Palo Alto Networks CEO Nikesh Arora says training AI on corporate data can elevate average workers to 95% of top-performer output

LLMs make expert knowledge scalable. That means the best use of your top people is no longer just having them do the work, it’s using their work to train agents that can do it better, cheaper, and at massive scale.

The highest-leverage experts don’t just solve problems. They train the models that solve them.

@tbpn @nikesharora Yap yap yap what’s with these word slop do nothing type people

Go do something

@ambig_art @tbpn @nikesharora "Go do something"

@tbpn @nikesharora I'd posit that this is not actually the biggest risk in business. A bigger one might be, say, your product/service sucks and you go bankrupt cause nobody buys it.

"The biggest risk in businesses is, you hire 100 marketing people. Not all of them are equally good."

@tbpn @johncoogan @nikesharora The “if I train…” part is carrying a lot of weight in the equation. Also, if you train using the experience and input of the top 5% they should expect to get rewarded accordingly.

@tbpn @nikesharora look at all the great hair in this video

@tbpn @nikesharora Slop.

@tbpn @nikesharora This is sooooo boring ! ! ! Go do something

What do you people do other than just word slop word commit about nothing?

@tbpn @nikesharora Yall need to up your dose of something here this is more boring than talking to gpt and that’s the lowest of low bars bcuz gpt is boring

@tbpn @nikesharora Can you morons do something already like enough with the nonsense talk start soldering go build something

@tbpn @nikesharora Go get off the spectrum and drill for oil

@tbpn @nikesharora Go drill for oil or something this is just sad

@tbpn @nikesharora The point of B players is that they’re not at the top of the distribution. “Democratizing intelligence” just means there’s a new meaning of B player. So lazy

@tbpn @nikesharora This is backwards. AI is a leverage multiplier, not an equalizer. The tide is rising but yachts are rising faster than dinghies.

@NickThoughtRepo @tbpn @nikesharora heres something i've done

your turn

@ypatil125 the hard part isn't even using lls for training, it's making sure the output is good enough to warrant automating in the first place. most teams can barely keep up with maintenance, let alone produce something worth replicating