CES 2017: NVIDIA Keynote Live Blogby Ian Cutress & Billy Tallis on January 4, 2017 9:15 PM EST
09:18PM EST - NVIDIA has one of the major keynotes for the CES show
09:18PM EST - Set to start in just over 10 minutes
09:18PM EST - We expect to see Jen-Hsun Huang on stage talking GPU, Automotive
09:19PM EST - Ian on text, Billy on photos
09:23PM EST - While NVIDIA has done CES keynotes for several years now, this is the first year that they are giving the "prime" keynote. The most important and well-attended of the show's keynotes
09:24PM EST - For a number of years Microsoft held this slot. Since then it has shuffled among product manufacturers
09:25PM EST - Venue is filling up - being a main CES Keynote means a couple thousand people
09:26PM EST - What makes the opening/prime keynote so important is that the Consumer Electronics Association likes to set the tone of the overall show with the keynote. So for NVIDIA this is a very big deal; it may be many years where they have another keynote this well attended
09:26PM EST - Some familiar faces in the crowd too
09:27PM EST - In terms of press, that is (GTC fills a larger space, but it's largely developers)
09:27PM EST - 5 minute warning
09:27PM EST - Ryan will be adding commentary
09:33PM EST - Here we go
09:33PM EST - Starting with a 50-years-of-CES video
09:36PM EST - 'The magic of CES is that there's always something that blows your mind'
09:37PM EST - Gary Shapiro, CEO of CES Assoc, to the stage
09:37PM EST - Introducing Jen-Hsun Huang
09:37PM EST - 'Gaming is the largest entertainment business in the world'
09:38PM EST - 'When NVIDIA invented the GPU...'
09:38PM EST - 'NVIDIA is at the forefront of artificial intelligence'
09:39PM EST - JHH to the stage
09:39PM EST - In his usual leather jacket, of course
09:41PM EST - Video showing 'the promise of AI', robots, and future things
09:42PM EST - 'Our imagination opens up amazing new worlds, NVIDIA brings them to life'
09:42PM EST - 'What comes next?'
09:42PM EST - Video over
09:42PM EST - JHH : 'We are going through the most exciting time ever in computing'
09:43PM EST - NVIDIA is dedicated to a model of AI and vision computing
09:43PM EST - 'We dedicate ourselves to tackle the most challenging computing problems in the world'
09:43PM EST - 'We love video games - the highest volume, the most computationally intensive actiont the world has ever known'
09:44PM EST - 'Our technology is also used in the cloud, building AI supercomputers'
09:44PM EST - 'Also, the most exciting: self driving cars and autonomous vehicles'
09:45PM EST - 'Researchers working on deep learning meant the big bang of AI happened'
09:45PM EST - 'It allows software to write software'
09:45PM EST - 'Allows the computer to recognize complex patterns'
09:45PM EST - 'Representing complex patterns from sets of layers of simpler patterns'
09:46PM EST - 'Edges, contours, features, then finally a face'
09:46PM EST - 'It needs to deal with variability'
09:46PM EST - 'Deep learning is the key'
09:46PM EST - 'This foundational technology is difficult - the handicap is the amount of data processing is enormous
09:47PM EST - 'Then the AI researchers met the GPU, and the achievements have been fast and furious'
09:48PM EST - 'A network has to learn context too'
09:49PM EST - 'Reinforcement learning through trial and error meant a network could learn to walk by itself'
09:49PM EST - 'The ability to percieve the world through technology'
09:50PM EST - 'the enabling technology behind all of this is GPU computing'
09:50PM EST - 'Geforce - the #1 gaming platform, 200m geforce gamers, 2x revenue in 5 years'
09:51PM EST - 'Before anyone gets a console, they have a PC'
09:51PM EST - 'Our market moves forward by the content being produced'
09:51PM EST - 'VR is coming'
09:52PM EST - 'Gaming is the worlds largest sporting event - 100M MOBA Gamers, 325M eSports Spectators'
09:52PM EST - 'MOBA is a game of strategy, teamwork, knowledge, hence why it gets so many spectators'
09:53PM EST - 'Streaming is now a $5b industry'
09:53PM EST - Announcing Geforce Expereince connects to Facebook Live
09:54PM EST - Capture a video image/stream direct to Facebook, as well as live broadcast
09:54PM EST - Connect via two clicks to facebook
09:54PM EST - Aaryn Flynn to the stage, general manager for Bioware
09:54PM EST - they're going to show footage of Mass Effect Andromeda
09:55PM EST - 'Mass Effect has a great story'
09:56PM EST - 'The people watching are going to tear down every frame of this upcoming preview'
09:57PM EST - 'Real in-game footage with a GTX1080'
09:57PM EST - Video now
09:58PM EST - Yup, it looks like Mass Effect
09:58PM EST - definitely engine footage
09:58PM EST - shooting flames from the hand
09:58PM EST - guns guns guns
09:59PM EST - (I've only ever played the first Mass Effect at any length, and that was what, a decade ago?)
