Everything Christopher Nguyen said on any show that made the record, most notable first. Each card names its show and opens the statement there.
Nguyen: Energy bounds put human-level AI compute 50 years away
“And if you compute the equivalent energy required per neuron, and extrapolating that, it'll take about 50 years to get the equivalent human brain.”
Nguyen: Cost parity for human-brain level AI will arrive in 16 years
“There's another metric you look at in terms of traverse edges per second. And when you look at that, and you say, well, when would it take for it to become human in terms of cost? That's in about 16 years.”
Nguyen: Human-AI augmentation is humanity's best survival path against superintelligence
“So, my idea is, what if we could augment ourselves with this AI stuff we're building? That's probably the way out. That's probably the way we get away with this.”
Nguyen: Machine Learning Is the Primary Purpose of Big Data
“So it turns out the reason for big data is machine learning.”
Nguyen: Brain-machine interfaces will create hyper-evolved humans within 50 years
“So, if we implement this, and we implement brain-machine interfaces, within the next 50 years, we can become that hyper-evolved being, right?”
Nguyen: Big Data Is Defined by Learning Thresholds, Not Volume
“So I like something that Peter Norvig, the director of research at Google said when he referred to big data, he says, big data is not just quantitatively different, but it's qualitatively different. In other words, there's something that happens when you have …”
Nguyen: Machine Learning Mirrors Human Learning From Experience
“And the way I think about big data is when machines learn from big data is very much like human beings learn from life experiences.”
Nguyen: Human Intuition Functions Like Parameters in a Machine Learning Model
“Well, what we think of as intuition are actually, you can think of as parameters inside a machine learning model.”
Nguyen: Modern Business Intelligence Uses ML to Predict Unknowns
“You can think of business intelligence going forward as the ability to apply machine learning algorithms to big data, and not just look at past questions, but also future questions, or asking to predict the unknowns from the knowns.”
Nguyen: Big Data Progress Is Driven by Cheaper Tech, Not Smarter People
“We don't necessarily get smarter over time. It's just that certain technologies get cheaper. They get, they become more available. So machine learning algorithms have always been around. The data that exists that you could collect has always been around. But i…”
Nguyen: MapReduce Was Intentionally Designed for Reliability Over Speed
“Interestingly, a lot of people may not realize that MapReduce was designed to be slow.”
Nguyen: Apache Spark Would Have Failed Earlier Due to Memory Costs
“Now Spark, if it was created six, five, six years before its time would have completely failed because memory was so much more expensive.”
Nguyen: Users Abandon Software Tasks if Latency Exceeds Five Seconds
“We had a phrase we call the five second barrier. And if the user can't get something done, you know, within five seconds, they won't ever do it. It's not like they'll do it at, you know, at twice the latency.”
Nguyen: Machine Learning Will Become a Feature of Every Application
“What you will see is that all of this machine learning will be a property of every application.”
Nguyen: Users Will Soon Expect Machines to Learn Like Human Colleagues
“Well, I claim that there will be a day very soon when then you will feel that about the machines you work with. In other words, you would expect that to be a property of all these machines.”
Nguyen: Parameter servers are the most efficient distributed deep learning architecture
“So, it turns out the fourth one is the most efficient one because it doesn't have to communicate the model back and forth anymore.”
Nguyen: In 2016, Google Brain is at the intelligence of a cockroach
“If you look at the Google brain, it's at the level of a, about a cockroach. Ok? And it's got about a million neurons.”
Nguyen: Blind people can learn to see through their tongues
“And after six or eight weeks of training, blind people can learn to quote, unquote, see through their tongue, because the brain learns to detect those signals.”
Nguyen: Model communication creates a persistent bottleneck in distributed GPU clusters
“No matter how many cores you have, Even if you have an infinite number of cores, there's gonna be some amount of time needed to communication, to communicate that model.”