Lazy Sunday #6

2024 is definitely going to be an interesting year. While Sam Altman wants to raise a couple of trillions to build more AI, Gates and Bezos are going on an investment run together again, slightly nimbler ticket size though. And talking ticket size, can LLM predict how successful founders can be, so the VC’s know how much money to put where?

I have to get one thing of my mind: Elon Musk’s prominence helps him incredibly to put NeuraLink on each and every bloody news page in the world. Even Putin talked about it in his interview with Tucker Carlson. Looking at it from the side lines, it sounds like Neuralink is the first and the best and only company. It is good to have more attention on BCI technology, it creates a discourse on possibilities, limitations and risks. But there is so much more to the industry than just Musk. I hope you grok this (pun intended!).

Let’s go!

Find my research

Can AI predict founder success?

Once, data analytics was reserved for hard data: data that could be tracked offline or online. It was more aimed at hard skills: sales target achieved or not? User growth achieved or not? CTR up or down?

But more and more analytics and prediction models creep into the “soft skills”. Can AI (or in this case LLM) predict founder success based on psychological attributes.

(screenshotted from website on 11.02.2024)

I don’t mind this exercise, as typically there are a set of traits in all combinations and permutations that can make a person a successful founder. But there is a high degree of bias creation if the model is not fed with the proper variety of training data. What are your thoughts? (Basisset)

Negative stories have impact on performance - with surprising results

The labour market, health emergencies, AI will take over jobs - the news is full of, well, bad news.

A study has looked into the impact of bad news on the performance of students.

On the one hand, negative stories, like the decline of the labour market (signal that finding a job is more difficult and challenging), can have a positive, motivating effect on students. Seen as a challenge to overcome, a problem to solve.

But at the same time, the same story can have a negative impact on motivation, especially it seems, when e.g. the financial background of the student is less secured. Instead of getting an extra boost in motivation, this group of students feel more stressed, which in return results in less action. This can be a dangerous spiral into depression and other health issues.

How is this important to you?

Firstly, external factors easily impact our overall state. Keep this in mind when communicating any kind of news to your team, especially if you want to make complex decisions right afterwards.

Secondly, considering all the lay-offs across various industries, there might be a group of people who need extra help and attention. For some a lay-off might be a motivator at the end, for others it might be the point of breakage and need for help. I think this is significant to consider for HR processes, there might be some HR teams, who despite lay-off’s, actually care about the people impacted. I very much believe so. (Nature)

Marketers - be aware of ChatGPT biases

I am not going to go into details of ChatGPT (and similar) LLM models’ boom. They are here, they are here to stay. If you haven’t tried any of them, I urge you to do so.

GPT-like products are used in marketing a lot. Most prompts offered are about writing content, rewriting content etc. If you have played around with it, you have probably come across the somewhat generalized answers these models exhibit. Even after pressing for a nuanced rewrite of the content, it is hard to extract a more creative answer or a more detail-oriented argumentation from these algo’s.

As it seems, GPT models are biased to a more high-level way of reasoning and decision-making. By the way, this is not limited to ChatGPT, as Bard - now Gemini - shows very similar traits.

Let me give you an example: if you decide to buy a product (e.g. software), you look over the general features you require (high-level construals = less detailed). After that, you narrow down feature sets and buying criteria, you start to compare details (low-level construals = more details) which ultimately impact your readiness to pay a higher or lower price. If a product has less of the detailed features you require, despite the fact it ticked the boxes on all the high-level features, it might fall out of favour. Interestingly, a study found that sellers will sell mostly on high-level features, while buyers will decide based on low-level features.

Coming back to GPT models: they are biased to high-level content production. This might be great for overall marketing direction, but will miss the level of detail buyers would want.

Two learnings for backed by science:

1. Bias is persistent and prominent in LLM’s. Just something to be aware of.

2. LLM’s are good for initial framing and directive work, but miss the level of detail to make content human. Consider this for your workflows.

(ScienceDirect)

A review of customer reviews

If you launch a product today, you can’t get around customer reviews. Word of mouth in all ears.