09:59PM EST - 'Captured in engine, representative of game experience'
09:59PM EST - March 21st release date
10:01PM EST - '1 billion PC users are not game ready'
10:01PM EST - 'no real way of installing a GPU into those PCs'
10:02PM EST - 'if we could put a PC into the cloud, like AWS, then consumers could game easier'
10:02PM EST - e.g. GRID using Pascal
10:02PM EST - 'Launch a game wherever, whenever'
10:02PM EST - 'Any latency ruins the experience'
10:03PM EST - Announcing GeForce Now for PC: Gaming on Demand
10:03PM EST - Turns any of your PCs into your most powerful gaming PC
10:03PM EST - Giving an example, PC laptop and a Mac
10:04PM EST - NVIDIA has been offering streaming for a few years now, both gaming and their virtualized desktop GRID solution
10:04PM EST - Launching Steam on GeForce Now on the PC, takes about 15 sec
10:05PM EST - A few more seconds, and your personal steam account is there
10:05PM EST - Also works on the Mac
10:05PM EST - All the same games from your personal account
10:05PM EST - You can buy a game in the interface, and games are downloaded in a minute
10:06PM EST - Running Rise of the Tomb Raider on the Mac
10:06PM EST - At a reasonable quality too
10:07PM EST - 'Video Games for the other billion users'
10:07PM EST - Available in March for early users
10:07PM EST - The current iteration of GeForce Now served up to NVIDIA's mobile devices is very game-centric. That is to say, you buy the games through the service
10:07PM EST - $25 for 20hrs of play
10:07PM EST - 'A gaming PC on demand'
10:07PM EST - Several grades of performance, the more performance = lower time per $25 credit
10:08PM EST - That they're instead offering a virtual machine on demand that can link into your Steam account is a significant shift in how the service is offered
10:08PM EST - Note, you also have to personally buy the games on Steam
10:08PM EST - I wonder if it counts the time you're not in a game
10:08PM EST - Now Android TV
10:08PM EST - But the lack of Steam integration has always been a friction point with GeForce Now. Customers don't want to buy a game twice
10:08PM EST - 'With a powerful computer connected to a store to running applications, an Android TV Console: NVIDIA Shield'
10:09PM EST - On the other hand, now they need to pay for service by the hour, as opposed to GeForce Now's flat costs
10:09PM EST - 'The performance is always getting richer'
10:09PM EST - Announcing a New Shield, supporting 4K HDR
10:10PM EST - Worlds first entertainment platform to support Netflix and Amazon content in 4K HDR
10:10PM EST - A steam for shield, that connects to your PC, to enjoy 4K gaming on your TV from the PC
10:10PM EST - 1000 games in the NVIDIA shield game store
10:11PM EST - The two most popular consumer electronic platforms are Smart TVs, but the other is the Amazon Echo
10:11PM EST - 'It brings AI into your home'
10:11PM EST - 'We thought, why have two devices when you can have one'
10:11PM EST - 'We worked with Google to create the worlds first TV with Google Assistant'
10:12PM EST - 'Now the TV can be controlled by natural language interaction'
10:12PM EST - Doesn't AppleTV already do that?
