Jensen Huang and Marc Benioff Discuss AI Agents Collaborating at Dreamforce 2024

please welcome uh Jensen Wong Jensen come on out to have you our industry is you know call it about a trillion dollars and it's an industry of tools you know computers um software their industry of tools for the very first time this is going to be an in Industry of skills and you capture that that phrase and you uh you call it agents but it's going to be for the very first time agents sitting on top of of tools agents using tools and the the opportunity for agents is gigantic as we know you know we

we now know we now know I'm hoping so by the way uh yeah it it sounds it sounds insane but here's the amazing thing we're going to have agents that obviously uh understand understand the subtleties of the things that we ask it to do but uh it can also uh use tools and it can reason and uh it can reason with each other and collaborate with each other and you know going to we're going to give we're going to give a problem to uh agent force and uh uh agents are going to go find other

agents uh that that can help achieve this Mission uh they'll work together assemble together and work together and solve this problem and and so I I think the the the Breakthrough uh for for me was was at a moment when when um we realized that unsupervised learning was going to be possible because the ability for humans to to uh be able to L label data at scale uh we would become the limit limiters of these digital digital AI to to uh uh expand their capabilities and uh when when unsupervised learning came along which allowed us

to use language models to create language models which codify human prior knowledge using that to now learn multimodality multimodal uh uh data uh from that point forward the scale was going to be exponential now the you know for for everybody here this is this is an extraordinary time because uh in no time in history has computer technology not only uh uh moved faster than Mo's law I mean we're Mo's law for example over the course of uh over the course of a decade would be about 100x um we are probably advancing at somewhere near 100

that was our other neighbor in Hawaii Gord more oh is that right great person great leader great person and and so we're we're we're at a stage now we're in an era now where where we're moving way faster than Mo's law and and um arguably easily Mo's law squared and and the reason the reason for that of course is at every single layer computers went from CPUs to gpus uh from Human engineered software to machine learning software and and now this feedback loop um that allows us to to um uh create new AIS uh and

these new AIS are helping us create new computer systems and these new computer systems are uh advancing at such incredible rates which allows us to create even better AI uh that that feedback that flywheel is really flying now and and so I think I think the the um uh the progress that you're going to have with a agents over the next year or two is going to be spectacular and surprising the challenges that we have in front of us are are many and worthy to be tackled right away um of course some of the some

of the most most um uh inspiring advances has to do with fine tuning supervised um uh guard and Guard railing and and um uh and uh um all of the work that's surrounds safety um everything from using AI to curate the data to create safe curriculum to teach the AI which then fine-tunes the AI on particular skills to grounded on values then then all of this wonderful AI technology to guardrail it um and Guard railing and then reflection uh using using Chain of Thought to reflect on the quality of the answers that it's producing you

know it's no longer AI this one shot producing information it's now reasoning about is this is the the the uh the answer I'm generating uh sufficiently safe and proper and is it the best possible answer I could be I could be providing it's reasoning with itself and it's uh it's reflecting on the answers before it answers and so I I think the the progress that we're Mak making in safety is phenomenal isn't it so important that people can actually understand uh this is what I call they need to get their hands in the soil they

need to understand the basic ideas to make this possible and that's what I hope you know happen with the thousands of people here that they can get in the code while they're here yeah to really understand so when they're going back to their offices you know that exactly what you're saying that we've demystified the AI that you know they don't have to worry about how they're going to DIY their AI that they can somehow put all the hard work that your team has done right into some very practical use because this leap of faith that

you've had this this strategic motion of your company has enabled so much of this now how do we put it in the hands of everybody to like just get to work in building an agent should not be some computer science uh fair project it should be something that we can easily do because it's going to be a lot more like onboarding employees than writing software so I think the the ambition of wanting to reinvent Computing uh to create what we now know is is is such a great Endeavor um so some somewhere between the the

being uh inspired by all the stimulus um the the the incredible challenge ahead of us you know it's kept us propelled um inspired and fired up for a long long time and now here we are we now have the instruments the tools uh this capability called artificial intelligence that goes solve all of those other problems that we've you know been excited about ever since we were kids and and so that's pretty exciting times yeah the summation of that whole thing right because now that I know like you were there when the tr80 model 1 was

there and we were at Radio Shack and we're working on that you know basic and then we're moving into the 6502 computers and the commodor and the Ataris and the apples and all of that and then all of a sudden yeah we're under these mini computers and mainframes and now all of a sudden we're in this like next Generation systems that can do things that we could have never imagined and that's right if somebody had asked us when we were in high school to start programming on those things we'd be like whoa yeah unbelievable times

and and so I think for for all of the industry this is what we're all fired up about because we're now seeing it and we get the benefit of being inside the industry looking at the early indications of all this work and and I'm excited to be here today uh helping to celebrate the beginning of the Enterprise AI ERA this is this is a big deal don't forget you know we say Enterprise is a name of an industry or you know um but the fact of the matter is this is how Society is built this

is how Industries are built and we're going to we're going to bring a level of uh automation capability that the world's never seen and and I love that you codified it as agents uh gave it a theme song um