Private investors tend to have too small equity allocations in their portfolios And we can only fix this through better financial education 🧵

A bit of a niche topic. But if you by any chance want to train an image segmentation model on satellite image data and your labels are given as vector format polygons (and not as labels in image pixel space) then this might be for you ;-) https://t.co/kN9ZixV3iR

Hoping that it might help other people discover something useful that they did not know yet, I created a list of my favorite tools and services:

If you are interested, you can find English and German transcripts here: https://t.co/yxylk6Q1Tw

Using this model I now could easily get the transcript for the podcast and translate it into English with the help of another awesome AI tool: DeepL.

With the advance of #ai and improved speech-to-text models I recently did build myself an automated podcast transcription tool that automatically generates a transcription for a given audio file (further details on this are a topic for another day).

A few years ago I did a #podcast on #assetmanagement and #automation . Back then, it somewhat bothered me that the podcast was basically useless to most people because it was in German and not in English.

Let me know if you managed to create a better result :-)

My first real-world challenge was not as successful as the mad scientist images. My three year old son wanted me to create an elephant that has the fur of a giraffe. Here is what I came up with:

Whatever image that you want to generate, you need to describe your image in a way that the AI model can understand what you want to achieve. This is actually not that straightforward (for the mad scientists I used a text input that I copied from https://t.co/sbXDHIN55X)

You can either create an image from a text input. Or use an additional "negative prompt" to exclude certain image motives in order to refine your result. Or use "image inpainting" to modify an existing image.

I wrote a blog post on text to image AI models: https://t.co/cvdKZVV6x9 It mentions three different publicly available pages that you can use to get creative yourself (for free), so that you can create your own fantasy mad scientists:

Roughly half of the differences can be explained through differences in local inflation rates ( https://t.co/DnXN4ppDPm)

Although in somewhat differing magnitudes ( https://t.co/DnXN4p8ANm)

Central banks have moved pretty unanimously in the same direction with policy rate decisions in the last months ( https://t.co/K4U0LqSvK8)

so true 😄

If the company can be bought for 100$, then we get 5% return. If the market price was 200$, however, then the return would be only 2.5%. The market prices that we pay are largely determining the returns that we will make on our investments 6/6

Another important take-away of this is: we can basically make the expected return of an investment anything that we want, depending on the market prices that we jointly set 5/6

Any economic growth would come on top of the 5%. If profits were increasing (from 5$ to 5.2$ to 5.4$ ...) then this would also lead to higher returns on the investment 4/6

If you can buy this company for 100$, then the return on your investment will be 5% each year (5$ divided by 100$). The investment will grow by 5% even though the economic fundamentals do not grow at all 3/6

Even without any economic growth stock markets can still generate positive returns. Example: Let's assume some company is generating 5$ every year in profits/earnings forever. In other words: the company is profitable, but its output is not growing 2/6