Cybiote

34 ms·

211.
by Cybiote 10y ago
I think you're missing the point there. Portia is able to, using just a handful of neurons (~600K), accomplish sophisticated vision tasks, complex planning, online learning and scheduling. Portia and insects in general are much more
212.
by Cybiote 10y ago
The large majority of deployed machine learning algorithms actually can't learn. The bias is so large that most (all?) specialized hardware assumes that only inference will be done. In contrast, there is evidence for online adaptabilit
213.
by Cybiote 10y ago
> The same technology which beat Go could also, in principle, be used to play a real time video game that requires hunting and stalking things. If you generalize things enough to mean application of linear algebra, the chain rule and map
214.
by Cybiote 10y ago
Interesting idea about the eye-tracking for VR: offloading computation to the brain! I'll make a correction that while minor, makes a world of difference: filling in the blank is more accurately understood as the brain making bayesian
215.
by Cybiote 10y ago
Telling people to remain alert will do no good because it is simply asking for too much. A person is supposed to sit still, not driving but maintaining their full attention while being ready to automatically take over in an emergency situat
216.
by Cybiote 10y ago
This makes intuitive sense but is not borne out by research such as [1]. The < 20 mph scenario, one of traffic, consists mostly of stress and the general unpredictability and frustration of the whole scenario. The problem of Level 3 auto
217.
by Cybiote 10y ago
With enough labor, you could annotate enough sentences to cover a lot of dialogue cases. Sections like "'stop!', he said angrily/dryly/mockingly are probably fairly common. You'd be modeling the next most proba
218.
by Cybiote 10y ago
That doesn't contradict anything I said, see my first paragraph. The core point is that you don't need to have full metacognition in order to build a culture that learns over time by explaining things and transmitting concepts. Yo
219.
by Cybiote 10y ago
Illusory Introspection and confabulation are fascinating but generally have more to do with what was going on in your head when you took that action? than can you explain why you think that? . XAI is not interested in psychological state
220.
by Cybiote 10y ago
A big part of this is that typing and dictation are very different from conversation. Conversation is free flow, very flexible to errors (since people can infer each other's true meaning) and interactive. However, in both typing and di
221.
by Cybiote 10y ago
I think the holiday pay example is more glaring. Seen in isolation, I would be confused as to what on earth it was getting at. Furthermore, the summary is no good. The abstract isn't either. My summary would be: British Gas continues
222.
by Cybiote 10y ago
I didn't think it was a failure either until I realized that I was letting future knowledge leak into the past! There is more than one internet company so upon reading that headline, given that it must be a novel event, my question w
223.
by Cybiote 10y ago
Maybe I need to clarify? A certain amount of energy must be expended to reach a particular velocity. For a relativistically relevant velocity, kinetic energy is given as: mass * c^2 * (1/sqrt (1 - v^2/c^2) - 1); it's clear th
224.
by Cybiote 10y ago
You can't travel as fast as you want because it will take ever larger amounts of energy to accelerate each fraction of a percent closer to the speed of light. You're effectively bounded above by the speed of light.
225.
by Cybiote 10y ago
I'm fully informed about this area of research. Including other research that found simple linear methods could also get good results over a large number of games and DeepMind's recent work where far less computationally involved
226.
by Cybiote 10y ago
That doesn't make sense. The brain is highly interconnected and there's no such thing as a "Go" or poker area. Many of the same structures are recruited for different tasks. For example, the part for recognizing faces wi
227.
by Cybiote 10y ago
The importance and whether there is any function of the column is not without controversy, it may well be an artifact of what tools were used in the early days. The brain is very flexible in its organization and attempts at organizational s
228.
by Cybiote 10y ago
I say that the computations in V1 and the computations in the CNN share a fair bit in common. However, I doubt that the layers in V1 and CNN have much in common for a couple reasons. First is because the layers in V1 will differ depending
229.
by Cybiote 10y ago
As I point out in another post (currently above) the layers of Neural nets have nothing in common with those in the brain. Only confusion will be met in heading down that path. What is had in common, and what was inspirational were (harsh)
230.
by Cybiote 10y ago
It's not misleading because they were careful to point out that these 6 'layers' have nothing in common with how layers are defined in Neural networks. In Neural nets, layers are effectively subgraphs that are bipartite. They
231.
by Cybiote 10y ago
That is a good explanation. I'll point out for anyone that clicks through, that unlike what he says, depth in ANNs do serve a purpose: to make learning more tractable by building representational abstractions. There are feedback and re
232.
by Cybiote 10y ago
Some of these functions are easy to compose to get something impressive, true. But that is not often the case. Take the case for games and imagine we wanted a meta-algorithm to select an algorithm to apply to each game. The intelligence wou
233.
by Cybiote 10y ago
> Atari games just from the raw pixels on the screen It's important to distinguish between what sorts of games work well under this method and what sorts do not. Games that are variations of pole balancing, like Pong, fare better th
234.
by Cybiote 10y ago
It doesn't need to be a fight. We can generalize to competitions, where we have examples such as athletes and gymnasts (who in particular are amazing). But you can also look at performers outside pure competitions too: acrobats and dan
235.
by Cybiote 10y ago
But it and methods like it are still very limited in what they can learn. For example, they can't learn relations involving antonyms. They can't tell apart hot from cold or big from small.
236.
by Cybiote 10y ago
I agree that AI is possible from text alone but only with the added stipulation that full understanding of text requires a very sophisticated learner. In order to predict what a human will say next, you need the ability to maintain extended
237.
by Cybiote 10y ago
It more looks to me that Turing was saying that the essence of the brain is not from its using electricity. Nor is it valid to say that computers and brains are alike because they both operate with electricity; instead, their essential simi
238.
by Cybiote 10y ago
I don't think that is quite right. This says it's applying basic ideas from program synthesis, which is not too distant from automated theorem proving style AI (specially when you get to SMT solver based synthesis). From skimming
239.
by Cybiote 10y ago
I think that sort of solution would work better as an educational tool or as part of an augmented programming environment. But I don't think it fits within the spirit of the task. It also would hit a ceiling really early on, if the par
240.
by Cybiote 10y ago
#2 Is incredibly ambitious but I expect it should be doable by a lone grad student over the course of a summer. Something like this would range in complexity from completely redefining programming languages (drastically lowering the skill r

More ›