10:12PM EST - 'You shouldn't have to lean over to a table to talk to the TV'
10:13PM EST - To be fair, the AppleTV requires a remote to do that. NVIDIA is suggesting an ambient assistant
10:13PM EST - Announcing NVIDIA Spot
10:13PM EST - Plugs directly into the wall
10:13PM EST - because the computing is done on shield, can have them installed over the house
10:13PM EST - far field microphone
10:14PM EST - voice triangulation via beamforming if multiple devices pick you up
10:14PM EST - all goes to one SHIELD over WiFi
10:14PM EST - Now a video for Spot
10:15PM EST - a guy in the video saying 'OK Google' a lot
10:15PM EST - Seriously, this video just shows how 'OK Google' is too many syllables
10:16PM EST - NVIDIA Home AI = Google Assistant + Smart Things + SPOT
10:16PM EST - 'We should build Jarvis for everyone'
10:17PM EST - new SHIELD for $199, pre-order now, available later January
10:17PM EST - SPOT to be released in the coming months
10:17PM EST - Now, AI for transportation
10:18PM EST - $10T transport industry
10:18PM EST - $199 is also where the last-generation Shield Android TV launched
10:18PM EST - 1B cars, 20m taxi rides/day, 1.2T miles from trucks a year, 500k buses in operation
10:19PM EST - 'The amount of waste in the transport industry is huge'
10:19PM EST - >so self driving helps reduce waste
10:20PM EST - JHH is basically saying self-driving cars helps reduce waste
10:21PM EST - 'GPU Deep Learning has made it possible for us to realise the self-driving vision in the next year'
10:21PM EST - 'Perception, Reasoning, Driving, HD Mapping, AI Computing'
10:21PM EST - 'Will deep learning we can percieve the world, rather than just sense it'
10:22PM EST - 'We can teach a car to drive by watching a human driver'
10:23PM EST - 'We can compare perceptions live with knowledge in the cloud to act, and you can keep learning'
10:24PM EST - 'Xavier AI Car Supercomputer: 8-core custom ARM64, 512 core Volta, 30 TOPs DL, 30W'
10:24PM EST - Announced last year
10:24PM EST - Runs Driveworks OS
10:25PM EST - 'Designed for ASIL D Functional Safety'
10:25PM EST - 30 TOPS = 30 tera-operations
10:25PM EST - For more details on Xavier, please see the original September announcement: http://www.anandtech.com/show/10714/nvidia-teases-xavier-a-highperformance-arm-soc
10:26PM EST - SHowing the BB8 autodriving car in video
10:26PM EST - Ryan is having a demo in it later this week
10:26PM EST - 30 TOPS DL is 50% higher than when Xavier was first announced
10:26PM EST - Picking up speed on on-ramp, merging at speed
10:27PM EST - video is cutting to different perception metrics
10:27PM EST - disengage autopilot via a voice command
10:27PM EST - I think Ryan will have fun later this week
10:28PM EST - A self-driving car with the theme song "do something crazy" is certainly an interesting combination...
10:28PM EST - 'These cars should be able to drive from A to B practically anywhere in the world'
10:28PM EST - 'It can determine its confidence level on paths it doesn't know, and hand back control if it doesn't have enough confidence'
10:28PM EST - 'AI should be a co-pilot'
10:29PM EST - Announcing the AI Co-Pilot
10:29PM EST - The car has perception, cameras and speakers. It knows where it is and the state of the driver/passengers
10:29PM EST - 'This car can perceive, and so if it runs all the time, it can either drive you, or look out for you'
10:30PM EST - 'When it doesn't have the confidence to drive, it still has everything working to tell you what it thinks and is completely aware'
10:31PM EST - Showing a cyclist ahead, or a motorcyclist behind changing lanes
10:33PM EST - AI CO-PILOT does face recognition on the driver - it knows who is driving and their demeanour
10:33PM EST - it does head tracking and gaze tracking, so it knows which way you are looking
10:33PM EST - Also lip-reading
10:33PM EST - 'Take me to starbucks' and it can tell
10:34PM EST - 'wouldn't be nice' - so this is still a goal for NVIDIA
10:34PM EST - Lip reading can be 95% accurate based on current research. Human lip reading is about 53%
10:34PM EST - >but then again, smart AI assistants are meant to be 95%+ accurate too
10:35PM EST - 'This is the NVIDIA AI Car Platform'
10:35PM EST - Drive PX, Driveworks OS, Auto-Pilot, Co-Pilot, Mapworks, Cloud HD Map, Cloud AI Assistant, NLU
10:36PM EST - DNN = Deep Neural Net
10:37PM EST - 'A co-pilot that will automatically open the gates on your driveway as you pull into the street'
10:38PM EST - JHH is describing features of the car platform, like Mapworks
10:40PM EST - Announcing partnerships with Zenrin and HERE
10:41PM EST - ZF is now a partner - they're the leading truck and commercial supplier in EU, top 5 worldwide
10:41PM EST - ZF is first to announce a Drive AI to the market
10:42PM EST - Announcing Bosch as adopting the NVIDIA Drive computer
10:43PM EST - Bosch is the largest supplier to the automotive sector
10:43PM EST - Major update on the partnership in March
10:44PM EST - Announcing, Audi and NVIDIA to partner to build next gen AI cars
10:44PM EST - On the road by 2020
10:45PM EST - Audi to the stage, Scott Keogh, President Audi US
10:45PM EST - Audi is a frequent NVIDIA automotive collaborator, so this isn't too surprising
10:46PM EST - 210k Audi cars sold in US last year
10:47PM EST - Within 4 days, a learning car can already navigate obstacles
10:48PM EST - 'We're talking level 4 autonomy by 2020'
10:50PM EST - 'Let's make sure none of our kids ever have to learn to drive'
10:52PM EST - Wrapping up now: GeForce Now, NVIDIA Shield and Spot, Drive AI Car Computer