To me the DCN picture looks ok, in convolution layer nodes receive input only from small area of the previous layer, in general this is true for convolution layers (kernels) and pooling layers. After that it has couple of fully connected layers where the connections go from every node to every node. Or have I misunderstood something in your comment?
An investor can get value from a company in other ways than net profit. For instance share buybacks, dividends, mergers and acquisitions, companies can pay back capital. P/E ratio is just one very simple number which is not applicable to all situations. For example REITs have often very high P/E because most of their profit is distributed as dividends. Using P/E to try to value Amazon is in my opinion misguided.
Revenue is the top line, net profit is the bottom line. There are many ways to get money out of a company which come before the net profit line in an income statement. Companies can also spin off parts of them in many different ways. The only thing I'm stating that P/E is often the wrong number to look at in a company.
Yes, I have played it. I think it's a good game. It has good controls, good music, nice "flow" to the game. It's not very deep game, but fun to play nevertheless. I heard someone describe it as the best restaurant simulator game available.
There are many people who have at least that sum of money and they appear in public all the time. I have heard this same thing about "him surely getting robbed or killed" several times now, it feels always a bit odd to me.
Isn't that part of the author's point? That Kurzweil is picking and choosing exponential developments because they fit his narrative? Since none of us know what the requirements of the Singularity are, it seems odd to extrapolate from a single abstract indicator and assume that it will happen "somewhere" on that line.
What happens when performance approaches infinity? Many AI problems have been solved because there is now simply more processing power available. For example computer Go using Monte Carlo methods. Just a couple of years ago people thought it was impossible for computers to beat human professionals.
I stand corrected, no official even game wins against professionals. Maybe I remember seeing unofficial games on KGS. The progress is impressive anyway, just a while ago the best computer programs were around 8-10 kyu level.
Usually it's the other way around. By law normal investors have to be educated about all possible risks involved. Accredited investors don't need such information since they supposedly know those things already.
