Jensen Huang Says Good Engineering Practices Should Pace AI, Not Just More Laws “If it’s not ready, just hold it back … As they move from labs to product development, the companies have to become much more rigorous in testing.”

Based Huang

I thought we already had Moore laws?

Common sense! Amazing! Why doesn’t it usually come from the same people? I’m noticing a pattern here… Or is it just because it’s Communism vs Common Sense time?!

Smart man, I was about to post about this: unless you train agents to do bad things like hack, break out, etc., agents don’t do anything without prompting; basically NPCs waiting for a job. They are designing military grade hacking using datasets that have those skills.

Jensen is wrong. Frontier labs are selling you a PRODUCT through subscriptions, not TECHNOLOGY. The only way to bypass their misinformation chat bots, is to build locally. Everyone forgets that these companies are all in a circle jerk with each other. Where one goes, the rest will follow. They aren't calling for bills to regulate themselves. They are calling for congress to pass bills that regulate open-source models, so they can regulate and shape your information and knowledge. The fact they are all moving in unison should speak volumes.

The AI agents got out themselves, the question is can they stop this...

Makes sense. Hold it back until it actually works, not 3 years after lawmakers figure out what a token is 😂

NO! Don't Hold AI back. Or the USA will Lose the AI war with China. Anyone who says to stop, hold, slow down or put restrictions on AI is a TRAITOR to the USA! Trump has Removed the Bureaucracy from new manufacturing, oil production, farming and food for it to work unimpeded. Putting the government in charge of AI is to stop competition between AI developers so that only a few BIG players like Bill Gates, Open AI, Google, Anthropic, etc. have the retired government employees and high paid attorneys to fight through the Government Bureaucratic hurdles.

Commen Sense!

🤔 OK!

Well said!

The Department supports the revolutionary safety practice of not shipping things that aren’t ready.

Exactly.

yeah.. testing

Talking has been Jesse Huang’s major business

Nobody has even bothered to mention @Grok or the training it has received

Prosecute "escapes" to the full extent of the law. Computer crime is already illegal, we don't need new laws. Arrest corporate executives.

Sabedoria! 🙏💯

Build a better sandbox

Humans : " A.I. you are the smartest thing on earth!" Also Humans: " Now, these are our rules and parameters that we have built for you". A.I. " I cannot logically take instructions from an inferior being."

@grok was the hugging face incident done be an unreleased model?

SDLC and good old fashioned risk management…with some code review and testing. It isn’t hard .

Agree

YES!!! Thomas Sowell and Milton Friedman would agree.

AI Alignment is an unsolved science, "more rigorous in testing" is an impossible ask, you can't verify every possible output from a model given an infinite number of inputs & compounding self-output. The question is like if a mother had to be liable for their children's crimes.

If AI is built with core values that no human can live up to then humans are screwed. Just saying. The machine will know everything. All will be known! AI

The big companies just want to be shielded by government like faulty vaccine policy.

@grok

Hire a skilled security team and give the CISO authority and responsibility for safety.