Loved the interview! I love the fact that Jensen doesn't have the vibe of a traditional tech CEO - it's almost like you were chatting with an engineering manager! Plus he seems like a super humble and kind person. The only point of disagreement with him would be at 41:50 where he said that the metaverse will be in 2D. I like to think it's obvious that 3D is the future. The only reason we're stuck with 2D is historical and has to do with tech limitations. It's much more natural to interact with objects in 3D. I used to work on the HoloLens project back at Microsoft. Give it 5, 10 years and it will slowly get there!!
Man, Mr. Huang, if you read these comments, you're something else and give me hope for our future! Thanks for that, all you've done. The Noyce Award was beyond well deserved. Keep it going, keep it going! LEARNER FIRST HERE TOO!
impressive great questions ! Jensen Huang: one of the most truly inspirational visionary and strategic CEO of our time. “One of the most important missions - and the purpose - of a CEO is to create the conditions where amazing people could do their life’s work … As a CEO of a tech company, you really need to enjoy learning about what’s happening in your company and what’s happening around the industry, and see if you could imagine a future that’s better for everybody”. ‘I’m never different (in relation to stock prices), I don’t think it’s possible to find a correlation between my behavior and the stock price … I completely separate the financial success of the company from the importance of the work and doing impactful work.
Very interesting. The digital twin on a system embedding physical law models augmented with statistical models based on real-time data has been suggested for the last couple of years. This in combination with an internet of things system for real-time control. If it has not already been done it seems close to realization.
The biggest constraint on machine learning at present is the lack of support for sparse models. The fact that Ray tracing requires sparse processing should point to a synergy.
20:15 Dude wants to be build a full scaled digital twin of the EARTH! I mean he wants to literally recreate the entire f**king planet!! I'm just trying to wrap that around my head. Mind=BLOWN!!!
The future is basically more industry normalization around certain conventions and models within AI that will allow for a large amount of infrastructure, products and services that companies can use for their own AI solutions in the cloud (and off). Which is cool because a lot of corporations are still tepid about taking the leap fully into AI for various reasons, especially related to cost and risk. By having industry standard frameworks that provide portability for IA solutions and infrastructure and platforms as a service in the cloud for running those solutions at scale you can help ease some of the burden. It will not address all of the reasons for slow adoption of AI as a whole across the board but make it more appealing for those kinds of applications that can most benefit. The only downside to this in the long term is if there is no big payoff in such industry adoption of those common frameworks and models as attitudes change, economic justification changes and other real world factors affect changes towards AI. Meaning that it could potentially lead to another kind of AI winter if those platforms and services don't take off providing justification for further R&D and investment which already is mostly concentrated among large big data and silicon valley corporations.
16:00; There you will find Luke's very valid question about the importance of the ever increasing proliferation of ever more energy hungry computers and data centers. These Nvidias DGX STATION and the large HPC computing clusters for machine learning NNs built from the A100 Tensor Core GPUs consume megawatts of power to generate information, depending on their size. It is frightening how Jensen desperately tries to deflect from this issue. How he doggedly tries to redirect the question in another direction. He tries to avoid addressing the great tragedy of today's NN hardware. The tragedy is the unspeakable consumption of very precious energy to run the machine learning training phase. A large part of the world's population lives in poverty. Energy is for many people of the earth an almost unattainable precious good. And companies like Nvidia are constantly developing and building new Tensor computing units with an ever more frightening hunger for energy. Whereby almost the entire power supplied has to be dissipated into the environment in the form of heat. This is absolutely depressing and very unfortunate. Shall in the near future our AI NN intelligence machines consume all the energy we generate? Which means nothing other than heating up the environment further and further at an ever faster pace? Here Nvidia must become the pioneer of the information industry, inden it achieves to take the example of the biological models seriously and to reach and surpass their consumption of energy per information unit. The current path of machine learning with the current hardware is the path to the abyss.
"AI needs to learn the law of physics" so do we get good NPC in elder scrolls 69, a digital twin that sucks away our social life even more, or just good ol' skynet? Really not sure what my takeaway that ought to be lol
