Reminiscent of HPE's comment at the In-Memory Compute Summit September 2017 - GSI's PIM paper would be interesting with a few changes. Also persistent memory will be announced most probably in May time frame which is required with GSI's which requires a non-destructive Read...,
I am concerned with the proclivity of police to "empty the clip" into an individual. At some point they seem to just make the switch to a violent video game without remorse or empathy for their actions. Shouldn't this be something that should be asked in gauging a police prospect before "arming" them with a license to kill?
Is this going pass the underwear stage stretching CNN to fit the Emperor's new clothes? Hinton's caught up in a mind boggling batch of mumbo-jumbo that proves no one understands how the brain actually works or is even close. Winter is nigh near. Till that time it's pin the tail on the donkey time...,
Why do we need to know how the brain works? We didn't build flying machines by creating artificial birds with flapping wings... likewise, nobody says we need to replicate wetware in order to create useful AI tools.
We seem to be lacking a theory of aerodynamics for neural networks. Most research seems to be judged by performance. The Aerodyanmic equivalent focusing on building a more powerful engine rather than wing design. The fact that we still visualize all the spurious w_ij weights in a network is a symptom of this problem. You wouldn’t show a fully connected circuit diagram for an 8 bit adder with spurious logical connections, but this is exactly what we do when we visualize NN topologies.
There’s some great work in Artificial Gene Networks that tries to tackle this problem. AGNs are mathematically the same as ANNs.
That's a fair point. And I don't mean to argue against continuing to try and understand how the brain works. But I don't think we should take "we don't know how the brain works" as a final argument against being able to build useful kind of AI.
Well, there's also observations from neuroscience. CNN's were directly inspired by observations of closely-grouped neurons firing when detecting certain images, and capsule networks are directly inspired by cortical columns.
I often see this metaphor as misleading somehow..
In both cases ( birds and airplanes) the physical phenomenon is exactly the same: lifting effect, which is related to the Venturi effect.. that is: the speed and the pressure of a fluid (in this case, air) depends on the geometry of the components affecting the flow of the fluid.
Note that the wing geometry in airplanes is actually similar to the birds.. (of course, airplanes have it optimized by CFD for the speed and altitudes they flight)
Although the “flapping” in birds’ wings also alters it’s geometry the main function is to act as the engines...
So, yeah, we need to discover the basic physical principle of the brain.
It's not a perfect metaphor, no. But pointing that out doesn't prove that we need a detailed understanding of how the brain works in order to build useful AI tools. Of course I think it goes without saying that more and deeper knowledge of how the brain works is desirable and would be useful. I'm just saying that that kind of knowledge isn't necessarily required in order to achieve useful ends.
Keep in mind the context of what I'm saying, which is responding to this:
Hinton's caught up in a mind boggling batch of mumbo-jumbo that proves no one understands how the brain actually works
I would argue that modern AI/ML is far past "mumbo jumbo" and provides useful tools, even without a detailed understanding of the brain.
1. Hinton's approach is not based on biology. He's not trying to make something that functions in a way similar to the brain.
2. Most people don't realize how much we already know about the brain. Numenta is one company that has working software based on those principles. We don't fully understand the brain yet but we're not clueless either.
Grinding wheat left byproducts of the stone burrs in the wheat product. This ground the enamel off peoples teeth leaving them nearly toothless by their mid-thirties. It wasn't till recently that "milling" became the norm in flour production. The wheat berry is not crushed in the milling process but is first burst and torn successively into every finer pieces. Some people have never tasted freshly milled flour made into bread. Total pity that...,
Getting the vast majority of your food from a single starchy source is also not great for dental (or general) health, even without the sand chewing part.
Robots can't taste. Something that no one really wants to take on for a number of reasons the first being that we haven't a clue how it works.
Variables in a Caesar Salad occupy wide spreads. Blending taste is still an art that most Chefs have trouble achieving as will the machine intelligence that conquers taste.
Wondering how machine intelligence will handle hunger pangs caused by the sudden thought of a Caesar Salad, ensconced with Pain au Levain Croutons, shaved Parmigiano-Reggiano and salted Anchovies on the side. Even the selection of the right sea salt affects the outcome of this wonderful concoction not to mention what mustard and Worcestershire Sauce is best used.
The other problem stems from the camaraderie that such a meal requires (haven't even got to the wine pairing yet) to be wholly satisfying. Some people have problems with the origin of the language that really counts as in "companion" - stems from "one whom with you break bread with"...,
I think you're romanticizing the concept of food preparation a bit much. You don't need to be able to taste, or feel hunger, to cook a good meal or choose a wine. I fully expect that a machine could not only do so, but ultimately do so more consistently than any but the most skilled of humans.
I fear such a thing would lead to Starbucksization of food and beverage production. As mentioned in plenty of other comments here, Starbucks baristas mostly just push a button to get a shot of espresso from one of their machines. They optimize for consistency and "good enough" - and indeed the machine pulls the same shot, every time, from beans which are over-roasted for consistency and good enough for mass consumption and a predictable experience across thousands of stores.
But that isn't what makes food and drink interesting. I enjoy going to a place because they pull a particularly great espresso shot and use it to prepare a macchiato in a way that I like according to proportions of milk and espresso that I think taste best... or a bar which has a particularly great wine collection according to my tastes... or a restaurant which has a fun twist on on a french classic which I like more than the original. Of course, each of these establishments could have a custom version of whatever algorithm you're envisioning which picks the perfect wines, pulls the perfect espresso shots, or prepares the perfect meal - but will they? Or will the platform that enables such extreme automation also enable such extreme commodification, like Starbucks vs a boutique coffee shop? I think it would.
I feel like you're talking about an experience that precisely emerged from the Starbucksization of food and beverage production.
Where I come from, staff in cafés would kill to make a consistent coffee every time (and they mostly do) but they're stuck with old machinery.
But now that Starbucks has delivered that, customers want to go back to an authentic experience that really has never existed.
The same thing will happen to food i.e most people would love to have decent restaurant meals delivered at a really low price. The fact that some people will try to taste gourmet food prepared by real chefs is rather irrelevant to say the least...
A machine doesn't have to optimize only for "good enough" or "burnt"; it optimizes for whatever parameters it was given. If there's a trick to "pull a particularly great espresso shot", then you can build a machine to do that. It's not even particularly hard, compared to, say, keeping up an intelligent conversation, doing laundry, or any number of other harder-to-automate tasks.
Of course - but people are going to control those machines, and the people who control the machines probably aren't going to be the people who (in another version of history where automation didn't take over) would be operating them with care out of personal passion.
In our Starbucks example - they also own Clover, the boutique automatic single-serve coffee brewer. It's a brilliant machine and makes a great cup of coffee. But it's artificially only available in certain markets, with certain collections of their own branded roasts. The macinery could be programmed to allow me to get a great cup of coffee at any Starbucks in America, but business incentives don't allow that (presumably - because Clovers are rare machines reserved for special store in major markets). Comiditization entails more than just automating the product and making things consistent and widely available. It also typically changes the incentives of the game and the people who are overseeing implementing them.
These machines aren't designed for foodies - they're designed to serve the rest of the populace that wants a decent, healthy meal at a reasonable price. That the machines try to appeal to foodies is just the side effects of marketing, in an attempt to provide some snob appeal to people who aren't true foodies.
There's always going to be a need for human food preparers, be they chef or cook, as foodies will seek unique experiences that cannot be provided in the large quantities that machines are designed to satisfy.
