Anthropic CEO: Open-Source AI is getting dangerous (2023)

novaRom•about 1 hour ago

Creating AI monopolies is more dangerous. How much better would it be for environment and privacy if instead of concentrating AI power in few data centers it's distributed across all compute and memory devices? If just few players control the narrative and the infrastructure it's a single point of failure or worse a single point of control. Bad for democracy, society, pushes inequality.

gorgmah•about 4 hours ago

They just want US authorities to ban big corporations from using models like GLM etc. so that they can keep selling their overpriced tokens. Funnily enough, a ban like this would close to impossible to enforce on individuals, so I guess their lobbying efforts will soon be met by lobbying efforts from big corporations that are losing from this.

butlike•14 minutes ago

I agree with you and just want to add how crazy it is to hear these statements from CEOs as you get older. They literally, literally just say the thing that will maximize profit for their company. Whether it sounds altruistic or not, it's ALWAYS to increase shareholder value (money for the company).

InkCanon•about 3 hours ago

Completely stopping people from using open source models is impossible, but if it became a federal law there could be incredible pressure exerted on US companies. For example, OpenRouter might be compelled to not accept open source requests on threat of fines.

bluefirebrand•about 2 hours ago

I'm just a guy who writes code and geopolitics and big business is usually above my head

That said, I have a suspicion that if the US exerts that kind of pressure on US companies, they will just pivot to using offshore contractors that are doing the things they want

pautasso•about 2 hours ago

We have no moat

https://newsletter.semianalysis.com/p/google-we-have-no-moat...

InkCanon•about 5 hours ago

Paraphrasing him, he says that open source software which tool tens or hundreds of millions to build should be classified as a different category. And the implication being they should be banned. Why?

kombookcha•about 2 hours ago

'These tools will become dangerously powerful, which is why nobody should be allowed to have them except through buying them from specifically me' seems a lot like motivated reasoning to me.

DaSHacka•42 minutes ago

Seems like they haven't learned from the last time they played with fire and got their models restricted to only U.S. citizens for "national security reasons".

RayBye•about 5 hours ago

Of course, we should only trust Anthropic on this matter.

unfixed•about 5 hours ago

Dangerous for their business model I guess.

petterroea•about 2 hours ago

If they are getting regulated to death they may as well bring the rest of the industry down with them, probably

Balinares•about 2 hours ago

The interesting thing here is that Amodei is implicitly admitting he foresees open-weight models (and not "open-source", please...) reaching Mythos-equivalent capability before much longer.

Which, I mean, d'uh; if you've been paying attention it's not a wild forecast.

But it's interesting that he's admitting it.

ChrisArchitect•about 1 hour ago

(2023)

Testimony from July 25, 2023 U.S. Senate hearing "Oversight of A.I.: Principles for Regulation."

https://www.judiciary.senate.gov/committee-activity/hearings...

clbrmbr•about 1 hour ago

Do we have recent commentary on open-weight from the Anthropic founding team?

rvz•about 5 hours ago

Dangerous for them. Great for everyone else who want alternatives.

That means open weight models are good enough to threaten Anthropic's bottom line so much that Dario needs to get governments to ban the release of open weight models.

khurs•about 5 hours ago

Disappointing behaviour from him.

Glad someone was watching and clipped it.

Soarez•about 4 hours ago

dario is a faux leader, one of many in the us. western tech is decrepit. best content these days comes from juejin

Anthropic CEO: Open-Source AI is getting dangerous (2023)

⚡ Community Insights

Discussion (18 Comments)Read Original on HackerNews