Post

P(aul) Frazee
@pfrazee.com

In 2027, mysterious hacks attributed to open source weights and definitely not connected to the companies that accidentally hacked a bunch of companies in 2026, which was okay

Quantian@quantian.bsky.social

lol two months ago OpenAI was signing a letter about how important open weights models were and now their house propagandist is saying “we will ban open weights and you will thank us for it”. You can’t hate this company enough

P(aul) Frazee
@pfrazee.com

The safe future is one where a few corporations run everything for us. That has made everybody happy and safe.

7 reposts192 likes

Brian Slesinsky
@skybrian.bsky.social

Open weights models will be misused because it’s the Internet. If we rule out regulation, the Internet just gets more dangerous. Maybe we huddle behind Cloudflare and hope the botnet swarm doesn‘t find a way through? I don’t like that scenario either, what’s a better one?

P(aul) Frazee
@pfrazee.com

Corporations never change hands and become corrupted against their original values or motives. I bet Anthropic will always be run by the right people

cyberrb25
@cyberrb25.bsky.social

Don't forget the other part: Where a few corporations run everything for us while the citizens' representatives think they're in control of what those corporations do (or even worse, are in cahoots with them).

Joe Fabisevich
@mergesort.me

To be fair both ends of the spectrum are an unstable equilibrium. The future where a few corporations run everything for us is bad because most people are powerless, but most are powerless in a future where everyone (technical or not) has to defend themselves against potential rogue AI cyberattacks.

Tim BooOOooOorga
@timburga.com

that's a false dichotomy though rogue attacks will continue to happen in both scenarios, so the differentiator is who will respond more effectively and as the history of the internet has demonstrated, the ingenuity of a bunch of pissed off oss nerds is usually the safe bet

Joe Fabisevich
@mergesort.me

The internet is full of scammers and cybercrime though. In fact, they take advantage of the exact same people who will be harmed if we *only* have open models with no guardrails built into them. I'm not saying we should only have closed models, but equilibrium is likely a balance of both.

Joe Fabisevich
@mergesort.me

No, I'm saying this from my experience working with trust and safety and government affairs for a product that had 500 million users. My point is that I don't actually want one extreme or the other, I think the right balance is both because *only* having open models has it's own set of consequences.