I dug into the internets to find studies. Of course there are! Short summary:

They work! Yes, you might know that, but: longer reviews have higher credibility for more complex products. Shorter reviews are more credible for less complex products. Meaning you got a complex product (e.g. b2b software) you might want to implement a minimum length.

Product reviews that divert from the average rating are perceived as less credible, although they might be true. Bandwagon effect - that can go both ways, by the way. (Research Paper)

On top of tech

AI uses smartphone keyboard usage to detect mental illnesses

I love technology that derives novel information from existing sources. A new app detects the psychological state of a keyboard user from the usage patterns applying - you guessed it - AI.

It aims to detect depression or even suicidal thoughts. (YR Media)

BCI technology as open source

As we know, most new tech is expensive. Thank the universe, the cost of computers came down from its initial $450,000 price tag. Apple Vision Pro anyone? Go somewhere near medicine, and it becomes even more 💰 : a fMRI machine can cost somewhere from 500,000 to 1,000,000 dollars in 2024.

Hence, it is paramount to find new ways to reduce cost and make tech more affordable. Enter OpenWater.

The company is led by technologist (ex Facebook, Oculus, Google) Dr. Mary Lou Jepsen. It provides neuromodulation and tissue scan technology (e.g. tumour detection) powered by light. It is a little more complex than shining a light on your forehead, but they de-complexed the setup significantly to make it light and cheap and decided to put all of it online as open source, while collecting 54 million dollars in funding. (Openwater / OpenWater Github)

Kernel and AppliedVR to reduce pain with VR

I am not a big fan of VR per se. I haven’t seen any meaningful use cases, that look good and realistic to trick the brain out of the feeling of wearing a cheap computer right on your face. But, as we know, even mice get VR headsets, so here we go.

There is Flintworks, who develops immersive clinical tools to treat e.g. PTSD for veterans. Really useful and shows great results in Australia.

And there is AppliedVR in collaboration with Kernel, which uses VR for immersive environments to treat pain perception. You know your knee pain or chronic lower back pain? Might be a thing of the past after some VR sessions.

But, and here is the interesting bit: while the VR headset provides the immersive technology, Kernel provides a helmet like device, that shines light on your skull to detect blood flow through your brain and hence the reaction of your brain to the VR content. The light technology follows a very similar trajectory as OpenWater in our story above.

By the way, Kernel was founded by Bryan Johnson, whom you might know, because he is the guy who is 46 and wants to look like 18. He also has a very fascinating podcast with Lex Friedman.

Google him. (Kernel / PRNewsWire)

Gates and Bezos do it again - Elemind gets a 12 million dollar funding

Bill and Jeff have a couple of things in common: they have both founded two of the biggest tech companies in the world. They are both divorced. They are both billionaires. They invest in neurotech. In 2023, they invested 75 million in total in Synchron. Long-time readers should know the BCI company, that kicks Elon Musk’s Neurolink (three billionaires in one paragraph!!).

Now they did it again and invested 12 million in Elemind, a neurotech company, powered by artificial intelligence algo’s to analyse brain activity from noninvasive devices.

What strikes me here, is that the company is definitely software first, with a lot of AI. Read their Manifesto (oh lord, I hate this word), it is interesting.

And the list of participating investors reads like a candy shop:

(BusinessWire / Elemind)

Coin sized device to reduce Parkinson’s symptoms

Big investments, small devices. In 2021, Charco Neurotech attracted 10 million dollars. It was Europe’s largest seed round in 2021.

Since then the company develops a coin sized device that costs approx. 400 dollars, which sits on the patients' skin and vibrates in a particular rhythm when required, reducing Parkinson’s symptoms. The company distributed the device to 2,000 people with a waiting list of 20,000 around the world. Great write-up on “Wired”. (Wired)

Misc but not least…

When somebody asks you to double-click on that specific topic during the next all hands…

Thank you all for reading again. It has been really fun and exciting for me to put these newsletters together.

I hope you learn something new every now and then. Let me know if there is something to change, add or remove at all.

If you think somebody might enjoy it, send them this link: https://bit.ly/42yxMBq

Until next week.

Alex