Hey I'm a data scientist with 5 years of experience building predictive models. I'm trying to start my own side-consulting company specializing in scoping machine learning projects and building PoCs and MVPs. I think I may be able to help you here. If you are interested in connecting, please look me up on Linkedin: https://www.linkedin.com/in/acauthon
Last year around Christmas time I was talking to my wife about what candy I wanted in my stocking. I said something like, "I really like those gold ball candies. I can't remember their name. Maybe ranchero something?"
Then less than an hour later I had an ad on my phone for ferrero rocher candy. It was incredibly creepy.
I know supposedly the Google phone and Google home (in my case a model from Lenovo) doesn't parse your conversation for ads, but I have no other explanation.
That’s incredibly creepy! I get the need for advertising but the sliminess has gone too far. I know it’s silly but I feel like there’s space right now for an ad company with morals that they can prove :)
I think there is a huge risk that the data being collected just isn't meaningful to the problem being asked.
For example, if you are trying to predict the weather you'd really like information on how big clouds are moving but all you have is wind speed and humidity.
Everyone acts like having fancier machine learning methods will solve their problem, but often the data just isn't good enough and getting better data is impossible.
> You can do things that don't scale to accelerate a good idea, but by definition you actually can't do that if the idea itself requires scale. The idea has to work on day 1, for customer 1.
He's talking about ideas that only work if the company has scaled to some non-trivial size of users or some other relevant factor. Uber would be an example. How could Uber work if there was only 1 customer? There would be no drivers.
If you need to "do things that don't scale" to launch your idea, your idea probably also needs to work with low scale of users/revenue/investment or you will be dead in the water.
Yes I'd call it a network effect, and the author does as well later in the article. The point is that the network effect is required for the idea to work. It doesn't have to be a network effect though. Other examples he gives:
> Third, how synchronous and mission critical is it? Be skeptical if it going down would cause interruption to workflows that couldn't be worked around or deferred. Be incredibly skeptical if this is true round the clock and on weekends. That kind of service level implies significant and robust automation and support, which require scale.
> Fourth, how much does the business model depend on volume? Losing a bit of money on first customers as you bootstrap and learn is not an issue. Be skeptical, though, if this would need to be sustained to bootstrap your way to some required volume which is quite a way beyond those first few customers.
Generally I also think of technical scale first, but I think using it more generally here is valid.
This is how you end up with students who are not prepared for the workforce because they haven't been trained on the tools that businesses actually use.
There is nothing ethically wrong with asking students to purchase proprietary services, but it is tone deaf to make them buy absurdly expensive software.
So the depositor put $1,000,000 in the bank and the bank loans $850,000 to a small business so it can buy more inventory. The small business goes to the widget manufacturer and writes a check which the manufacturer deposits into the bank. So now the bank has 1,850,000 in deposits and 850,000 in loans.
The bank takes the new deposits and loans out 85% of it ($723k) to another small business. This small business goes and uses it to pay its employees and they all deposit the money in the bank. Now the bank has $2.573m in deposits.
The employees and the manufacturer can all withdraw the money at any time. So I struggle to see how this isn't creating money.
The fractional reserve system lets the bank loan the same dollar out multiple times. How isn't this creating dollars? If the bank takes one dollar and loans 85 cents to you and 73 cents to me, isn't there more money?
> The employees and the manufacturer can all withdraw the money at any time
This isn't quite correct. If all parties go to withdraw their money at the same time, the bank will not be able to give it, because they only actually have $1M in reserve.
In practice, if there are more withdrawals than than the bank has in reserve, it is usually able to get short term loans from the central bank which than ACTUALLY creates the money and loans it to the bank.
However, over the long-term if too many of the bank's loans default, they won't have enough assets to cover the deposits, which is how banks usually fail.
And you might think, "Well, there are so many banks, that money probably won't end up back at mine," but yes, there are so many banks, all of them giving out loans. My million ends up all over town and the world, but so does yours.
Because doing it one time, or 10 times, or 1,000 times doesn't change anything about how it works. If everybody pays all their debts, it all adds back up to $1,000,000 in cash (plus interest for the bank(s)).
No it doesn't. In my example the bank loaned out $850k to the first small business and $723k to the second. That's $1,573,000. The bank started with $1million.
It absolutely matters how many times the dollars come back to the bank because that is the pathway that lets the bank loan the same original dollar out multiple times.
I think you’re forgetting the bit where the business that took the $850,000 loan is left with $850,000 debt and $0 balance after they spend all of it. Every time it passes through the system the amount recirculated simply decreases by the reserve %. It all still adds up to the original amount (plus interest).
In your example though, the bank has a different problem of offering unsecured loans, which could lead to some losses for them.
He doesn't literally mean country clubs. He is saying that the fed's current policies allow rich corporations (controlled by rich people) to get loans at very low interest rates.
This is an over-simplified view of the market. The market isn't a zero-sum game. It's often in an investor's best interests to publish their research on a position (after they have taken it) to convince others to jump in the trade with them. It doesn't do you any good to find a diamond in the rough if no one else knows about it.
It is a minefield though and you should never 100% trust one source on any given investment, but to act like there is some kind of mathematical truth that no one ever publishes any useful stock analysis is just silly.
I think this could be fixed by making each job be posted for a fixed block of time, say five minutes. Then it is randomly assigned to one of the shoppers that selected that job. This would remove the incentive around reaction speed and I don't think the customer would care about the slightly increased time for an order.
I was here to say the same thing. But you don't even need to extend the time frame to five minutes to defeat the bots. Lisa Marsh, in the article, complained "no human can click that fast." I bet she only needs 10-15 seconds to assess an order.
Not for nothing, this is the same mechanism by which the "unfair" advantages of HFT practitioners can be defeated. Run mini-auctions every whole second and randomly allocate fills if there are ties for best bid and/or best offer.
Of course, as long as the exchanges deeply care about vanity metrics such as trading volume (trading volume !== liquidity) such measures will never be broadly enacted.
If my company could lower the cost of its product by half and still make the same yearly revenue, it would. And all the executives would get huge bonuses.
