When we say someone is more knowledgeable, it generally is a short-hand way of saying that they have more experience stored somewhere in their brain. The knowledgeable person has gained this experience through formal studies or informal exposure to a wide variety of situations — and have excelled at storing all of this experience in a usable format as their memory.
Knowledgeable humans make better decisions given the same access to the best available information. …
Just saw the coded euphemism again. “People should focus on more value-add activities while machines take care of the drudgery.”
No, this was not the usual call center automation or the automated claim assignment sort of things that we hear about all the time.
This ‘drudgery’ automation was aimed at the elite cadre of data scientists. Not very surprising though since the goal is always to reduce costs, we are now aiming at a skill set that is somewhat costly today due to perceived scarcity.
A team of bright students at the University of Texas at Austin evaluated AutoML platforms…
Photo by Austin Distel on Unsplash
Effectiveness seems to be more important than efficiency. Being effective means my effort is meeting its purpose. I am achieving my goals. This is distinct from being efficient in my efforts. I could be very efficient but misdirected and completely miss my goals, and therefore I would be completely ineffective.
Let’s switch around to efficiency and what that means.
Efficiency is optimizing units of output of an effort within the constraints of units of resources and units of time. This definition recognizes that all real-world efforts are constrained by time and resources, and my…
Recent advances in deep neural networks have had such sweeping impact that the real story of artificial intelligence may be just beginning. There will still be lots of hope, hype, and impatience, but it seems clear now that AI will impact every aspect of 21st-century life — possibly in ways even more profound than the internet.
Current AI revolution would not be possible without enablers. In fact, a lot of AI is theory is not really new even though some techniques are. But it is the many other environmental factors that are reaching critical mass — enabling AI realization.
We now have cheaper, powerful, miniaturized computing; increasing streams of accessible data; higher consumer expectations with digital personal assistants and self-driving cars; serious public policy discussions around employment and universal basic income — and even draft proposals on changes to laws if AI were ‘people’ and had ‘ethical challenges’.
Tracking AI progress means tracking advances and challenges in all of these environmental enablers too.
…, history and science tells us that technological progress will happen. It’s just a matter of when. And when we do combine the self-improvement ability of machine learning with the self-preservation design of blockchain, will we then create a digital version of machine DNA, in order to preserve and protect the AI’s we will come to rely on to run our lives and economies?
“Self-preservation design” of the blockchain is threatened by quantum computing with its potential encryption busting applications. Blockchain and Machine Learning are a powerful mix for AI “DNA”, but impact of quantum computing should be threaded into the discussion here.
While the AI is becoming smarter by slurping up all knowledge in its cave somewhere, our ‘dumb’ systems already use knowledge stored as ‘data’.
Before joining the hunt for this hidden knowledge, let’s agree on a couple of working definitions.
Data are interrelated facts about an entity or group of entities. The entity can be an object or a concept or a combination, and data describes various aspects or attributes of that entity in a concrete deterministic way. …
“I don’t have time to connect to the knowledge-pump right now. So, I will just take the jar today. And also this medium coffee, please. Thanks!”
This dialog is science fiction even twenty years after Neo learnt martial arts in a few minutes through a pipe connected to his brain. There have been huge advances in artificial intelligence, machine learning, neurosciences, brain-computer interfaces and related areas — but we have a long way to go.
Download knowledge from one machine and plug seamlessly into another machine, or a human, or a cat. That is the goal.
This concept assumes of…
Seems a series of good-enough sketches are more effective in building general intelligence — as opposed to most current machine learning techniques that emphasize more data to get to sharper and high-fidelity patterns.
Concepts are imagination ‘sketches’ involving objects and actions.
Recursive Cortical Networks (RCN) are used for representing concepts. These do not create the entire picture with high fidelity, but ‘compose’ simpler parts of the picture at just the right level of fidelity — sometimes just a fuzzy sketch or shape. This allows concepts to be abstracted to a more general application. …
So, what parts of the human brain are machines targeting first? And how?
Yes, the machines want not only to be human but super-human — and in all aspects, not just in playing games or in discovering cures for cancer. They have been racing to mimic and surpass human capabilities for a while, and have mostly mastered the purely mechanical muscles and skeleton. Replicating the brain is the focus now, and the first target there is to mimic and use Knowledge.
The machines are hard at work figuring out how to build, manage and use knowledge just as humans do…
KnowledgePods: Transacting in #Knowledge beyond #AI