causes me to reflect on contrasting currents in tech culture. I remember growing up with the Apple/Mac rumor culture around the time Steve Jobs came back, and how people had conditioned themselves to get hyped for any little tiny leak about upcoming products. A culture which obviously persists now, albeit in streamlined, advertiser-friendly blog spaces. By contrast, MacWeek magazine had a columnist calling himself Mac the Knife who claimed to have clandestine rendezvous with shady trenchcoat-clad characters in the back alleys of Cupertino... And somehow the new product reveals were almost always somewhat less whelming than the rumormill had built them up to be. Part of the Jobs idolatry that still dominates Silicon Valley is the clear strategy among empty-suit grifters like Altman that such hype is vital but Apple didn't do enough with it; that you should always be marketing what's around the corner rather than keeping it hidden away under lock, key, and NDA.
Contrast this with open-source culture, warts and all. What's in the repository is the basis of what comes next. You think superintelligence is imminent? OK, where's the code stubs that will serve as the foundation? Make a pull request for your mega-brain's medulla, let's review it. It's also a big reason why the current round of AI doesn't fit with open-source culture, no matter how many people are trying to force it. It is inherently an obfuscatory technology. Not just due to the sheer size of the data sets and weights involved, but also through the weird non-deterministic practice of configuring software through natural-language prompts. GIGO at scale, but you can keep the hype going by promising a lower percentage of garbage in the future.
The GIGO pipeline: Sure, it's garbage, but hear me out! What if we increased the throughput?