How to Remember What You Read - Farnam Street
Farnam Street | 20 mins
The benefits of reading are negated if you don't remember what you read. This article discusses a tested system to increase retention.
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
Arxiv | 32 mins
We introduce a new language representation model called BERT, which stands for Bidirectional Encoder Representations from Transformers. Unlike recent language representation models Peters et al. (2018); Radford et al. …
A Few Beliefs
Collaborative Fund | 3 mins
Having your views confirmed is a powerful and addictive drug. The worst financial decisions happen when people risk what they need in order to gain something they merely want. Unsustainable things can last years or decades longer than people think. Tell people what they want to hear and you can be wrong indefinitely without penalty. Being loyal to people who deserve your loyalty is a wonderful thing. The luckier you are the nicer you should be. Past performance increases confidence more than ability. Define what you’re incapable of and stay away from it. You’re not proven until you’ve survived a calamity. Read fewer forecasts and more history. Many bad investment decisions are good ideas taken too far. Spending money to show people how much money you have is the fastest way to have less money. The most reasonable plan is planning on your plan not going according to plan. A lot of denial masquerades as patience. A lot of people have a hard time distinguishing between what happened and what they think should have happened given their world view. About once a decade people forget that bubbles form and burst about once a decade. The best way to deal with uncertainty without hiding in a bunker is to save like a pessimist and invest like an optimist. With the right incentives, people can be led to believe and defend almost anything. Expectations move slower than reality on the ground, so a lot of frustration comes from clinging to the trends of past eras. Most people can afford to not be a great investor but they can’t afford to be a bad one. Beliefs that haven’t changed in 20 years are as suspect as beliefs that changed overnight. Progress happens too slowly to notice, setbacks happen too fast to ignore. We are extrapolating machines in a world where nothing too good or too bad lasts indefinitely Optimism and pessimism always overshoot because the only way to know the boundaries of either is to go a little bit past them. The world is governed by probability, but people think in black and white, right or wrong – did it happen or did it not? – because it’s easier. It’s hard to empathize with other people’s beliefs if they’ve experienced parts of the world you have not. The cure to overconfidence is constantly reminding yourself that you’ve experienced maybe 0.00001% of the world. The line between bold and reckless is thin and often only visible with hindsight, so be careful who you admire and what you admire them for. Something can be factually true but contextually nonsense. Being nice to people is the easiest career competitive advantage. Reputations have momentum in both directions, because people want to associate with winners and avoid losers. If your expectations grow faster than your money you’ll never be happy no matter how much you accumulate. People who are exceptionally good at one thing tend to be exceptionally bad at other things. Not caring about temporary things, and obsessing over permanent things, is underrated. Good marketing wins in the short run and good products win in the long run. When reading the news, you can increase your productivity 100x by asking, “Will I still care about this a year from now?” Uncertainty amid danger feels awful. So it’s comforting to have strong opinions even if you have no idea what you’re talking about. Daniel Kahneman says a key to investing is having “a well-calibrated sense of your future regret,” which might actually be the key to understanding all forms of risk. Debt removes options, savings add them. People like weekends because it’s when they have the most control over their time; financial goals should keep this in mind. It’s easiest to convince people that you’re special if they don’t know you well enough to see all the ways you’re not.
Attention Is All You Need
Arxiv | 21 mins
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks that include an encoder and a decoder. The best performing models also connect the encoder and decoder through a…