These images combine photography and computer generated elements in an effort to visualise the output of various mines in South Africa. The CGI objects represent scale models of the materials removed from the ground. By doing so, the intention is to create a kind of visualisation of the merits and shortfalls of this industry that has shaped the history and economy of the country so radically.
Via The Prepared,.
Backpropagation of error (backprop) is a powerful algorithm for training machine learning architectures through end-to-end differentiation. However, backprop is often criticised for lacking biological plausibility. Recently, it has been shown that backprop in multilayer-perceptrons (MLPs) can be approximated using predictive coding, a biologically-plausible process theory of cortical computation which relies only on local and Hebbian updates. The power of backprop, however, lies not in its instantiation in MLPs, but rather in the concept of automatic differentiation which allows for the optimisation of any differentiable program expressed as a computation graph. Here, we demonstrate that predictive coding converges asymptotically (and in practice rapidly) to exact backprop gradients on arbitrary computation graphs using only local learning rules. We apply this result to develop a straightforward strategy to translate core machine learning architectures into their predictive coding equivalents. We construct predictive coding CNNs, RNNs, and the more complex LSTMs, which include a non-layer-like branching internal graph structure and multiplicative interactions. Our models perform equivalently to backprop on challenging machine learning benchmarks, while utilising only local and (mostly) Hebbian plasticity. Our method raises the potential that standard machine learning algorithms could in principle be directly implemented in neural circuitry, and may also contribute to the development of completely distributed neuromorphic architectures.
It’s easy to think that arguments have just three terminal truth values: right, maybe, and wrong. In practice, arguments (and in particular, the sort of argument that we use to justify actions) have many possible truth values. These include things like ‘got the details wrong, but is by-and-large correct’, or ‘is correct but for a different level of abstraction ; doesn’t apply here’, or ‘is partially correct, but isn’t as useful compared to a different framing of things.’
people with limited understanding of business think that business is all about making profits. But those who actually run businesses know that running a business is all about managing cash flows
Many of the games you play in business is centred around managing cash flow. This might seem bizarre — because a cash flow game is essentially taking the money you make, and moving it back and forth in time. Once you get this, however, you’ll understand that raising capital is merely one instance of this game. Which probably means that if you want to reason from first principles, you’ll have to start with cash flow in mind.
Incredible, the museum experience at home (even better?).
/The consequences of intervention are rarely what we expect or desire./
Kibbutz is bottom‐ up socialism on the scale of a small community. It thereby avoids the worst problems of state socialism: a planned economy and totalitarianism. The kibbutz, as a unit, is part of a market economy, and membership is voluntary: you can leave at any time. This is “socialism with a human face” — as good as it gets. Being a member of a kibbutz taught me two important facts about socialism. The first is that material equality does not bring happiness. The differences in our material circumstances were indeed minimal. Apartments, for example, if not identical, were very similar. Nonetheless, a member assigned to an apartment that was a little smaller or a little older than someone else’s would be highly resentful. Partly, this was because a person’s ability to discern differences grows as the differences become smaller. But largely it was because what we received was assigned rather than earned. It turns out that how you get stuff matters no less than what you get. The second thing I learned from my experience of socialism was that incentives matter. On a kibbutz, there is no material incentive for effort and not much incentive of any kind. There are two kinds of people who have no problem with this: deadbeats and saints. When a group joined a kibbutz, the deadbeats and saints tended to stay while the others eventually left. I left. In retrospect, I should have known right away, from my first day, that something was wrong with utopia. On my arrival, I was struck by the fact that the pantry of the communal kitchen was locked.
Progressivism rests on two critical assumptions. The first is that we know how to improve society: “social science” provides us with a reliable basis for the necessary social engineering. The second critical assumption is that government is a suitable instrument for improving society. My second and third lessons taught me that these two critical assumptions were unfounded and unrealistic.
A small group of postdocs, graduate students, and undergraduates inadvertently formed a longitudinal study contrasting expected productivity levels with actual productivity levels. Over the last nine months, our group self-reported 559 tasks, dates, and completion times — expected and actual. Here, I show which types of tasks we are the worst at completing in the originally planned amount of time (spoiler: coding and writing tasks), whether more senior researchers have more accurate expectations (spoiler: not much), and whether our expectations improve with time (spoiler: only a little).
Overall, we complete 53% of our weekly tasks in the originally planned week. The actual number of hours required to complete a task is, on average, 1.7x as many hours as expected
Coding / writing take longest (compared to expectation).
This site is about my experience with the Wirth school of languages, based on the ideas and implementations of Prof Niklaus Wirth, Kenneth Bowles, Per Brinch Hansen, colleagues, and their students. And my experience with the various variants, from the P2 and P4 compilers originating in Zürich ETH, via UCSD Pascal P-System to the Borland compilers and Modula and Oberon systems. All applicable to small computers and device control.
On this website you will find information on Pascal for small machines, like Wirth compilers, the UCSD Pascal system, many scanned books and other files on UCSD Pascal, Pascal on MSX and CP/M, Delphi programming on PC, Freepascal and lazarus on Windows and Raspberry Pi, Oberon systems. Many sources of early Pascal compilers!
Anti-portfolio is a great idea.
Most people use Einstein as the canonical example of a super-smart person. The smartest people I know tend toward von Neumann. Runner up is probably Grothendiek.
Layout Parser: I’ve had a few projects I gave up on because OCR isn’t good. This tool might change that.